Phil Arena

Miscellaneous Links

2012-03-01T19:57:00.002-05:00

Below the fold.

1. Excellent NYTimes piece on the price of gas and the impact of the crisis with Iran. I'm so used to criticizing the NYTimes, I'm a bit in shock, but there you have it. It really is a good piece. It does a nice job of describing the tough line the Obama administration must walk, with sanctions showing real promise, but the price of gas threatening to become a political liability.

2. Iran and strategic ambiguity. This is something I've been thinking about for a while, and had flirted with the idea of writing a paper about. More transparency can lead to a higher probability of war, which is broadly consistent with what Wolford and I have argued about intelligence.

3. Misperceptions, foreign policy and Iran. Like much of the commentary on the importance of psychology to IR, I mostly agree with the substantive point, but I can't help being irritated by the "you can't understand the world if you're not thinking this way" tone, given that so much of what is being said could easily be reconciled with an incomplete information model of "rational" unitary actors. Read Kydd's book and then come back and tell me that there is no way to understand the role of perception without turning to psychology.

4. Two former Senators suggest direct link between Saudia Arabia and 9/11. Very telling that the government is trying to prevent questions from even being raised. Here's a question I'd love to hear the left answer: how can oil be the reason the US went to war with Iraq when it is so transparently the reason the US will not go to war with the Saudis, and will help Bahrain put down democratic protests when the Saudis request such? (There's also the small matter of most of the contracts to develop oil fields in Iraq going to Chinese companies, while only one went to US companies, but who cares about facts anyway).

5. Cats and mice, regimes and oppositions. Jay Ulfelder on the risks of explaining protest movements in isolation. Very sobering.

6. Could the ICC help bring about a political solution in Syria? I wish my thinking was this sophisticated when I was an undergrad. If you aren't already, you really should be reading Anton's blog. Incidentally, if you're interested in game theoretic models of bargaining between dictators and the ICC, I suggest you read this paper (forthcoming at JTP) as well as this working paper, both by Emily Ritter and Scott Wolford.

7. Who says money can't buy happiness? Of course, it's observational data, based on self-reports. This is no smoking gun. But I'm not sure I see the story for OVB here the way I do with the democratic peace or negative advertising or military buildups or other issues where I've been more skeptical.

8. "Blogging is quite simply one of the most important things that an academic should be doing right now." Very much agreed. The publication process is broken. I've had papers accepted that had worse reviews than papers that keep getting rejected (and just so happen to advance arguments people don't want to hear). The idea that 2 not-so-randomly selected peers should have the right to decide whether the rest of the discipline should even be exposed to one's argument is absolutely insane. The wonderful thing about blogs is that everyone can decide for themselves whether an argument is persuasive.

9a. Increased enrollment has decreased the average wage premium associated with a bachelor's. Which shouldn't surprise us, since the primary benefit of a college degree probably has a lot to do with signaling. And b. a straightforward extension of the classic Spence model provides a pure signaling explanation of why the wages of high school graduates are decreasing. Alternatively, c. a college degree might not lead employers to update about the abilities of potential employees so much as remove uncertainty. If that's the case, we should just require all high school seniors to take an exit exam, whether they intend to go to college or not, and allow that to do at least some of the work of reducing uncertainty for employers.

10. How politicians view the stockmarket. And strong partisans too.

February Roundup

2012-02-29T11:01:00.000-05:00

February's most viewed posts:

1. Zombies, Negative Advertising, and Military Buildups.

2. Review of Clarke and Primo.

3. R2P and Consistency.

4. The Pacifying Effects of Aircraft Carriers?

5. The Political Impact of Audience Costs.

Democracy: The Devil is in the Details

2012-02-26T23:51:00.008-05:00

Over at Howl at Pluto, LFC is critical of how much time I spend arguing against what others say about how democracy influences international relations.

Maybe Phil could consider taking an occasional break from criticizing...and focus on the particular forces that drive bad, suboptimal policy in the particular democracy known as the United States.

Kindred at IPE@UNC comes to my defense, arguing that

a regular theme is that [Phil] is skeptical of claims in the academic literature that "democracy" does this or "democracy" does that...In other words, I imagine Phil would be fine with the bureaucratic-politics-plus-interest-groups story that LFC puts forward, and nothing he wrote contradicts it.

Kindred is exactly right about what I'm trying to accomplish, and that my attempts to discredit the prevailing view of democracy is not in any way incompatible with the more nuanced view LFC suggests.

But, in LFC's defense, I think it's reasonable to ask me to spend a bit more time focusing on what I do believe, and perhaps a little less on what I don't.

I'm far more confident in my claims about the deficiencies of extant approaches than I am in my own views of how things do work. I sometimes make careless statements that I later wish I could retract even with respect to the former (and I do hope that Brenton is aware that I appreciate it when he keeps me in check), but even so. My views about why the romantic narrative we'd all like to believe is true doesn't seem to actually fit the world we live in much of the time continues to evolve. And though I don't always succeed, I generally try to minimize the probability of saying things on here that I'll later read and say to myself, "what the heck were you thinking?"

That said, I do of course think it is important to address such questions, and I'm working on precisely those questions as we speak.

With the obvious caveats then that I may or may not still believe this six months from now, here's an overview of my current thinking on how domestic politics influences international politics in democracies.

1. The assumption that leaders first and foremost want to remain in office is a valuable assumption.

Like all assumptions, it is a simplification, and I wouldn't claim that it is always true. But this is definitely not where my gripes with the extant literature lie.

2. The assumption that the public closely monitors the foreign policy choices of the government, has coherent preferences on the finest detail of trade talks, and can credibly threaten to defect to the other party if the leader deviates from their foreign policy agenda in the slightest degree is really problematic.

Assumptions don't need to be true to be useful. If relaxing them would only complicate the math, without actually pushing our results in a different direction, then I'm perfectly comfortably making them, no matter how unrealistic they are. This assumption strikes me as wildly unrealistic, and not in an innocuous way.

3. Election outcomes largely -- though certainly not exclusively -- turn on the state of the economy.

That's an equilibrium outcome. If parties fielded exceptionally poor candidates or staked out sufficiently unpopular positions or failed to raise adequate funds, that would be a different story. But provided the major parties are all behaving the way we tend to expect them to, there is a serious upper bound to the effect foreign policy is going to have on an election.

4. The big exception is for wars. Especially ongoing wars.

This is well recognized in American politics, where some of the best forecasting models include no other variables other than those that reflect the state of the economy and the impact of the war (see here). But this is not just true in the US. The Labor Party (and its forerunners, like Mapai and Alignment) was continuously in power for nearly 30 years in Israel (from 1948 to 1977). When it finally lost power, it was because of the 1973 Yom Kippur War. Similarly, the Indian National Congress Party was in power from 1947 to 1977, and lost to the diverse coalition operating under the name of the Janata Party because Indira Ghandi declared a state of emergency in response to accusations of fraud during the 1971 election, in which the Second Indo-Pakistani War probably should have seen the defeat of the Congress party.

We can go on and on. Wars matter for elections. I don't think anyone doubts that. But I'm not sure how much foreign policy influences elections during peacetime. I won't go so far as to say it plays no role. When the fundamentals are such that a tight election is expected, the positions the parties take and retrospective evaluations of the incumbent's handling of foreign policy will almost certainly play a role. But I think more IR scholars need to recognize that in many elections, not just in the US but generally, the fundamentals don't predict a competitive election.

5. The public does not abhor war. Nor are they necessarily all that attentive to objective information about the progress of the war.

It is overly simplistic to assume that leaders of democracies will avoid wars they cannot win quickly and easily. As I argue in a few of my papers, and as others have argued as well, the link between accountability and war outcomes is more complicated than that. Leaders can shift costs onto a minority of the population. The opposition can influence the degree to which the public perceives the war as a political issue. The media can influence how people believe the war is going. For various reasons, it is possible for democratic states to fight costly wars that they are unlikely to win without being punished for it. It is also possible that they'd be punished if they didn't.

6. Special interests will find it difficult to pressure governments into adopting policies that will be unpopular with the public, provided such policies are highly salient. But they will often find it easy to pressure government into adopting policies that harm the public when the effects of these policies are not highly visible. They will also find it relatively easy to block legislation that would threaten their interests, provided the public never knew such legislation had been proposed.

This has a lot of implications, both for domestic and international politics. It speaks to the debate about threat inflation and defense spending. It helps explain why Americans sweeten everything with corn syrup, rather than, you know, sugar. It gives some insight into USFP towards Cuba and Israel.

But I'm skeptical of stories that link powerful interests to policies that are highly salient. I don't think there's any doubt that the US would be spending less on the military, trading more with Cuba, and pressing harder for an independent Palestinian state if not for the power of lobbyists. But I am far from persuaded that the best explanation for the 2003 Iraq War is that big oil and defense contracting firms expected to profit from it.

In short, good many of the suboptimal policies we observe in the United States are driven by small groups trumping the interests of large groups, in classic Olsonian style. There are limits to this though.

7. When we talk about how domestic politics influences the conflict behavior of "democracies", we're talking about a handful of states.

Most of the world's democracies would be predicted to engage in somewhere close to zero wars in any given decade even if they weren't democratic. Interstate war (the only type of war we care about when we talk about the democratic peace or the tendency of democracies to win their wars) simply isn't that common (thankfully). Most of the interstate wars in the CoW data set took place in Europe, and most of those involved the major powers. There are a great many reasons why we should not expect to see another European war any time soon. Democracy may be one of them, but observational equivalence is a bitch.

If you have a story to tell about how democracies behave during war, or when they will or won't go to war, it damn well better fit the United States, Israel and India.

8. The United States seems to behave differently than other democracies.

Israel and India do indeed tend to fight short wars. And win them. There are exceptions, but if you looked just at these states, you'd say that the dominant narrative looks okay.

But in the past 70 years or so, the United States has fought five major interstate wars, and though one of those lasted 100 hours and ended in overwhelming victory, the others are among the longest wars of the postwar era, and not one of them has indisputably ended in clear victory. The United States seems to be on a quest to behave precisely the opposite of how IR scholars think democracies behave. And yet, even though the literature I'm referencing is almost exclusively written by US scholars, no one seems to have noticed. The mind boggles.

Why does the US seem to behave so contrary to the predictions of Reiter and Stam (and many others)? Perhaps because of the unique position it occupies within the international system. Perhaps because the US, unlike many of the other democracies that are actually at risk of going to war, has a presidential system with term limits. Still another is that the United States is separated from all potential adversaries by vast oceans.

9. Democratic governance tends to improve the lives of many. But that is only because a larger winning coalition means more people get to be winners.

The claim that democratic leaders must provide public goods in order to stay in office is not just absurd, but, in a sane world, would make any privileged white male who has ever said it aloud feel deeply ashamed.

Now, obviously, I too am a privileged white male. And I don't want to "win" the argument by calling into question the motives and sensitivities of others. But come on.

The word "public good" has an actual meaning. And it is precisely that meaning that allows it to play its prominent role in selectorate theory. Make no mistake. One of the most prominent theories linking domestic politics to international conflict actually argues that leaders of democracies have no choice but to enact policies that benefit everyone. There are no winners and losers in democratic politics. Only winners and even bigger winners.

This is an absolute insult to:

African Americans
Native Americans
Irish Catholics
The Inuit
Aborigines
Palestinians
The Untouchables

and many, many others.

In politics, there are winners and there are losers. And the fates suffered by the losers have often been unspeakable. That's not just a problem of small W systems. Nor is it an unfortunate relic of some bygone era. For those outside the winning coalition, life can be downright Hobbesian right here in the good ol' US of A.

Yes, democracy means there are more winners than there would be under alternative forms of government. But it does not work for everyone, and that makes a difference.

10. When democracy "works", it ensures that the policies enacted have the support of a relatively large subset of the population. No more and no less.

This means there's room for tyranny of the majority, which speaks to the previous point about winners and losers.

But I mean more than just that. There's also room for rewarding policies that basically do not work, and do more harm than good even to those backing it. So long as voters don't realize that what they've demanded is bad policy.

Good thing we have absolutely no reason to question whether voters know what the likely welfare implications of their preferred policies are, right? Clearly, a trade agreement or immigration policy that would bring net benefits to society as a whole would not be unpopular with the public.

In other words, I think that at least some of the time, when we observe suboptimal policies being enacted by the US (or other democracies), that's evidence of the system working, not of some powerful elite subverting the will of the people.

I have a lot more to say about the topic, as you can probably imagine. But I'll stop for now. Hopefully this gives you a better sense of what I think about how domestic politics actually does play out in democratic states, and what implications this has for international relations.

Blood, Treasure, and Support for War

2012-02-26T04:59:00.001-05:00

The notion that those who pay the costs of war should be least likely to support military ventures is quite old. These days, it is often associated with Kant, though of course he was not the first to propose the idea. In broad form, this underlies much of the thinking about the democratic peace.

But it's not actually clear that we can go from the intuitive claim that those who bear the costs of war are unlikely to support it to the conclusion that democracies will abhor war.

There is, after all, significant variation in how the costs of war are distributed within democracies. And leaders of democracies appear to be very much aware that it behooves them to figure out ways to limit the human cost of war. By shielding the average voter from the worst of war's ravages, which they can do through a variety of strategies, not the least of which is avoiding conscription, leaders of democracies may be slip loose of the constraint electoral accountability supposedly provides.

No one makes this case more clearly, or more persuasively, than Jonathan Caverley. He argues that states can shift the burden of war onto higher income citizens by fielding capital-intensive militaries rather than labor-intensive ones. In his words,

[H]ow a state intends to fight is likely to influence when and how often a state will fight. In the event of conflict, a capitalized military not only results in the median voter doing less of the fighting herself, but will generally allow someone else’s resources to fund the costs of war. Wealthy but inequitable democracies with ready access to capital and military technology may be quite willing to build large militaries and initiate disputes, because
arming and war are, in the minds of the voters, cheap. (emphasis in original)

A prime example of this is the Vietnam War, as Caverley has argued elsewhere.

As I mentioned previously, an implication of this argument is that the Vietnam War should have been less popular with higher income individuals than lower income individuals. Clearly, this wasn't the case, since the war was more popular with Republicans than Democrats (at least after 1968, when it became Nixon's war). The Republicans are, after all, the party of the well-to-do.

But maybe it's not so simple.

I decided to take a look at the 1972 NES data and see if, after controlling for Party ID, which correlates with both income and support for the war and therefore might act as a confounder, there was any evidence that income accounted for variation in attitudes towards the Vietnam War.*

If I model the relationship between income and the probability of responding that the US should have stayed out of Vietnam as linear, I find that higher income is associated with support for the war (i.e., lower probability of responding that the US should have stayed out), even after controlling for Party ID.

But, then, the US didn't entirely replace labor with capital. Vietnam wasn't fought exclusively with unmanned vehicles, the way future US wars may well be. And there was a draft. As Gartner and Segura demonstrate, the war was initially just as popular with African Americans as whites, but over time, as African Americans suffered a disproportionate share of the casualties, a racial gap in support for the war opened up. So perhaps we need to make a slight adjustment to Caverley's argument. Relative to past wars, or wars fought by other states, the Vietnam War represented an effort to shift a greater share of the cost onto high income individuals. But that doesn't mean that lower income individuals weren't bearing a greater burden than other Americans.

So I decided to try a quadratic fit. And lo and behold, we find that, at least in 1972, respondents from both high income and (especially) low income households were more inclined to say that the US should have stayed out of Vietnam than were middle income voters.

This suggests that democracies can potentially build a broad base of support for war among the middle class. If the human costs of war fall disproportionately on the lower class and the financial costs of war fall disproportionately on the upper class, the median voter may well indeed feel that war is cheap.

Just a wee bit problematic for most explanations of the democratic peace, wouldn't you say?

*I created a binary DV that takes on a value of 1 if the respondent answered that the US should have stayed out of Vietnam (response category 5 for VAR 720050), a value of 0 if the responding answered that the US did the right thing in intervening in Vietnam (category 1), and is missing otherwise. I similarly dropped observations where the respondent did not provide a clear answer for the Party ID (VAR 720140) or household income (VAR 720420) variables. Note the income measure is ordinal. For details, see the NES codebook.

Miscellaneous Links

2012-02-25T16:34:00.000-05:00

Below the fold.

1a. New IAEA report says Iran is accelerating uranium enrichment, prompting rather different responses from the Israeli and b.) American intelligence communities. "There is no dispute among American, Israeli and European intelligence officials that Iran has been enriching nuclear fuel and developing some necessary infrastructure to become a nuclear power. But the Central Intelligence Agency and other intelligence agencies believe that Iran has yet to decide whether to resume a parallel program to design a nuclear warhead — a program they believe was essentially halted in 2003 and which would be necessary for Iran to build a nuclear bomb."

2. The technical details matter. Along the same lines as what Zach says here, note that a world in which Iran could move relatively quickly to test its first nuclear bomb is really not the same as one in which Iran has tested a nuclear bomb.

3. Jay Ulfelder on threat inflation. "It’s easy to see how stories of things like terrorist sleeper cells, “dirty bomb” attacks, global pandemics, and nuclear strikes from Iran would appeal to editors and consumers seeking novelty and poignancy, and it’s easy to see how those stories would, via substitution, distort our perceptions of the threat those things pose." Clearly, Jay hasn't read Reiter and Stam on how democracy promotes a healthy marketplace of ideas.

4. Staniland on Slaughter on Syria. "The Syrian government’s repression is absolutely abhorrent...but let’s not pretend that [Slaughter's] preferred strategy is something other than regime change through proxy war."

5a. NATO announces that all foreign advisers will be pulled out of their posts in Kabul. It's almost like tolerating a culture of Islamophobia makes it hard to maintain support from any local allies that just so happen to be majority Muslim. Who'd have guessed?

6. Do democracies win their wars because of a greater ability to keep secrets? (H/T Doug Gibler) I'm not sure we actually know that democracies are in fact more likely to win their wars, but Colaresi makes a fairly persuasive case, and his argument is a badly needed counter to the narrative focusing on how democracy promotes transparency. I'll have more to say about this piece later.

7. Persistent fighting and shifting power. Bob Powell does what Bob Powell does, and provides a model that can explain why fighting might persist for long periods, end with a negotiated agreement, then start back up again. Such a pattern is, he argues, particularly likely in consolidating states.

8. Sacha Baron Cohen, as Admiral General Aladeen of the Republic of Wadiya, is outraged that he was not invited to the Oscars.

When do Human Rights Treaties Reduce Repression?

2012-02-24T16:52:00.000-05:00

With the generous support of the Graduate Student Association, we were fortunate enough to have Emily Ritter come up to Buffalo this week and give a talk on that very question, based on a couple of related projects she is working on with Courtenay R. Conrad. I'm going to do my best to summarize the argument from the first paper (available here).

International human rights treaties (IHRTs) make it easier for citizens to seek legal redress for violations of their human rights. But this greater ability for a state's domestic courts to protect the rights of its citizens might have two very different effects.

First, the state is going to expect to pay a greater price for repression. The cost of repression is assumed to depend, among other things, on how powerful and effective the domestic courts were to begin with, and the extent to which commitment to an IHRT empowers the courts.

Second, anticipation of the state's increased cost for repression may encourage dissident groups to mobilize against the state at higher levels.

Of course, the more the people mobilize, the more attractive repression becomes.

So what's the net impact? Does the increased fear of litigation rein in the state? Or do IHRTs perversely lead to more repression, by encouraging the people to mobilize?

Emily and Courtenay make a persuasive case that the answer is that the former effect prevails, provided that the leader is sufficiently secure in office. Otherwise, we don't see much effect one way or the other.

The intuition is that insecure leaders don't feel as threatened by mobilization.

Wait -- insecure leaders don't feel threatened?

That's right.

Think of it this way -- if you would otherwise have expected to remain in office until you die in your sleep in your nineties, the prospect of mass protests or riots or whatever culminating in you being forced to enact critical reforms makes a huge difference in terms of how long you expect to remain in office. So when the people start to mobilize, you are going to find repression pretty attractive.

On the other hand, if you otherwise would have expected to be stepping down in a few months anyway, perhaps because you are ineligible to run for another election (if we're talking about a democratic leader) or because things have been unstable for some time anyway and the economy is in the tank (if we're talking insecure autocrats), the benefit of suppressing a mobilized challenge to the state is a lot smaller.

Granted, there aren't many leaders who are very sure they're going to be leaving office in a matter of months. But thinking about what happens at the extremes helps to clarify the general tendency. Essentially, leaders who expect to remain in office longer are playing a higher stakes game. And so they're more willing to resort to desperate measures to keep from losing that game.

If you're thinking about Assad and the atrocities taking place in Homs right now, before you insist that he is doing what he is doing because his hold on power is threatened and therefore object to the argument Emily and Courtenay present, consider this -- how much longer would Assad have been in power if there had been no protests? Somewhere just shy of forever, right?

The claim isn't that leaders whose hold on power becomes weaker as a result of the mobilized challenge itself won't repress. Indeed, these are exactly the leaders we expect to do so. It's that those leaders who were initially facing a high baseline probability of remaining in office are the ones who have the most to lose when challenged by their people.

Yet such leaders are also sensitive to changes in the ability of the court to constrain them.

So Emily and Courtenay expect that we should find that it is only among the most secure leaders that judicial effectiveness and commitment to IHRTs will reduce repression -- because it is only the most secure leaders who are likely to be sufficiently threatened by mobilized challenges to be tempted to engage in significant levels of repression anyway.

I'll skip the details of the empirics, just as I've skipped the details of the formal model, but here are graphs of the key results.

Note, these graphs are plotting the change in the probability of systemic torture, as a function of changes in judicial effectiveness (first graph) or commitment to the Convention Against Torture (second graph). Thus, when you see that the line starts out high then drops sharply as we move to the right, you should note that what we are seeing is not that insecure leaders repress a lot while secure leaders do not (which would contradict the theoretical argument). Rather, what we are seeing is that, as job security increases, the difference between the probability of systemic torture as a function of judicial effectiveness or commitment to an IHRT goes from being indistinguishable from zero (as expected -- you can't reduce the probability of repression much when it was low to begin with) to being negative and statistically significant.

There's much more detail to their argument than I've presented here. And I've only talked about the results from one of the papers Emily discussed. In the second one, Emily and Courtenay turn their focus to the level of mobilization against the state, and the effects thereon of commitment to an IHRT and the leader's job security.

But the take home point from this brief post is that international law matters, contrary to the received wisdom in IR. It doesn't always have the effect that we'd like, to be sure. But if we focus our attention just on those cases where it is most likely to make a difference (leaders who would otherwise have had the greatest incentive to repress), we do indeed find the effect we'd like to see.

More on Negative Advertising

2012-02-23T09:04:00.005-05:00

John Sides was kind enough to respond -- at length -- to this post of mine. He makes several points, both in response to me and comments left on the original post of his that kicked off this discussion.

His point about experimental studies is well taken. As I admitted in my post, this is not my area of expertise, and I was unaware of such studies. The argument I was making applies to observational studies.

That said, I still think Sides is making a stronger argument than the evidence warrants. Look below the fold for some thoughts as to why. (Note to regular readers -- I'll be posting about dead bodies again soon. Fear not.)

There are two points that I'd like to make. The first concerns the excellent paper by Matt Blackwell.

First, Blackwell himself uses far more guarded language than Sides does. Here is a key passage:

If the negative campaigns are instead weaker, perhaps because campaigns go negative when they are in trouble, then results only grow stronger for non-incumbents. The results for incumbents highlight the potential violations of ignorability for that group. The results are fairly sensitive to the degree of confounding, both in the point estimates and the confidence intervals. Note, though, that this confounding would have be above and beyond any information contained in polls and pre-campaign measures of competitiveness.

It is clear that there is some sensitivity to the sequential ignorability assumption and we could attempt to gather more data (quantitative and qualitative) to justify which direction this confounding is likely to lean.

Emphasis mine.

Sides is absolutely correct that Blackwell's paper speaks directly to the concern I raised. But I think he is reading more into the results than Blackwell himself would encourage us to do. In my original post, I argued that if we believed negative advertising "works", but also believed that campaigns go negative when they are in trouble (not my original words, but Blackwell says it better, so I'm using his), and further believed that it is difficult to fully disentangle which campaigns are in trouble, then we'd expect to fail to uncover evidence that negative advertising works even if it actually did. Blackwell makes essentially this point.

He also stresses, admittedly, that the confounding I am so concerned about would need to be above and beyond what is contained in polls and pre-campaign measures of competitiveness. Though I'm not sure I'd be too quick to dismiss such a possibility, I will at least amend my previous statement to say that journalists who have strong priors that negative advertising "works" would not be required to update those prior beliefs nearly to the degree Sides seems to think that they should in response to the best current available evidence from political science.

Suppose we have two alternative states of the world. One is that, all else equal, negative advertising systematically increases a candidate's share of the vote. The degree to which this is true would of course depend on how many ads are run by each candidate and many other factors, but this view stipulates that, for all the variation in the magnitude of the effect that might be driven by such concerns, it is nonetheless true that there is typically going to be a non-trivial effect.

The other view is that negative advertising, at best, works some of the time, but is also likely to be either ineffective or even to backfire in many cases.

The real question of interest is what would someone who assigned a very large probability to the former view believe after reading Blackwell's paper? According to Bayes' Theorem, such a person would revise their belief downward provided that we believed that the probability that Blackwell would find what he finds given that we live in the first state of the world is lower than the probability that he would find what he finds given that we live in the second. I'm not sure exactly how we know what those probabilities look like, but I'll grant that it's entirely possible that the probability that we'd observe his results might well be higher if we live in the second state of the world. But there's no getting around the fact that there's a lot of uncertainty here, and by Blackwell's own admission, the results are sensitive to assumptions we make about things we can't observe.

In short, I overstated things when I said there is no grounds for any belief updating. But Sides is overstating things as well when he says that the journalists' view is a zombie. Blackwell's paper does not show that. We cannot, by any means, rule out the possibility that the journalists' view is essentially correct.

At least, assuming the journalists' view is what I characterized as the first potential state of the world. This leads me to my second point. Sides undoubtedly pays more attention to media coverage of campaigns than I do, so perhaps I should simply defer to him, but it is not my impression that journalists believe context is completely irrelevant, as his most recent post seems to indicate. Again, I don't follow campaign coverage nearly as closely as I'm sure Sides does, but what I have read leaves me with the impression that journalists are already assuming that ads that run close to an election matter more, that a candidate is unlikely to gain an advantage by running attack ads if his or her opponent is running an equal number of attack ads against them, etc. But maybe I'm wrong. My goal here isn't really to defend journalists. As any regular reader of this blog knows, I spend plenty of time taking the media to task myself.

Rather, as I said in my first post, my goal here is to push back against the idea that we can conclude that policies do not work if social scientists fail to find evidence that they do when analyzing observational data. This is a point that applies to negative advertising, military buildups, Keynesian stimulus, and a whole host of other policies that are more likely to be attempted when the outcome of interest is already likely to be unfavorable. If I have misunderstood what Sides takes to be the journalists' view, and his argument amounts to no more than the claim that they need to acknowledge that context matters, then I will gladly declare that the two of us are in substantial agreement. If Sides is saying that those who believe we live in the first state of the world that I described above need to accept that political science research provides compelling evidence that we instead live in the second state of the world, I must respectfully disagree.

Zombies, Negative Advertising, and Military Buildups

2012-02-22T13:34:00.003-05:00

(I posted this yesterday, then took it down. This version of the post makes the same substantive point, but hopefully does so in a more careful and appropriate manner.)

John Sides likens the idea that negative advertising works to zombies -- a dead idea that just won't die (in the vein of Zombie Economics). With the obvious caveat that I'm nobody's expert on negative advertising, I respectfully disagree.

First, though, let's be clear about what Sides is actually saying. Framing the belief that negative advertising works as a zombie suggests a much stronger critique than Sides actually provides, and to his credit, Sides is quite transparent about that.

So let's set aside talk of zombies and focus on the real argument Sides makes:

Which is to say: we haven’t remotely arrived at a place where “research” suggests that negative ads “work.” I’m not asking that reporters or commentators or foundations say that it “never works.” I don’t think the literature suggests that either. I’m just asking for some engagement with this research. I think good reporting demands it.

I don't see any problem with wanting journalists to engage with political science, of course. But the presumption here is that the empirical evidence that we have should cause someone who had a strong prior belief in the conventional view to revise their belief. And, as I'll discuss in just a minute, I'm not persuaded that there are grounds for such a conclusion.

Now, Sides is careful to say that he's not actually claiming that negative advertising doesn't work. I don't want to misrepresent his position. But what is there to be gained from engaging in a literature unless we think that those who do will be forced to revise their beliefs, at least to some degree?

Also, let me be clear that I'm not saying that Sides should revise his beliefs. What I'm saying is that, given that journalists have a strong prior belief that negative advertising works, the important question is whether the political science literature should make them less confident about that.

Here's why I don't think it should.

If we assumed that 1) negative advertising typically turns voters off, 2) yet typically increases the relative share of the vote that will be garnered by the candidate running the negative ads, and 3) candidates have reasons to want to avoid the former effect when they already believe they are likely to win, we would expect to find that negative advertising does not work when we compared the vote shares of candidates who go negative to those who do not.

The reason I am writing about this is not that I have any particular interest in negative advertising. Rather, I have made precisely this argument with respect to the effect of military buildups. Many people in IR believe that military buildups promote war. Others argue that it prevents war. A large number of studies have tried bring data to bear on the question and figure out who's right. Some find support for the view that it promotes war, some do not, but virtually none of the studies provide any support for the view that military buildups "work". So if we were to call on journalists to "engage" with the literature that hasn't "remotely arrived at a place where 'research' suggests that [military buildups] work", that would be analogous to Sides' argument.

The problem is that, as Kyle Joyce and I demonstrate in this working paper, we would expect to find that military buildups do not work even if they did. And worse, this problem is almost certainly not one that can be solved by matching or structural estimation, let alone the mere inclusion of control variables.

If we would fail to conclude that negative advertising works even if it actually did, then there are no grounds for expecting someone who currently believes that they do work to change their mind.

That doesn't in any way prove that they do work. If Sides' prior belief is that they don't, I'm not going to argue that he should come to believe that they do. But I don't know that, in this particular case, we've got much ground to stand on when we admonish journalists for failing to engage with the relevant poli sci literature. Because I'm just not sure it tells us much yet.

Now, in fairness, the debate is quite complicated. There's a lot more to this story than the mere observation that candidates who go negative are no more likely to win than those who don't. The absence of evidence that the Daisy ad moved opinion is problematic for the traditional view. There are, of course, anecdotes in favor as well though. Such as Mitt Romney's victory in Florida. And I'm sure there are many other points that hopefully you'll forgive me for overlooking, being that this is not my area of expertise.

Nonetheless, the main point is that we cannot simply look at the fates of those who adopt a certain strategy, compare it to the fates of those who do not, and draw from that any strong inferences about whether that strategy "works". So long as we believe that the actors whose behavior we are studying is strategic, and so long as we cannot perfectly distinguish between actors who would have the greatest incentive to adopt the strategy, the problems with such arguments are difficult to overstate.

Please also see this followup.

My American Foreign Policy Class, Part II

2012-02-21T19:10:00.003-05:00

This set of slides completes my discussion of how the main foreign policy tools would be evaluated optimally. Towards the end, I began to discuss not only what we would expect policy to look like if USFP simply sought to promote the national interest, and to do so as efficiently as possible, but also whether the policies we see in reality bear any resemblance to such expectations. That's a question I'll address in far more detail in the latter half of the course, but, for now, I've argued that foreign aid disbursements actually do look a fair amount like what we'd expect if the US was setting policy optimally, while trade policy is pretty clearly suboptimal.

It should go without saying, but let me nonetheless add the caveat that my goal is to challenge my students by forcing them to think about things differently than they might be used to. All of the issues discussed in this class are far more complex than the treatment I give them. I focus a lot of attention on pushing back against arguments that students are likely to have heard (and continue to hear) from talking heads. That doesn't mean I don't think there's even a grain of truth to the arguments commonly found in the popular press, or that I'm unaware that the politics of trade is a whole lot more complicated than the way I present it here, or that the US is actually getting what it pays for from all of the recipients of its foreign aid, etc etc etc.

Nor am I trying to encourage my students to go out there and advocate for reform. As I'll argue in the next set of slides, there's a very good reason why US trade policy is suboptimal, and the barriers to reform are considerable. My goal is here to help them critically evaluate US policy and, more importantly, to understand why it looks the way it does. I don't think we can answer questions like "how does domestic politics influence policy choice" if we don't first establish a clear baseline of what policy would look like in a world where USFP maximized the national interest (assuming such a thing could be easily defined -- a topic I'll actually discuss at the start of the next section). When I argue that trade policy is suboptimal, it's not because I'm promoting an agenda. At least not deliberately. My primary goal is to lay the foundation for arguments I'll present later about the impact of interest groups.

When is a War a War in the Eyes of the Media?

2012-02-21T15:27:00.004-05:00

Defining "war" is no easy task. We political scientists have a variety of definitions.* So you certainly can't fault the media for being unaware of the esoteric debates about coding rules that take place within the academy.

What I do find a bit puzzling, though, is the internally inconsistent standards the media applies to armed conflicts. Specifically, It seems to me that virtually any level of violence that takes place after a border has been crossed will justify using the word "war", while the standard for determining whether a state is experiencing a "civil war" seems to be just a wee bit more demanding.

When the armed forces of one state cross an international border and engage in hostilities with some force on the other side of that border (be it the armed forces of another state or not), the media mindset seems to be "if it bleeds, it leads." And how better to move copy (or, these days, attract pageviews and generate links) than to call the conflict a "war"? Who wants to read about a "skirmish"?

Consider, for example, the conflict between Israel and Hezbollah in 2006, or the conflict over South Ossetia in 2008. These were limited conflicts that resulted in small numbers of battle fatalities (between 320 and 720, depending on whose estimates you believe, in the case of the former, and less than 400 for the latter). Contrast that with the coverage of the ongoing conflict in Syria (where even conservative estimates have more than 1000 battle fatalities occurring**). Or recall the debate about whether Iraq was experiencing a "civil war" in 2006.

Some scholars have already proclaimed that Syria is in the midst of a civil war. Now, one can quibble with academic definitions. Just because political scientists say it is a war doesn't mean it is a war, you might say. Fine. But even if we applied the standards the media itself applies to other conflicts, this is most definitely a civil war.

So what gives?

I'm no media critic. Perhaps everything I'm about to say is pure bunk. But, for what it's worth, my hunch is that the reluctance to refer to refer to internal conflicts as "civil wars" stems from one or more of the following:

1) a mistaken belief that the US Civil War is a typical civil war (the same way American journalists insist on talking about who "wins" PR elections as if that has anywhere near the same meaning it does in a FPP context), and thus an expectation that there will be two sides, both of which will be well organized, and that the fighting will take the form of conventional battles

2) a perhaps reasonable belief that the act of labeling an internal conflict a "civil war" is an implicit criticism of any policies that sought to prevent such and/or an implicit argument in favor of intervention

3) a reflection of the tendency of the media to uncritically adopt the language and framing of those who issue public statements -- which governments tend to do (and in which they tend to insist that they've got everything under control), but rebel organizations often do not.

But, like I said, I have no idea if that's all nonsense. What do you all think? Do any of these make sense? Some more so than others? What other reasons might there be for why (US) media outlets are so reluctant to label internal conflicts as "civil wars" when they are so quick to throw the word "war" around if a border got crossed?

*Actually, I would argue that we have a relative small number of definitions and a wide range of coding rules. I see no good reason to treat "coding rule" as a synonym for "definition". But, as with so many other matters, I fear I am in the minority on that count, so let's just pretend I never said anything. It's not like anyone reads these footnotes anyway.

**I am, of course, setting aside civilian deaths. I do so not because they are unimportant. They obviously are extremely important. Rather, I am not counting civilian deaths, as I did not when discussing the battle fatalities associated with the two interstate "wars" referenced above, because I do not want to conflate a campaign of government repression with a civil war. The two often go together, but are nonetheless distinct, and one cannot blame the media for failing to refer to instances of the former as "civil wars".

US Electoral Politics and the Probability of an Israeli Attack on Iran

2012-02-19T18:38:00.006-05:00

I continue to believe that the probability that the standoff with Iran will end with either the US or Israel using military force is lower than people realize. But it certainly does seem to be on the rise. I'm feeling less and less comfortable with the fact that I made a bet with an Israeli friend of mine that there will be no Israeli attack.

And, just now, I'm thinking that the probability of an Israeli strike on Iran before the year is out just might be significantly lower if not for US electoral politics.

Look below the fold for some thoughts on why that might be.

There are 3 assumptions that warrant some discussion.

A1. A sustained effort at sanctioning Iran would have a non-trivial chance of causing Iran to abandon its nuclear program.

A2. The sanctions regime would remain in place longer if it either was not a presidential election year in the US, or if Obama's electoral fate was already determined.

A3. Israel is willing to wait at least a few more months, giving sanctions a chance, before they decide whether to attack.

Regarding the first, Iran is signaling that it wants to talk, and is at the same time resorting to more and more desperate attempts to fracture the coalition arrayed against it. Taken together, these developments suggest that Iran just might be starting to wonder how long they can live with the sanctions. And no surprise. Iran's currency has lost more than half its value, and Iran has begun to default on loans, leading major suppliers of agricultural exports to cease or delay shipments. Without Indian rice and Ukrainian maize, Iran may soon begin to starve.

Sanctions often do not succeed in changing the target's behavior. But neither are they typically this effective in wreaking havoc on the target's economy. Make no mistake -- the crippling sanctions many have been calling for are here.

The question is how long they will be.

It's a war of attrition at this point. Sanctions aren't just hurting Iran. Many of the nations involved will be facing elections soon. That includes Iran, several EU countries, and, of course, the US.

The sanctions have begun to drive up the price of oil. As I mentioned the other day, that's good news for Russia, and indeed for all the other major oil exporters. But it's bad news for the US.

And the GOP is doing everything it can to make sure that voters are aware of it. Sure, some of them are blaming high gas prices on Obama's economic and energy policies. And it's entirely possible, so far as this non-economist can tell, that the average price of gas in the US would be lower today if Obama had been pursuing different economic and energy policies over the past 3 years. But it doesn't take an economist to see that a dramatic runup in oil prices that began more or less around the time that the sanctions on Iran started to target the oil industry and central bank probably has an awful lot more to do with those sanctions than it does any domestic policies of the US. Put differently, the GOP may or may not have a valid point with respect to the baseline that was in place when the standoff with Iran really started to escalate, but the short term changes we've seen since then simply cannot be explained by policies that haven't changed in that time. In short, the GOP is creating a strong incentive for the administration to abandon sanctions. Despite the fact that they are simultaneously criticizing Obama of not doing enough to stop Iran from developing nuclear weapons. And despite the fact that most of the Republican candidates for president would be willing to consider a US strike on Iran -- which I can assure you would not bring oil prices down.

Now, to anticipate an obvious criticism -- no, I don't imagine that voters would completely ignore high gas prices if Republicans weren't calling attention to them. And if rising oil prices were to push the US into another recession, then Obama would be done for no matter what. There are definitely limits to how much pain the US would be willing to tolerate in order to keep sanctioning Iran even if we took domestic politics completely out of the equation.

But that doesn't mean there's no role for opposition rhetoric to play, at the margins, in shaping how long the US will want the sanctions to remain in place.

We've got pretty good evidence that the public punishes incumbents for poor policy outcomes to a greater degree when the opposition criticizes the policies that (are perceived to have) led to them. If the Republicans were out there saying that a nuclear Iran is an unacceptable threat to both the US and Israel (oh wait...they are), and that therefore Obama should be doing everything in his power to prevent Iran from developing nuclear weapons (again, check), and thus the US public should accept that the price to be paid for addressing this threat is high gas prices (not so much), that might give the administration the breathing room to keep the pressure up on Iran. I'm not claiming that such statements from the GOP would completely shield Obama from economic retrospective voting. Of course it wouldn't. But, again, we have pretty good evidence that government policies that are backed by both parties attract significant support from the public. I'm not sure how much longer the US could afford to keep the sanctions in place in a world where the GOP wasn't using gas prices as a cudgel with which to beat the administration, but it looks to me like Iran is suffering badly enough that a few extra weeks could make a difference.

Now, I'm not passing judgment on the GOP here. I don't expect parties to pass up an opportunity to maximize their chances of taking power all that often, even when their opportunistic behavior might have some tragic results. I'm just cynical that way. I don't think the GOP is unique in this regard either. I've argued elsewhere that the US might not have invaded Iraq in 2003 if the Democrats had been willing to oppose the war, and the US involvement in Iraq's civil war might have ended in 2007 if not for the Democrats' inability to commit to refraining from using the war as a wedge issue. Perhaps I am wrong about those arguments, just as I may well be wrong about this one. But either way, I want to be clear that I'm not trying to score points politically. My view of the democratic process, based on my research and that of others, is that opposition parties have an incentive to do their best to get the public to punish the incumbent for pursuing policies that the public might well want the incumbent to pursue if they had a better grasp of what was going on.

Of course, even if you don't believe me that the GOP's rhetoric might influence Obama's willingness to maintain sanctions on Iran long enough to have a real chance of affecting their behavior, you'd still be hard pressed to argue that electoral politics didn't matter. If the next presidential election wasn't until 2016, Obama would have a lot less reason to worry about how the crisis with Iran might influence the price of gas within the US.

Back to the crisis over Iran's nuclear program.

The Israelis have made it clear that they are losing patience. But there is pressure on them to wait and give sanctions a chance. If you believe they are immune to this pressure, then it may not matter that US electoral politics is likely to bring the sanctions regime down earlier than might otherwise be the case. If, however, you believe the Israelis can be persuaded to wait a little longer, then it matters how long the US is willing to keep up the pressure on Iran.

To sum up, I think the probability that Israel will attack Iran is larger than it would be if not for electoral politics in the US creating a disincentive to giving sanctions time to work. There are a number of assumptions underlying this argument that some may find objectionable. I've tried to be transparent about what they are. I think these assumption are reasonable, but I'm not going to pretend I think it's impossible to disagree with them. And even if the assumptions are warranted, the argument I've laid out is probabilistic in nature. There may yet be no Israeli attack on Iran, even if the sanctions regime collapses before Iran backs down due to fears of what high gas prices would do to Obama's prospects for being reelected. But electoral accountability, one of the hallmarks of democratic governance, just might be making an already volatile situation that much less stable.

Miscellaneous Links

2012-02-18T16:38:00.000-05:00

Below the fold.

1. In defense of particularism in American Foreign Policy. Jay Ulfelder offers some thoughts on Kennan, the Truman administration's response to the Chinese civil war, and the implications for the debate about R2P and Syria.

2. US, EU welcome Iran's offer to restart talks. Talks are often just talks, but not always. It is of course too early to tell, but this isn't the only sign that sanctions might be working.

3. Iranian naval ships seem to the Syrian coast. Many of the advocates of a Syrian intervention oppose military action against Iran (though there is some overlap -- see here). You'd think then that some of the eager proponents of an R2P style intervention in Syria might have noticed that such an action is extremely likely to result in a war with Iran.

4. Russia, despite opposing sanctions on Iran, is profiting handsomely from them. "Russia is now the world’s largest oil producer, pumping about 10 million barrels of oil a day, slightly more than Saudi Arabia." (Did you know that? I didn't.) "Analysts estimate that Iran jitters have added $5 to $15 a barrel to the global price of oil, which means an extra $35 million to $105 million a day for the Russian industry. And the taxes the Russian government has received from those sales have been a political windfall for Prime Minister Vladimir V. Putin as he campaigns to return as Russia’s president."

5. Venezuela, too, stands to profit. And they need additional oil revenue a lot more than they need Iran.

6. Brinkmanship isn't the right model for the EU crisis. I agree. I think it's crisis bargaining. We're looking at the risk-return tradeoff, not dancing on the edge of a cliff while being handcuffed together. Subtle difference, maybe, but it has implications -- not the least of which is that Germany isn't flirting with mutual annihilation so much as generating a risk of a suboptimal but nonetheless survivable outcome in hopes of getting a better outcome.

7. Wartime elections. My advisee, Nick Nicoletti, summarizing his dissertation work. He argues that, under the right conditions, the media might correctly report that the war is going well, yet the public will replace the incumbent with an anti-war opposition party. Similarly, he argues that the opposition will sometimes pledge to continue a war that they know is not going well, and which the media accurate reports is not going well, and the public will put them in power. Most importantly, his model suggests that what drives these tragic outcomes are core features of the democratic process.

8. "False-positive" psychology. What's that? The stars next to my coefficients aren't as magical as I think they are? How can that be?

9. Realtime smartphone polling. React Labs, a joint venture that includes UC-Davis political scientist Amber Boydstun, is looking for beta testers. So if you're planning on watching the debate this Wednesday, please so consider signing up.

10. The US is not planning to attack Iran. According to the Onion. Four years ago.

EITM 2012

2012-02-16T16:25:00.000-05:00

Information here. This year, it is at Princeton.

I attended EITM as a participant in 2008 and as a mentoring-faculty-in-residence in 2009 and 2010. Those were very rewarding experiences. I strongly encourage grad students to apply. You will be exposed to a variety of views about how to make the most of theoretical and empirical models. You will meet a lot of great faculty and future colleagues. And you will received detailed feedback on your research.

Review of Clarke and Primo

2012-02-16T15:52:00.006-05:00

I've plugged the book a few times on here already. Now that I've finished reading it, I'm going to offer a more detailed review. And then stop bringing it up every other post. (Probably).

What role do models serve in guiding social inquiry? What is "the scientific method"? Is there such a thing? How should theoretical models and empirical analysis inform one another?

These are among the important questions tackled by Clarke and Primo's A Model Discipline: Political Science and the Logic of Representations.

The authors argue that political scientists operate under an outdated and deeply flawed conception of "the scientific method" that has long since been discredited by philosophers of science and abandoned by practitioners of the natural sciences. This view is known as "hypothetico-deductivism" (H-D). Guided by their cult-like belief in H-D, the way the modal political scientist "does" science is to formulate a theory (which will usually be informal, if present at all), "derive" a hypothesis therefrom (read: state a hypothesis and assert that it follows from the previous 5 pages of poorly organized lit review), then "test" the hypothesis, and judge whether the hypothesis was supported by the analysis (i.e., conduct some statistical analysis, gaze at the stars, and then commit the fallacy of affirming the consequent).

One of the few complaints I have about the book is the space devoted to persuading the reader that this is indeed how political scientists "do" science. Perhaps it's just because I am among the minority of political scientists who was already more than a little sympathetic to their views, but I needed no convincing. And, looking at the Scope and Methods courses taught in the departments I've been in, what I've heard at the EITM institutes I've attended, and my interactions with faculty and grad students at other departments, it seems to me that there is very little doubt that most grad students in the US are hearing the message that this is "the" accepted way of doing things, even if they might be aware that there are a few wrong-headed crackpots who disagree. But maybe there are some readers out there who will doubt that most grad students are taught "the scientific method" entails formulating a hypothesis, testing it, and then evaluating whether to accept or reject the hypothesis on the basis of that empirical test.

Another small complaint I have is the way they speak about the EITM movement as though it is a monolith. There are those who hold the views they attribute to everyone associated with the project, to be sure. But as someone who has participated (as both a student and as a mentor) in three EITM summer institutes, I can say with some confidence that there are many people involved with EITM who do not share the view of Morton and Granato. That the views expressed in this book will be received well by more EITMers than the authors realize. But that is, of course, a very minor quibble.

At any rate, Clarke and Primo go on to demonstrate the problems with the H-D approach that many in the field embrace and indeed preach. To oversimplify their argument, there are two separate but equally important reasons why we should reject H-D.

1. We can gain valuable insight from models that do not generate hypotheses that garner "empirical support."

2. It is fundamentally impossible to test theoretical models, and extremely difficult to use statistical models of observational data to test hypotheses.

In short, testing is neither as important as it is widely presumed to be, nor anywhere near as possible as it is presumed to be.

So what should we be doing instead?

Clarke and Primo argue that there are 4 roles a theoretical model (be it game-theoretic, computational, or informal) can serve, and four roles an empirical model (be it quantitative or qualitative) can serve.

The four roles for theoretical models are: foundational, organizational, exploratory, and predictive.

The four roles for empirical models are: theory-testing, prediction, measurement, and characterization.

They argue that prediction, in the sense of forecasting, is not what most of us are after. It might have been nice if they'd have had just a little more to say about such models, but that's not really the main focus of the book.

Foundational and organizational models have no need to be tested. The former serve to discipline our thinking and to inform future, more detailed, modeling efforts. They give us a baseline against which we might compare other models. The latter primarily seek to account for empirical regularities which have already been established. They give us a way of understanding why that which we think we know to be "true" is "true". What then are we supposed to test? What we already knew?

Exploratory models might be evaluated by the extent to which they generate observable implications that correspond to what we observe empirically. So there is indeed room for dialogue between theoretical and empirical modeling efforts. But it is important to note that some theoretical models do not aim to generate new observable implications, and to judge them by their ability to do so is to miss the point.

With respect to empirical models, they argue convincingly that theory-testing is the task to which empirical models are least suited.

That does not mean empirical models aren't useful however. It just means that we should focus more on their ability to generate insight through their impressive ability to assist efforts at measurement and characterization.

By characterization, they mean that empirical models serve to help us make sense of the patterns of association found among our data.

This isn't mere hair-splitting. When you analyze your data, you are engaged in the same sinful act of simplification and reductionism that I engage in when I analyze my formal models. There are of course differences between empirical models and theoretical models, but they are fewer and more modest than many empiricists realize.

When you fit a model to your data, and impose the assumption that the pattern of association between X and Y is linear, then determine that, given your many unstated assumptions, there is less than 5% chance that you would observe a coefficient estimate as large or larger than the one you have identified by random chance alone if the truth was that there was absolutely no relationship whatsoever of any kind between X and Y, you have not demonstrated that there is a linear relationship between X and Y. You have not demonstrated that there is a monotonic relationship between X and Y. You have not demonstrated that there is an unconditional relationship between X and Y. Your empirical model has nothing to say about why X and Y covary, and probably has a lot less than you realize to say about whether X is causing Y, Y causing X, or whether both are being driven by some third factor that was omitted from your analysis. What you have shown is simply that, if we assume a lot of things that we know are almost certainly false, then we would conclude that it is probably more appropriate to view X and Y as exhibiting a linear relationship (whose slope is given by your coefficient estimate) than it is to assume that there is no relationship of any kind whatsoever. No more and no less.

That can indeed be very useful. But we need to start seeing that for what it is. A simplification. One that deviates from the Truth with probability asymptotically approaching 1. Not an objective discovery of reality.

So how then can theoretical models and empirical analysis inform one another? If we are never objectively discovering the truth, how can we say that we are testing our theories?

We should abandon the notion that we can or should be in the business of falsification. Questions of true and false are a waste of our time.* What Clarke and Primo suggest we should, and can, do instead is focus on constructing explanations.

There are two prominent views of what it means to explain, though neither is without its flaws. One focuses on classifying the event or class of events we seek to explain into a broader category, whose properties are well-known. We might explain the lack of a global effort to curb CO_2 emissions, despite widespread agreement (at least amongst elites) that climate change is a real concern and CO_2 is a key contributor to it, by observing that this problem is an example of the Tragedy of the Commons, for example.

The other focuses on identifying a mechanism and demonstrating that said mechanism reliably produces the outcome it supposedly explains. This notion of explanation lies at the heart of the causal inference movement. However, Clarke and Primo rightly note that even if we can be entirely sure that E(Y|X=x_1) =/= E(Y|X=x_2), that still doesn't provide much of an explanation. We still need to know why this is true.

Here, then, is where they see the potential for fruitful integration of theoretical and empirical models. Empirical models can characterize patterns of association, but cannot explain them. Theoretical models can explain why we might observe a pattern of association, but cannot tell us whether such a pattern is observed in reality. If we have a theoretical model that identifies a mechanism by which a certain pattern of association might be observed, and we do indeed observe said pattern of association, that goes some ways towards generating insight.

But it does not "prove" the model right.

Explanations do not need to be true any more than models do. It can be useful to compare them to another and determine whether one explanation outperforms another with respect to some specific task. But even that is not always necessary. Explanations are at best partial. The question isn't whether they are right, or whether they crush all other ways of thinking about the problem. The question is whether they give us a new way of understanding the problem. One that we find useful.

Inevitably, the question of whether a given explanation is useful is going to be something of a judgment call.

But if you ask me, that's a lot better that than the cult of statistical significance.

In closing, Clarke and Primo offer a badly needed corrective to the dominant thinking about "the scientific method". The book offers a cogent account of the problems with this approach, and provides examples from recent work in political science of how we might do things differently. As provocative as I'm sure their claims will be seen by many, in the end, they aren't actually calling for such radical changes. In fact, in many ways, their arguments present less of a threat to the standard way of doing things than do the arguments of some within the EITM movement (which, again, is not monolithic). Formal theorists (be they game theorists or computational modelers) need not submit to pressure to find and "test" some observable implication of every model they analyze. Those empiricists who neither have anything remotely resembling a theory nor any interest in anything remotely theoretical need not bother pretending otherwise. But we all do need to stop pretending that the only contributions to knowledge come from theory testing.

It simply can't be said enough. Models are objects. They serve a purpose. There are ways of evaluating whether they serve their purpose well. Asking whether they are "right" is not one of them.

Since we are all modelers, every one of us, this is a message we should all embrace.

Yet something tells me it will be a very long time before the H-D hegemony shows any signs of cracking.

Btw, if you're interested, I've blogged about related concerns before. See here in particular.

*I'm not making a post-modern statement here about the impossibility of making any objective statement about the world. I'm comfortable with the idea that certain things can be known. Such as whether an event took place. But I would not classify the causes of said event as something than can be objectively known. Relatively few political scientists are interested in simple questions of whether something occurred or not. Rather, we are in the business of answering questions about why certain events take place, or even in constructing general explanations that might apply to an entire class of like events.

The Pacifying Effects of Aircraft Carriers?

2012-02-14T17:11:00.008-05:00

Since World War II, there has been but one war that involved two states who both possessed aircraft carriers (Argentina's Veinticinco de Mayo saw some action at the beginning of the Falklands Islands War).*

You, of course, will be familiar with this from all the books and articles you had to read for comps about the pacifying effect of aircraft carriers.

What's that? You haven't read any such works? But you've read all about how mutual possession of nuclear weapons prevents (or does not prevent) conventional conflict, right?

Yes, they're different. In that one very obviously ought to influence decisions about whether to engage in conventional military conflict. I'm speaking, of course, about aircraft carriers.

No, no, no. They're different because nuclear weapons encourage restraint. And that's why there have been zero wars between pairs of states that both possessed nuclear weapons.

Some say so, yes. But the pesky Kargil War (which was also a war fought by two states that were, as of the start of the war, both democracies**) pitted nuclear-armed India against nuclear-armed Pakistan. If there's a difference between nuclear weapons and aircraft carriers in terms of promoting peace, the evidence isn't to be found here.

The difference, you'll insist, is that we have theoretical reason to expect jointly nuclear states to be less war prone. But do we really? And is there no reason to expect aircraft carriers to prevent escalation?

What we always hear is that nuclear weapons increase the costs of war. And a straightforward implication of most formal models of war is that an increase in the expected costs of war decreases the probability of war (though for an exception, see here).

But if nuclear deterrence works the way everyone but Schelling thinks it works, then this actually makes very little sense. Especially in jointly nuclear dyads.

If you know that a nuclear attack against me will cause me to retaliate with a nuclear strike of my own, you will not use your nuclear weapons. And neither will I.

And if we both know that, why should that stop me from sending in the tanks?

Indeed, we have some evidence that possession of nuclear weapons influences who gets what, but does not influence the probability of conventional armed conflict.

There's a whole lot less reason to believe that states who possess aircraft carriers will refrain from using them. Sure, their use doesn't up the costs of war to nearly the same degree as the use of nuclear weapons would. But, at minimum, there remains debate about whether anyone should expect nuclear weapons to be used in a war between two nuclear-armed states.

Of course, we've got serious concerns about observational equivalence here. The list of dyads who are jointly democratic is very similar (thought not identical) to the list of dyads who are jointly highly economically developed,*** which is very similar (though not identical) to the list of dyads that are jointly in possession of aircraft carriers, which is very similar (though not identical) to the list of dyads that are jointly in possession of nuclear weapons. Given the relative rarity of war (for which we should all be thankful), the task of disentangling the impact of any of these effects from the others is daunting indeed.

So it could be that all of them have some independent pacifying effect. Or perhaps even none of them. I don't know. (If some enterprising grad student out there wants to go and sort these claims empirically, that would be great. I'm not sure how easy that would be, and there are some serious issues about causal identification, but I'd still be curious to see what the results look like.)

For now, though, I'll say this -- we have a very good understanding of how and why interstate crises are less likely to escalate to conventional, interstate wars when the expected costs of those wars increases (see just about any paper in the crisis bargaining literature). It's pretty hard to dispute that joint possession of aircraft carriers would increase the costs of war. It's debatable whether joint possession of nuclear weapons should, if you buy the logic of MAD. Arguments have been made about how joint democracy and joint economic development might increase the costs of war, but the causal chains here are a whole lot more complicated. (We're also told that democracies and/or developed states find it easier to engage in costly signaling, but the logic there is really problematic.)

Curious, then, that of all these effects, the least ink has been spilled about the one that has the simplest logic.

*H/T Jeremy Wells, who pointed this out to me.

**Thankfully enough for the careers of many quantitative IR types, this war gets dropped out of most data sets thanks to Polity assigning it a missing value due to the regime change that came about at the end of (and as a result of) the war.

***While there are some statistical concerns with this analysis, at least with respect to the claim that the economic effects vitiate the impact of democracy (see here), I think Gartzke has made a fairly compelling case that joint economic development is associated with peace (though, as with the democratic peace, he argues that there is no monadic effect -- i.e., developed states fight undeveloped states plenty often). Part of his story concerns costly signaling, and I'm not sold on much of that (for reasons I've discussed elsewhere), but he also articulates a few different reasons why highly developed states will suffer greater costs of war, and these are unrelated to the technology they will bring to bear in a conflict.

Miscellaneous Links

2012-02-12T16:06:00.000-05:00

Below the fold.

1. Zach Jones on why you should be skeptical that Israel will attack Iran. I agree. In fact, I'm so sure that he's right that I recently bet an Israeli friend of mine that Israel would not attack Iran. Of course, if you knew my track record with betting on political outcomes, that should probably cause you to update your priors in the opposite direction...

2. This is what bargaining looks like: Israeli settlement edition. Ruling Likud may propose legislation granting tax exemptions for donations to "Zionist settlement." Apparently, the language of the bill doesn't clarify where those settlements would have to be, and so presumably would include any beyond the Green Line. Great example of why the "bargaining" in bargaining models need only refer to efforts by one actor to acquire more of the disputed good short of war through a fait accompli rather than actual negotiations.

3. Opposition candidate says Pakistan will shoot down US drones if his party takes power. It's almost like the Pakistanis don't appreciate the US keeps killing their innocent civilians and lying about it afterwards. Geez.

4. Glenn Greenwald on how supportive progressives are of the US president killing US citizens. Hurray for the cult of personality and excusing abuses of power so long as "your guy" is the one doing it.

5. The West is stoking the Syrian conflict by supplying arms. So says...wait for it...Russia. Yeah, that Russia. They've got chutzpah, all right. Gotta give 'em that.

6. Andrew Exum on why the West should not arm the Syrian rebels. Money quote, "That doesn't mean the rebels don't stand a chance -- they can always carry out a guerrlla campaign using raids, ambushes and IEDs. But it does mean that schemes to train and equip the rebel groups will be more about doing something that makes us feel better about ourselves rather than an act that seriously changes the game in Syria."

7. The Duck on lessons of Libya, Parts One and Two. I'm sure it will surprise no one to hear me say that I'm more inclined to agree with Porter than Kelly.

8. Graphs, tables, Stata, R, and LaTeX. Excellent resources from Daina Chiba.

9a. The sequential urinal game. Yes, ladies, we actually do think this way. b. XKCD on same.

10. "People reading our [posts] will see our intelligence, recognize how clear and logical our arguments are, and..." Because two XKCD links in one post is never too many.

Predictability, Rationality, and Homo Economicus

2012-02-12T13:06:00.016-05:00

This is a follow-up to yesterday's post, which has attracted a lot of good comments. Of course it is also a "follow-up" to many other posts I've written. See here, here, and here specifically. And see also this excellent post at PM's question time.

In fact, the quote "Even though one class of theorist managed to colonize formal logic first in the social science, there is no requirement that all theories expressed formally be of their ilk" sums up, quite succinctly, so much of what I've been trying to say on the topic, that I'm more than a little jealous of PM for having written it.

Though I think that one sentence really says it all, I'll elaborate a bit more below the fold. Because accepting what it says means letting go of a deeply held belief that you've probably been conditioned to accept for years. It's not a complicated idea, but it's one that most social scientists have a hard time wrapping their heads around.

In the comments to yesteday's post, Nick writes:

"Incentives definitely drive behavior for the most part and the human element is predictable - at least in a probabilistic way. Nevertheless, despite Gary Becker's model, I still think there are human behaviors - like addiction - which are strange from a cost benefit standpoint."

If we stick with the mindset that "cost", "benefit", "incentive" and "utility" all refer strictly to materialistic concerns, of the sort that matter so much to Homo Economicus, then Nick's statement is eminently reasonable.

But the point I'm trying really really hard to make is that there is absolutely no reason to assume that this is what those words mean.

If a "cost" is anything that detracts from a person's subjective utility, a "benefit" is anything that adds to it, an "incentive" is anything that influences cost-benefit calculations, and "utility" is simply a measure of anything and everything that a person values -- whether it is acquire money, saving time, achieving sexual gratification, maintaining good social standing, minimizing cognitive dissonance or even cognitive effort, passing on their genes, or whatever -- then this is utter nonsense.

I mean no disrespect to Nick when I write this. I'm just trying to make a point that is very hard to get people to accept.

Everyone responds to incentives. All the time. Perfectly. Mechanically. Without fail.

Because that's what incentive means.

Now, you can respond that what I just said renders "rational choice" non-falsifiable. I've addressed that claim before, but I'll play along. If we choose to let "rational choice" mean what it is implied to mean by such a statement (which is of course fundamentally different from what others think it means, since others insist that it is not only falsifiable, but false), then yes, it is.

So what?

I'm not in the business of testing the hypothesis that people are rational. My need to pass normative judgment on the wisdom of other people's decision is minimal. I'm in the business of trying to understand human behavior (particularly in the context of international conflict).

The language of "utility" and "cost" and "benefit" is a system of mental accounting. Nothing more and nothing less. It helps us keep track of why we think this, that or the other thing might influence behavior. If we choose to formalize our arguments, rather than working with mental models, we have a better chance of ensuring that our conclusions follow from our premises. Sometimes that's really important, because the story we're telling has enough moving parts that the implications of our assumptions may not be obvious. Sometimes, there is nothing to be gained from doing so. The argument about whether to formalize one's argument is an entirely separate argument from the one about whether people respond to incentives or whether costs and benefits determine their behavior. To dispute the latter is to demonstrate that you are working with an unnecessarily narrow definition thereof. One that, in all likelihood, was imported from your understanding of what people believe in economics (which may or may not be an accurate reflection thereof).

The fact that people make decisions at time point t (or, taken to extreme, in each and every odd-numbered time point) that they regret at t+1 (or each and every even-numbered time point) does not mean that their behavior cannot be described in terms of costs and benefits. If the utility derived from satiating a short term jones for a high is large enough, then the mere fact that the person will wish when it is all over that they could kick the habit doesn't mean that they've behaved in a manner that isn't perfectly in line with their subjective utility considerations. To claim otherwise is to impose your view of how much value that person ought to attach to a short-term high (little) and how costly they should find it to quite (not very). If your goal is passing judgment on people, that's fine. But otherwise, I just don't see what we get from claiming that addiction can't be explained with costs and benefits.

Now, this is where Matt will point out that if we explain behavior by saying that people derive utility from it, we haven't really explained anything.

I agree.

I've never said otherwise.

There are two important points here. First, the fact that an explanation is vacuous does not mean it is wrong. Second, the fact that an explanation is right does not mean it is useful.

Thanks for those bits of fortune-cookie wisdom, fil.

The point is, in situations where all we can say is, "well they must have gotten utility from it somehow", such as drug addicts, suicide bombers, and people who routinely seek out abusive relationships, the question that social scientists should seek to answer is why a person would derive utility from such a thing, and, more importantly, how we can reliably distinguish those individuals who would derive utility from such acts from those that do not. If we have both a story for where such a preference comes from and the ability to verify the observable implications of our argument, we just might be on to something. I think, on a certain level, we all share that view. But then why are there people out there insisting that because suicide bombers, drug addicts, and people who tolerate abuse all exist, that this proves that people aren't rational or don't always respond to incentives etc etc etc?

Maybe the following example will clarify what I'm trying to say.

Suppose I have a machine that is supposed to function as a calculator. But it's a pretty bad one, because it always adds 7.5 to the correct value whenever you ask it to perform either addition or subtraction. And it always divides the correct value by 1.9 whenever you ask it to perform multiplication or division.

There is no question that this machine produces the wrong answer in every case. But would anyone describe the output of such a machine as "unpredictable"? Would we come up with a name for the discrepancy between its output and the truth? The "fil's messed-up calculator element"?

We are not at the point where we can say that all we need to do is subtract 7.5 from our answers (or multiply them by 1.9). I sincerely doubt that we ever will be. I don't even think that should be our goal. Simplification is at the very heart of science. Those who aspire to perfect explanatory power are tilting at windmills.

But the point remains the same. We know that people do not behave the way some of our simple models predict. But the discrepancies are anything but random. We do not yet understand exactly what causes them. We do not yet know exactly when to expect them. But we know that there are certain cases where they do not show up. EVER. That tells us something very profound.

There is a logic to human behavior. We have not fully decoded it, and I rather suspect (even hope) that we never will. Frankly, I wouldn't want to live in a world where we had, though I do of course hope that we move a bit further away from complete ignorance. The fact that we'll never (hopefully) be able to fully characterize the decisions people make doesn't mean that the phrase "the human element" is anything but romantic tripe that belongs in Hollywood and on Hallmark cards rather than in social science. Nor does it mean that people do not respond to incentives (which is defined as something that causes people to behave as they do!). Nor does it mean that cost benefit calculations only apply in some settings.

Once shed of the unnecessary Homo Economicus baggage, the terms "cost", "benefit", and "incentive" apply in all settings. They have to.

You may not feel the need to write down a formal model to represent the costs and benefits that you think create incentives for individuals to behave one way or another in some situation. That's fine. And if we can't say anything about why the benefits of some action are large, and the costs small, for some individuals, while the costs are large and the benefits small for others, then we haven't really said much of anything. No argument there. But don't tell me that people don't respond to incentives, or that costs and benefits only apply in some situations. That's nonsense.

The Remarkably Predictable Human Element

2012-02-11T14:50:00.012-05:00

One often hears that the reason we social scientists have it harder than natural scientists is "the human element." Of course, no one ever explains what exactly this means, because apparently it just goes without saying that human beings, by their very nature, are prone to doing all manner of utterly incomprehensible, random, senseless things.*

Something along those lines may indeed constitute an important difference between the social sciences and the natural sciences. I don't know. But what I do feel pretty sure about is that "the human element" cannot possibly be anywhere nearly the obstacle to the social scientific enterprise as we've decided that it is.

Why do I say that?

Because we're ignoring all the countless ways in which human behavior is extremely consistent, logical, and predictable. Of course, I understand why these things do not readily spring to mind. No one has ever written papers about them. We don't have any reason to notice them. But they really are everywhere.

You have, at some point in your life, undoubtedly come across many examples of this:

The picture shows a worn footpath right next to a paved walkway. The worn footpath represents the eminently predictable decision of many to optimize their effort to get to where they are going by traveling a shorter path. This happens all the time.

I'm not saying no one ever walks on the designated path. Some people feel compelled to. They think it's "cheating" not to. That's fine. I find it a bit weird, almost charming, even, that people feel this way, but whatever. It's not so much the choice between the paved walkway and the impromptu shortcut that I want to draw your attention to.

Rather, ask yourself this -- when was the last time you saw a makeshift footpath that had been created by countless numbers of people choosing to walk a longer path to get to where they are going?

Yeah, me either.

Or how about this -- when was the last time you saw someone go into a grocery store, fill up their basket or cart, get to the register, and haggle about the price, demanding that they pay more?

Ever meet anyone who said, "I hate politician X. Candidate Y is a creep too, but at least he's the lesser of two evils. For that reason, I'm going to vote for X."

Perhaps you've even noticed that when debating the most complicated matters of public policy, particularly with people outside the academy, the number of arguments your conversation partner seems to be sampling over in selecting their response is not only countably finite, but probably numbers no more than three. If that's too much jargon for you, what I'm saying is that you can be damn sure that if you mutter the word "inequality" or "gay marriage" or "immigration" within earshot of most people, they will very predictably respond with one of two or three stock answers that their side of the debate always falls back on. There is a certain amount of uncertainty here, specifically over which of the stock arguments they will supply, but I'd hardly call such behavior unpredictable. If we were really talking unpredictable, people might occasionally express arguments you haven't heard a million times before.

Now I'm just being pedantic, right? The debate about how best to study human behavior has never been about whether people would do any of that.

True. But why wouldn't they do these things? Is it because it would make no sense for them to do so? Because it wouldn't be in their interests? If so, what happened to your firm belief that people don't slavishly do what's in their interests? That human behavior doesn't always make sense?

Scholars have been trying to make sense of human behavior by focusing on the incentives people face for a long time. We've made a lot of progress. But there's a lot of variance left unexplained. A huge amount, really. No question about it.

But whatever it is that we're missing, it simply can't be that human behavior is random. It just can't. If that were true, we would expect to see a whole lot of weird shit that we don't see. If you compiled a list of all the possible aberrations from what most people think "rational" behavior is, I think you'd find that no more than a very tiny fraction of them occur with any frequency. That any of them do is interesting, and worth studying. No question about that. And I'm not suggesting anyone should go write papers about why no one tries to overpay at the grocery store, or why people don't vote for the candidate they like least, or anything of that nature. But I am saying that, upon observing that a large fraction of published papers in social science purport to show evidence of "irrational" behavior, we ought not infer that a large fraction of human behavior is "irrational". Because we're ignoring -- for good reason -- the countless intuitive, trivial, uninteresting, cases of people doing exactly what you'd expect them to do.

Lest you think I've set up a strawman, here and here are but two well-known examples of papers that explicitly assume that there is a random component to human behavior. These papers employ the quantal response equilibrium, which allows us to fit statistical models to observational data quite well (indeed, is guaranteed to allow you to say that your model fits the data...in much the way that every single regression equation is 100% true, if we count the error term), but it can't explain why we don't see all the crazy and whacky stuff we might imagine that literally does occur with probability 0.

Because, if you'll forgive me for getting a little technical, QRE models predict that the probability of one action being chosen over another is determined by the difference in utilities between the outcomes produced by said actions. Thus, actions that yield very similar utilities are chosen with nearly equal probability. This intuitive property allows us to explain why it is that most of us, if asked to play the role of P1, would chose SQ in the game below, even though SQ is not part of the unique SPE.

But suppose we add the option of sticking with the status quo followed by rolling around in peanut butter. QRE tells us that P1 would choose such a strategy just about as often as they would the SQ. And yet, we don't read a lot of newspaper stories about leaders rolling around in peanut butter.

Now, maybe there would be a non-trivial loss of utility for engaging in such an action because the public would think the leader was insane if he did so. I guess that's reasonable enough. Note, though, that in that case, we'd still predict that autocrats should be rolling around in peanut butter during crises all the time, since they don't need to worry about losing power if people think they're eccentric.

Now, I'm definitely not saying that we can't learn anything of interest from QRE, or from random utility models more broadly. They are incredibly useful for answering certain questions. We might alternatively think of the "random" components in a RUM as representing private information or measurement error, in which case we get the nice statistical properties but need not assume that the agents themselves are behaving a nonsensical manner (see here). My gripe here is not with RUMs, which I actually think should be used more widely (I'll say more about them in a future post). Rather, I'm taking issue with the idea that deviations from the predictions of simplistic formal models somehow prove that, because of "the human element", the behavior we seek to study is non-nonsensical, or "irrational", or random, or unpredictable. That may sound nice, but I'm afraid I just don't buy it.

And, I rather suspect, neither do you. Not really.

As others have pointed out, if you really believed that, you would be crazy to ever set foot in an automobile. Every time you travel any distance in one of those death machines, you are quite literally staking your life on the belief that human behavior is highly predictable. That the probability that any given driver will stop at red lights, drive on the right side of the road, and give you some warning when they are coming to a stop, is not only relatively high, but asymptotically approaching 1.

When we see that people do not immediately "take" in centipede games, or that they often cooperate in single shot Prisoner's Dilemma, or frequently make more generous offers in ultimatum or dictator games than Homo Economicus would, perhaps we should be humbled. If we don't believe that such behaviors are primarily an artifact of the observer's paradox (and personally, I suspect that they largely are), we should consider whether alternatives to Homo Economicus might be useful. But one thing we absolutely should not do is conclude that human behavior is unpredictable or illogical or "irrational" or any of that.

Show me a voter who votes for his least-preferred candidate, a consumer who demands to pay more than the sticker price, a pedestrian who goes out of his or her way to make their journey more laborious, or a leader who rolls around in peanut butter in the midst of a crisis, and I'll reconsider my position on "the human element". Until then, color me skeptical.

*I often wonder why it is that most every reasonably well-educated person learned at some point that no argument can be resolved by appeals to "human nature", yet those same people use words and phrases like "the human element", "humane", and "inhumane" as if they are any less problematic.

Sanctions as a Commitment Generating Device

2012-02-07T20:48:00.005-05:00

Sanctions are often discussed as alternative to the use of military force. There's something a little strange about this, since we know that sanctions are often a precursor to the use of force (see here).

The simple interpretation of this association is that sanctions don't work, and when states really care about the issue, they sometimes resort to an instrument that does (read: military force).

In the American Foreign Policy class that I'm currently teaching, I present a model that provides an alternative explanation. One that does not require us to assume, falsely, that sanctions do not work.

The challenger must first decide whether to impose sanctions. Regardless of whether she does or not, she then issues an ultimatum. If the defender gives in to this demand, the game ends. If the defender resists, the challenger must decide whether to go to war or not.

There are a number of key assumptions here that are so common within the crisis bargaining literature that I'm not going to dwell on them. But there are a few that are unique to this model, and so warrant some justification.

First, I assume that sanctions can be imposed in stages. This is not always assumed in the formal literature on sanctions. But it's pretty evidently true empirically, so I'm pretty comfortable with this.

Second, and more importantly, I assume that the costs of war inherently include the foregone benefits of continued economic exchange. There is some dispute over whether countries always and everywhere sever trade ties during war (see here), but it is pretty typical to assume that, as a general rule, they do. If this is true, the maximum costs that can be imposed through sanctions must needs be smaller than the minimum costs of war, as the latter subsume the former. Put differently, since one of the reasons war is costly is that it disrupts trade, severing economic ties (which is basically what sanctions do) is like putting a down payment on the costs of war.

Stopped for a sec to make sure that's really what I wrote, didn't ya?

Surely that isn't why states impose sanctions, even if it does also happen to have that effect. Right?

Well, I'm not so sure.

What the model suggests is that when the challenger has an inherently credible threat to use military force if its demands are not met (i.e., finds the status quo unacceptable relative to war), she will never impose sanctions. That doesn't mean she'll always go to war. Sometimes, she'll issue demands that risk war, but end up being lucky enough to get what she wants from the defender without having to make good on her threat to attack if she doesn't. But no matter whether the defender chooses to concede to the challenger's demand, when the challenger's threat to attack would be credible regardless of whether she imposed sanctions first, sanctions are not imposed.

On the other hand, when the challenger would not have a credible threat to use military force even if she imposes sanctions (i.e., when she finds the status quo sufficiently attractive relative to war), the challenger always imposes sanctions. And always receives concessions as a result. How large these concessions are depends on how much additional pressure the challenger can put on the defender by ratcheting up the sanctions. The coercive power of sanctions here is driven entirely by the difference between how much harm is inflicted upon the defender initially compared to how much harm can be imposed after the sanctions are stiffened. The model suggests that making the first round of sanctions imposed on a country as crippling as possible undermines the ability to engage in successful economic coercion. The defender has to know that giving in will spare him the worst of what awaits him should he continue to resist.

Finally, when the challenger finds the status quo relatively unattractive, but not entirely unpalatable, sanctions can generate a commitment to use military force. Here, sanctions grant bargaining leverage not directly, as above, but by rearranging the challenger's incentives in a such a way that makes threats to go to war credible. In some cases, war will indeed follow sanctions. In others, sanctions will be followed by demands that the challenger would not have dared make if she hadn't first tied her hands by placing a down payment on the cost of war, but the defender will concede to those demands and so war will be averted.

Overall, then, the model suggests that sanctions can either be seen as an alternative or as a precursor to military force. Which of these relationships obtains depends upon the challenger's evaluation of the status quo. The positive correlation between sanctions imposition and the use of force may not represent the failure of sanctions. In fact, paving the road to war may, at times, be their very purpose. In such cases, the hope is still that war will not be necessary. But the only way for a challenger who is known to find the status quo reasonably attractive to extract concessions from the defender is to first make herself more willing to go to war. She might do so through imposing sanctions, since sanctions reduce the opportunity cost associated with war. Should she use sanctions to generate a credible commitment, she will, at times, end up fighting a war that would not have occurred had she not imposed sanctions. But had she not imposed sanctions, she'd also have failed to extract any concessions from the defender. Unless the challenger subscribes to the "peace at any price" view of foreign policy, that might well be seen as the lesser of two evils.

If you're interested in the details, see this post for more about the class, as well as a link to the lecture slides.

My American Foreign Policy Class, Part I

2012-02-07T20:47:00.000-05:00

Nearly half of the class is going to be dedicated to assessing what USFP might look like if we believed that policy was made with no other consideration in mind but serving "the national interest" (under the assumption that such a beast exists). There will then be a few weeks on why we might not expect USFP to actually look like this, due to various domestic political pressures. The final third or so of the course will consist of a discussion of USFP from 1945 up through today, interpreted through the lens of the theoretical expectations developed in the first two parts of the course.

Given this approach, it would probably make more sense to wait until I've finished the entire section on optimal foreign policy before posting, but that's still a few weeks off. So here's the first set of slides, which covers the basics of crisis bargaining, militarized coercion, and economic coercion.

Miscellaneous Links (updated again)

2012-02-07T09:34:00.002-05:00

Below the fold.

1. Saideman offers some thoughts on the Afghanistan mess. Key quote: "Did we know this in 2009 when Obama was trying to decide on a new Afghan policy? Yes, which is why he took a great deal of time and care to consider it. Why didn't he just fold the tents then? It would have been politically costly at home and abroad."

2. So does Zach Jones. Very much agree with all of Zach's points. The Taliban may not be loved, but that doesn't matter. So long as the alternatives for ordinary Afghanis are as poor as they are, US resolve wanes as it has been, and ISI support for the Taliban keeps up, there is no doubt about how this all will end.

3. UN's inaction emboldens Syrian regime. Suggests leaders choose how much restraint, if any, to show based in part on likelihood that their behavior will provoke intervention.

4. Regional implications of Syria's conflict. I've been saying for some time now, Iran-Israel isn't the real story. It's Iran-Saudi Arabia.

5. US drones targeting rescuers and mourners. Andrew Sullivan likes to say that if Obama was a Republican, he'd be on Mount Rushmore by now. I think he's right that mere partisanship prevents conservatives from realizing how pleased they should be with Obama, but it is equally true that partisanship has blinded many liberals to just how thoroughly Obama has come to embody everything they hated about Bush's foreign policy.

6. US focus on regime change closes off option of negotiation with Iran. Which may be the goal anyway.

7. MAD or MADE? No evidence is provided in support of claims linking Pakistan to China, but still an interesting argument overall about economic interdependence and parallel to nuclear deterrence.

8. Another go on Beckley, GDP, GDP per capita, etc. I could not possibly agree with Dan Nexon more if I tried. Obviously, GDP is not the whole story. Beckley's response to Voeten does a nice job of pointing that out. But GDP per capita is still a poor substitute. If someone wants to graph (GDP per capita - 10,000)*(ln(pop)) for the US and China over time, maybe we'll be able to have a reasonable discussion about trends in the distribution of "surplus wealth". But if we're all going to insist on simple measures, GDP per capita really isn't going to cut it.

9. Economic growth and democratic survival. Evidence linking the two is far weaker than many realize. Bonus: Ulfelder provides estimates of the likelihood that the three Arab Spring "success stories" will be democratic five years from now. Outlook pretty good for Tunisia. Appears to be good for Libya, but Ulfelder has doubts about that one. Egypt is a toss-up.

10. Worst advice you can give students when they go on the job market? Be yourself. Key quote: "The fact is, Dear Readers, 'yourself' is the very last person you want to be. Why? Because no matter where you are in your career, but most especially if you are just starting out, or (god forbid) a grad student, you are, as an academic, insecure, verbose, defensive, paranoid, beset by feelings of inadequacy, pretentious, self-involved, communicatively challenged, and fixated on minutiae." I'm a few years out of grad school, and I'd still say, "Guilty as charged."

11. Causal identification and policy relevance go hand in hand. Yes. I have long been irritated by those who argue against "overly" mathematical approaches to social science, which allegedly prevent us from asking policy relevant questions (see here, for example). Since when is asking a question itself interesting? If you can't provide a compelling answer, exactly how "relevant" is your study anyway?

12. Why we argue online.

The Political Impact of Audience Costs (updated)

2012-02-02T18:07:00.007-05:00

As I've said before, I have a number of concerns about the theoretical arguments about audience costs. I'm not sure why I should expect voters to behave in such a way as to give rise to them, and I'm really not sure we understand what behaviors we would observe if they existed. But let's set all of that aside for the moment and simply ask: what would the political impact of audience costs be if we thought that Tomz's experimental study accurately reflects the way real world voters vote -- at least in the US.*

There are lots of reasons to believe that respondents in survey experiments are more inclined to punish leaders for not sending the nation's sons and daughters off to war than real world voters (whose actual sons and daughters might die) would be. But forget that. Let's even set aside what other experiments find about how easily leaders can manipulate the public (for example, in a paper forthcoming at JoP, Levendusky and Horowitz find that the president can completely offset any punishment by explaining to the public that new information came to light). For the purposes of this post, I will assume that leaders who threaten to intervene in an ongoing crisis then fail to do so will see the approval of their handling of that crisis drop by 26%, the largest effect Tomz finds under any condition.

Note, that means that the study everyone cites as having validated audience costs arguments finds that, no matter how you slice it, a minimum of 74% of US voters do not punish leaders for getting caught in a bluff.

The point I want to make is that elections do not simply turn on the public's evaluation of the president's handling of crises that didn't escalate to war. Wars often fail to influence elections, provided they are not still ongoing at the time of the election. Consider, for example, the Persian Gulf War, which brought George H.W. Bush's approval ratings to unprecedented heights, yet nonetheless failed to keep him in office.

I will, however, make what I consider a rather implausible assumption that the only foreign policy issue is going to be this one crisis. We'll assume away the possibility of ongoing wars (you may remember that there was once a time that the US was not perpetually at war). We'll assume away debates between the candidates about how to handle ongoing issues, or potential future crises. We'll ignore differences in terms of proposed levels of defense spending, support for Israel, etc. To the extent that foreign policy influences elections in the simulations I'm about to describe, that effect will be entirely captured by whether the president was caught bluffing in a recent crisis.

What I will not assume away, however, is the importance of other factors besides foreign policy.

Here's the code for my very simple simulations.

set obs 10000

drawnorm approv growth ep, means (55, 2, 0) sds(8, 0.6, 0.005)

gen prob_reelecN = 0.01*approv + 0.1*(growth - 2) + ep
gen prob_reelecB = 0.01*approv + 0.1*(growth - 2) - 0.026 + ep

gen draw = uniform()
gen reelecN = 0
gen reelecB = 0
replace reelecN = 1 if draw < prob_reelecN
replace reelecB = 1 if draw < prob_reelecB

gen critical = 0
replace critical = 1 if reelecN==1 & reelecB==0

I create 10,000 observations. For each, I assign a value for presidential approval, economic growth, and unobservable factors.

Approval is distributed normally, with a mean of 55% and a standard deviation of 8%. This gives us an observed range of 25 to 86, quite similar to the actual observed range for election years.

Economic growth is distributed normally, with a mean of 2 and a standard deviation of 0.6, giving us an observed range of -0.3% to 4.2%. That doesn't capture the lows that we've unfortunately seen, but isn't too far off overall. And is certainly close enough for the purposes of this illustration.

The unobservable factors might represent campaign effects. In these simulated data, such effects will range from just about 2% for to approximately 2% against the incumbent. That's more or less in line with what some estimates suggest. Nothing huge, but certainly important enough to matter if the fundamentals are neither strongly in favor of, nor strongly pushing against, reelection.

I then construct probabilities of reelection both with and without bluffing. These probabilities largely reflect the president's approval rating (and all that that entails), but are also pretty sensitive to economic growth (with above average growth being rewarded and below average growth being punished). I assume that a decrease of 10 percentage points in approval of the president's handling of the one crisis that matters for evaluations of foreign policy will result in a decrease of 1 percentage point in the probability of retaining office. Thus, a drop of 26% in approval of the handling of the crisis translates into a drop of 0.026 in theprobability of reelection.

Next, I create binary variables indicating whether the president would be reelected in the event that he did not get caught bluffing in a crisis and whether he would if he did. These are random variables that take on a value of 1 with the probabilities described above.

Finally, I create a binary indicator for whether audience costs would be critical to the outcome of the election. This variable simply denotes whether the president would retain office if he is not caught bluffing but would be removed from office if he is.

This variable takes on a value of 1 in 2.5% of cases.

If, like me, you have concerns that Tomz's results represent something of an upper bound on plausible estimates of the size of audience costs, given that it seems that presidents can untie their hands relatively easily by telling the public that new information came to light; or that it is unrealistic to think that the only consideration that will weigh on voter's minds when thinking about foreign policy is whether the president made an empty threat to intervene in some crisis; or that foreign policy is unlikely to have even this much influence if there is not an ongoing war (in which case the war will predominate foreign policy evaluations), then you probably ought to conclude that audience costs do not matter. Perhaps it is true that some voters do think this way. But it's far from clear why that should matter much to leaders.

You don't have to agree with me that we as scholars can go ahead and forget all about audience costs, but you should at least be prepared to articulate which assumptions of my argument you reject.

*The audience costs proposition should apply to all states, with democracies of all types exhibiting greater audience costs than non-democracies. But experimental studies only survey US respondents, and it is only in presidential systems that the public directly elects the executive anyway, so for the rest of this post, I'm going to limit my attention to the United States.

R2P and Consistency

2012-02-01T19:11:00.006-05:00

Anne-Marie Slaughter argues:

If you believe, as I do, that R2P is a foundation for increased peace and respect for human rights over the long term, that each time it is invoked successfully to authorize the prevention of genocide, crimes against humanity, grave and systematic war crimes, and ethnic cleansing as much as the protection of civilians from such atrocities once they are occurring, it becomes a stronger deterrent against the commission of those acts in the first place.

Dan Trombly says:

Precedent is only a deterrent if the power and will of a country to enforce that precedent at all places and times remains constant. Given that everybody knows the resources and willpower of countries supporting R2P are finite, countries will generally (and correctly) bet that repressing a local effort at regime change is a more reasonable policy than succumbing to a revolution for fear of being deposed.

I think both get it wrong in these passages, though overall I think Dan gets the better of this exchange.

One of the things that must be true for R2P to deter atrocities is that interventions today increase the belief held by those who would commit tomorrow's atrocities that if they do so, the US (and/or some other actor) will intervene against them. Slaughter claims that every intervention in the name of R2P increases this belief. That could be true, even if the US did not always intervene when atrocities were being committed. But that doesn't mean proponents of R2P should be so quick to dismiss critiques focused on consistency and hypocrisy.

For the sake of argument, let's assume that leaders update like good Bayesians.*

Let n denote that the US engages in "normal" politics (uses force to promote economic or geostrategic interests, and is willing to sacrifice its normative goals in pursuit of said interests.)

Let r denote that the US has a sincere commitment to R2P. (We'll set aside what precisely that means).

Let n_ix be the updated belief that the US engages in "normal" politics, given that the US has intervened in country x.

Let ix denote a US intervention in country x.

By Bayes' Rule, n_ix = (pr(ix|n)*pr(n))/(pr(ix|n)*pr(n) + pr(ix|r)*pr(r)).

Suppose that 0 < pr(ix|r) < 1, indicating that the generic probability of an intervention would always be positive but also always be less than 1, even if the US had a sincere commitment to R2P.

If, and only if, pr(ix|r) > pr(ix|n), then n_ix > n. That is, if and only if the probability that the US would intervene in country x if the US was sincerely committed to R2P exceeds the probability that the US would intervene in country x if the US was engaged in "normal" politics, then intervention in x would increase the belief of a good Bayesian that the US is sincerely committed to R2P, and thereby decrease the likelihood of future atrocities (at least at the margins).

That tells us that the US need not intervene everywhere in order for R2P to have meaning, contrary to Dan's claims. But it also tells us that the US needs to be seriously worried about looking disingenuous if it expects to deter future atrocities, despite the dismissive attitude many proponents of R2P have expressed towards such concerns (I'm not citing anyone because I don't want to pick on anyone, but I think we've all seen people brush the consistency criticism under the rug as though it were no big deal).

As Dan Trombly says,

Would not the US intervention in Syria just be a cynical ploy to deny Iran an ally in the lead up to a US or Israeli strike on the country? Or an attempt to weaken Russian influence in the Eastern Mediterranean to secure Israeli control of oil and gas fields and trade routes? Or another attempt to distract the Arab world from the behavior of America’s Gulf monarch clients?

Replacing the generic x with an S for Syria, Dan is suggesting that pr(iS|n) is pretty large. Is it larger than pr(iS|r)? I don't know. But the very best you can say is that an intervention in Syria would have a very modest impact in terms of deterring atrocities. And I'm skeptical about even that much.

I think Dan nails it here:

If the US is so committed to R2P, why doesn’t it do more to Bahrain or Saudi Arabia instead of signing arms deals with them?

All it would take is one act that no one would ever expect the US to take if the US was engaged in "normal" politics to convince the world that the US has a sincere commitment to R2P (i.e., to get what is currently a large n to be replaced by a small n_ix). Ceasing to prop up brutal dictators who just so happen to be vital to US economic and strategic interests would be a good way of doing that. Far more so than intervening in Syria, which the US would probably be willing to do even if it had no real commitment to R2P.

I'm not holding my breath while I wait for the US to sever ties with the Saudis though.

* Some people will no doubt question this, but if they update in a manner roughly consistent with Bayes' Rule, allowing for some (random) error, then nothing changes. If you think they update in a fundamentally different way, that's a problem for my argument. I've never seen any evidence in support of such a claim, myself.

Diplomacy and Cheap Talk

2012-02-01T16:24:00.011-05:00

One of the less salient arguments in Fearon 1995 was that we should not expect wars to be averted by diplomacy. Wars occur primarily* because of the incentive to misrepresent private information. It is not simply the presence of incomplete information, but the fact that states have no incentive to tell the truth.**

At least, so says Fearon.

A number of scholars (see here, here, here and here) have since claimed that diplomacy does indeed have the power to prevent war. This is sometimes framed as a criticism of the argument that cheap talk is uninformative (especially by Sartori). I find that rather misleading, though, because what these authors are in fact arguing is that diplomatic statements are not cheap talk.

There are various ways in which it might be costly for states to claim to be stronger/more resolved than they are. The one that has been most thoroughly explored by the literature is reputational (see Sartori, Guisinger and Smith, and Kurizaki). States that claim to be resolved and then back down suffer a very real, if indirect and intangible, cost -- they lose the ability to be believed when claiming to be resolved. That's not "cheap" talk in any meaningful sense.

In a paper forthcoming at ISQ, Ramsay provides another example of "cheap talk" that isn't actually cheap revealing information. In his model, each state adopts a literal bargaining position. If the distance between what each side says it is willing to accept is too large, war occurs. Thus, putting forth an extreme bargaining position risks war. That is a very real and tangible disincentive for making false statements about what you would or would not accept.

See also this experimental study and the thoughtful discussion of it in this blog post. The mechanism here may be something different altogether. Of course, it's hard to know, since students often behave the way they think they are supposed to behave in experiments, regardless of how hard we try to induce incentives...

Anyway, I think the theoretical arguments of Sartori (and others) and Ramsay are interesting, and the mechanisms they describe strike me as fairly plausible. There is ample evidence that states worry about their reputations -- not only for resolve, but also for honesty -- just as there is evidence that states believe that bluffing is dangerous. There is also abundant evidence that states believe that diplomacy reveals information and thereby prevents war, as Kurizaki details nice in various working projects. If you subscribe, as I do, to the view that any activity political elites invest a lot of time and effort in is probably an activity that serves a useful purpose for them, however much it seems like that ought not be true to political scientists, then your reaction to Fearon's claim that diplomacy doesn't prevent war is probably one of skepticism.

However, that same line of reasoning points to another puzzle.

(The fact that "something is missing" from existing accounts does not invalidate them, of course. I'm the last person to say that simplification is itself a sin. I just want to pivot off these arguments to point to something else that we might want to explore.)

If we "know" that diplomacy works because states wouldn't bother with it otherwise, then I can't help but wonder why face-to-face interactions remain so central to international diplomacy. If states are discouraged from bluffing by concerns for their long-term reputation, or because the statements they make influence the probability of war, then why would it matter how the messages they send to one another are conveyed? Why not simply send an email and be done with it? Historically, there may have been few other options, but in the modern era, there are far easier, quicker, cheaper ways of communicating than attending summits or maintaining embassies.

Of course, one possibility is that face-to-face interactions are little more than political theater, while embassies primarily serve as an exercise in mutual hostage taking. But leaders certainly behave as if they learn valuable information from diplomatic summits and from their embassies. And I don't think the latter is simply because spooks so often pass themselves off as diplomats.

This got me to thinking -- what if the physical act of communicating with representatives of the other state might convey information that the content of their statements does not?

Suppose, for example, that diplomats, much like poker players, have "tells". Then if we take the canonical ultimatum crisis bargaining model that Fearon used to demonstrate that diplomacy does not matter and introduce the assumption that there is a small but non-zero probability that a state inadvertently reveals its true type when communicating with its opponent, regardless of the actual content of their statement, then diplomacy would reveal information, but it would be essential that diplomacy take place in person. It's much harder to read a person over email, after all.

But what would this really imply?

It has become something close to conventional wisdom in IR that if the primary cause of war is information problems, anything that facilitates the credible revelation of private information must obviously prevent war. But this simply isn't so, as Scott Wolford and I demonstrate here. Information revelation, especially if it is imperfect can either promote or prevent war. And both leaders and diplomats must surely realize that their ability to read one another's tells is not perfect. What I'm suggesting is that face-to-face diplomacy generates exogenous signals that are credible ("tells" by definition are inadvertent) but also quite noisy (and so would not allow the receiver to update to certainty -- i.e., would mitigate but not eliminate uncertainty).

If I'm right about this, if face-to-face diplomacy matters because it creates the possibility of inadvertent information revelation, who would we expect to benefit from diplomacy? It is sometimes taken as given that all states have an interest in diplomacy, but I'm not sure that follows. Engaging in diplomacy means increasing your odds of learning secrets the other side wants to keep, but it also means increasing your odds of accidentally revealing your own secrets. The full implications of that are not obvious to me. There's potentially some very interesting dynamics there that need to be sorted out...

I would say, though, that one thing this suggests is that we could learn a lot from comparing how students react to diplomatic statements from other students, delivered in person, to how students react to statements that they know were generated by a computer algorithm (or even those that were sent by other students, but where the receiver and receiver did not interact face-to-face). We know that people react to ultimatum offers differently when they come from other people versus computers (alas, I can't find the citation at the moment). Do diplomatic statements work the same way?

Were it not for the fact that I'm already working on too many papers, I'd be tempted to run with this. I'd start with a formal model, which I'd use to assess the relationship both between diplomatic ties and war and to develop expectations about who establishes ties with whom, then evaluate the observable implications in terms of who does in fact have ties with whom. If no one has explored this idea more fully a few years from now, maybe I'll come back to it...

*The cheap talk argument does not apply to commitment problems. Yet neither does diplomacy. So the fact that information problems are but one of the mechanisms Fearon laid out (albeit the one that is often thought to be more prominent) doesn't undermine the point of this post.

**We sometimes shorthand this by saying war is caused by uncertainty. In the right context, that's fine, but it's important to note that this uncertainty is not simply a random nuisance that bedevils both states. Rather, we might expect it to arise as a direct result of the decision states make. Uncertainty is bad from the perspective of those of us who would like to see fewer wars, but, sadly, it does not appear that we have any reason to assume that states share our desire to eliminate uncertainty.

January Roundup

2012-01-30T13:33:00.000-05:00

January's five most viewed posts:

1. How Not to Think About China's Rise

2. Is China a Security Threat to the United States?

3. Stop Asking "Do Sanctions Work?"

4. Actually, Dr. House is Not My Role Model

5. How Not to Write Referee Reports

Good Reasons and Bad Reasons to turn to ABM

2012-01-27T23:10:00.000-05:00

A few people have asked me my thoughts on agent-based modeling. I do have some, and you can find them below the fold, but let me preface my remarks by saying that you should no more trust what a game theorist says about agent-based models than you should trust what an agent-based modeler says about game theory -- which, as I'll argue below, means you probably shouldn't trust it very much.

For that reason, I have asked Kyle Joyce, a good friend of mine who teaches Complex Adaptive Systems at UC-Davis, to write a guest post responding to this one.

First of all, some terminology. If you're interested enough in this topic that you're reading this blog post, you probably are already familiar with all of this, but it can't hurt to review.

Model - an object, used to represent reality. By necessity, it captures some aspects of reality in detail, while simplifying or neglecting other aspects.

Empirical model - a particular type of model. Used to summarize patterns of association among data. Examples: structured, focused comparisons; process-tracing; statistical models.

Theoretical model - a particular type of model. Used to evaluate the logical implications of a set of assumptions. Examples: mental/informal models (most common type), computational,game-theoretic.

Computational model - a particular type of theoretical model, where in the implications of the assumptions are established not through deductive logic, but are arrived at through inductive inference. Examples: agent-based models, numerical approximation of equilibria.

Agent-based model - a particular type of computation model, where independent agents follow a set of rules and the analyst observes the outcome that obtains, often analyzing simulated data using standard statistical techniques.

The problem facing all models, as Scott de Marchi eloquently argues, is the Curse of Dimensionality. The goal of analyzing a model, be it theoretical or empirical, game-theoretic or computational, is to better understand the simple world it represents, in hopes that this helps us better understand the world in which we live.* The Curse of Dimensionality refers to the idea that there are infinitely many ways to specify our models. We must always worry that we have overlooked important variables in our empirical models. We can never know if we have made simplifying assumptions in our theoretical models that were far less innocuous than we believed them to be.

Herein lies the appeal of computational models. If you aren't constrained by the need to be able to identify a closed form solution analytically, you can set up a more complicated model. You can vary more things. You can make less demanding assumptions.

It is precisely that appeal that makes many game theorists skeptical of most applied work done with ABMs. Note, I chose my words carefully here. The skepticism to which I refer concerns the actual work that is done in practice by many agent-based modelers, not with the method itself. ABMs are often seen as "models for people who don't actually believe in modeling."

That is not a criticism of the method itself. It would be ridiculous to tell someone that they should not use a method because many of those who do use it use it poorly. I have seen such arguments made about game theory, and I find them infuriating.

What it means is that there are good reasons to turn to ABM, and there are bad reasons. If you think carefully about what the tool can do for you, what it cannot, and, perhaps most importantly, what you can get from other tools, then you can unlock the profound potential of ABMs and make a powerful contribution to the literature.

I have met grad students who have told me that they are interested in ABMs, but, when I ask why, simply answer, "Because I don't think the world is that simple."

Neither do I. Believe it or not.

It is not necessary to believe that your assumptions are literally 100% in order to believe that you can learn something of interest from your model. I must be starting to sound like a broken record at this point, because I feel like I say this in a lot of my posts, but the extent to which your average social scientist fails to understand the power and necessity of simplification continues to stun me.

If you are turning to ABM because you think simplification is inherently wrong, odds are you aren't going to produce particularly interesting results with your ABM either. The Curse of Dimensionality does not go away when you move to the world of computation. It becomes more manageable, sure. But it's still a real beast. Ultimately, if you haven't got much intuition for how the output of your model is likely to change when you turn this knob versus that one, you are going to do piss-poor work no matter what. That, or you are going to be the type of person who takes so long to finish projects that you'll never defend your dissertation, or never publish after getting a job, or never make tenure. There is no escaping the need for you, the analyst, to do a lot of the heavy lifting. If you turn to ABMs because you like the idea of being inductive, of taking your hands off the wheel and letting the computer find the answer for you...you're doing it wrong.

A related bad reason for turning to ABMs is the belief either that "people are not rational" or "the world is not in equilibrium". People who make these statements are demonstrating that they do not understand what these words mean to game theorists. See here for my take on "rationality."

In ordinary language, equilibrium is often taken to mean a stable point, or a situation that is unlikely to change. It is therefore tempting to think that game theoretic models are only useful for analyzing static situations. I think that's why so many people say that game theoretic models are incapable of analyzing "dynamics", though I understand that some people use the term "dynamics" to mean something other than its every day meaning would imply. Regardless of what "dynamic" does or does not mean, it is simply not true that game theoretic models are only suitable for analyzing situations in which nothing ever changes.

In game thoretic parlance, an equilibrium is a set of strategies (and beliefs, depending upon the type of equilibrium) such that no player has an incentive to unilaterally deviate from their strategy profile. A strategy profile specifies what actions a player should take at each possible decision node, even those that are not in fact reached in equilibrium.

That's a mouthful. What does that mean?

Well, if you send your solders into a battlefield, and give them clearly defined rules of engagement, and the enemy sends his soldiers onto the battlefield, with clear rules of engagement, the one thing that is just about certain not to result is stability.

If you observe that reality is messy, that stuff is constantly changing, that situations are always evolving, and from this infer that the world is out of equilibrium, you have demonstrated that you don't know what the word means.

If you want to conclude that the world is not in equilibrium, you need to tell me how it is that you know that we are observing actors behaving in a manner inconsistent with their equilibrium strategy profile. In order to do that, you first have to know what their equilibrium strategy profile is. In order to know that, you first have to know what the person wants and what they know about the world, and a whole lot else besides. I have not yet met anyone who knows these things. If anyone did, I can't imagine why they'd bother analyzing models of any kind, since they'd apparently already know everything worth knowing.

Basically, what I'm saying is that you should not turn to ABM because you think game theorists make untenable assumptions. And you probably don't want to make a habit of making ignorant statements about what game theory is incapable of in each and every paper you write, the way many ABMers have. I understand that the response you're likely to get from reviewers when you send your papers out will often be, "why aren't you modeling this with game theory?" So you probably do have to have a "this is why my model is not a GT model" section in your papers. That's unfortunate, but I get it. But, given that this is the way our discipline works right now, keep in mind that if the reviewer is the type of person who thinks it is appropriate to say such a thing, they probably are not the type of reviewer who will overlook the cheap shot criticisms that so many ABMers put forth.

Like I said, though, these aren't critiques of the method itself. These are reasons why many game theorists see that a paper contains an ABM and groan. They've learned through experience that this means they are going to have to wade through several pages of being told how complex** the world is, how evil simplification is, and how inadequate game theory is.

If you take just one thing away from this post, let that be it. We've heard that song and dance before. And we're not buying it. Learn a new tune.

Happily, there are a lot of other songs you can play with your instrument. The Curse of Dimensionality is real. And though I think it's easy to oversell how damning it is for those with a refined intuition about which assumptions matter and which ones don't, and to undersell how much of a problem it will always be for those who lack such an intuition, regardless of which method they choose, I really do not doubt that ABMs offer a powerful way of addressing this problem. Sometimes, we are interested in questions that simple models can answer perfectly well, thank you. Sometimes, we are not. The trick is knowing which is which.

There is no doubt in my mind that ABMs have a lot to offer political science. I use (relatively simple) computational models all the time. Many game theorists do. The manuscripts I have sent out for review have, for some years now, all ultimately contained proofs for my theoretical propositions. But I often struggle with a particularly complicated proof and use simulations to numerically approximate equilibria first and give me a sense of whether I'm heading in the right direction. So far, that has always enabled me to ultimately construct a complete proof of what I wanted to say, but it's not hard for me to imagine being interested in a model where I couldn't quite get there in the end. In such a situation, I would turn to computational modeling.

I wouldn't, however, go on any rants about "rationality", "equilibrium", "complexity", "dynamics", "adaptation", or any of that first. I would just make the exact types of assumptions I've always made and rely on statistical analysis of simulated data generated by my model rather than relying on algebra and proofs. Your mileage may vary. But if nothing else, be aware that the argument in favor of ABMs can, in principle at least, be divorced from all of that.

*Some people have convinced themselves that empirical models are not models at all. This is bunk. But I assume that anyone reading this post is already on board with that and is more interested about what I have to say about different types of theoretical models, so I won't dwell on that point.

**Miller and Page provide a definition of complex systems that thy contrast with complicated systems. It is a useful distinction. Unfortunately, it seems to me that in practice most people don't go much past arguing that the world is complicated before they go on to say that they must use an ABM to study it.

Miscellaneous Links (updated)

2012-01-26T20:05:00.003-05:00

Below the fold.

1. China has already surpassed the US. According to CINC scores.

2. Kindred's take on the whole thing. I should have expected him to remind us to think of system dynamics. Excellent points, all around.

3. I suppose I'll even link to Walt's take. Not that I doubt any of you who might care would have missed it. Incidentally, I sure would love to have someone explain to me why Walt wears his "realism" on his sleeve, as a part of his personal identity, when he so often writes stuff like this, "Moreover, five decades of misguided policies have badly tarnished America's image in many parts of the world, and especially in the Middle East and Central Asia." Perhaps I just do not know what "realism" means.

4. Drezner weighs in again, mostly responding to Walt. An excellent post. The point about the nostalgic fantasy deserves to be repeated often and prominently. And it's far from clear why it's bad to overestimate the capabilities of the US but not of China. One might even think that the goal should be to get as accurate a picture of the distribution of power as possible...

5. Of course, it's not clear realists do either. It's a theory about how things actually are. Or maybe it's not. Because it's really more of a foreign policy orientation than anything. Or maybe it's not, since realists don't advocate the policies that realism prescribes. Or maybe....

6. PM's Question Time adds some thoughts re: House and rational choice. Yeah, I really should have mentioned Gary Becker.

7. How Iran Learned to Stop Worrying and Love Being Bombed? Dan Trombly offers some criticism of the coercive power of aerial bombardment. And uses the phrase deus ex demokratia. For that reason alone, this is one of my favorite posts ever. Second best line: "In other words, short-term limited strikes are unlikely to impose severe enough costs to make Iranians anti-war or anti-regime."

8. Perhaps the Israelis really are going to strike Iran. I go back and forth on how unlikely I think it is that they will not. I still think it is relatively unlikely, but storm clouds do seem to be gathering.

9. Zach Jones on Org Mode, transparency and computational modeling. I'll finally be posting some thoughts about ABMs tomorrow myself.

10. Scott Wolford summarizes his class' discussion of commitment problems and war. Scott has been blogging his International Security class. I hope he continues to do so. This is a great service for those of us who can't be there.

11. Josh Goldstein on the Sunni-Shi'ite divide. Goldstein suggests it's as much about oil as anything. I think he's right.

12. Jay Ulfelder on electoral systems engineering as conflict resolution. Jay makes two great points. We have very little basis for drawing inferences, and we have very little reason to believe we can get states to adopt the electoral systems we think are best for them anyway. Of course, the idea that we don't know what we're doing, and couldn't implement it if we did, has never stopped political scientists from being convinced that it is their job to prescribe policies and thereby solve the world's problems.

13. This is what bargaining looks like: The Onion edition.

Actually, Dr. House is Not My Role Model

2012-01-25T16:54:00.004-05:00

Vikash Yadav over at the Duck writes:

House is the most rational person in the world; House is a complete drug addict. These two statements are not a contradiction within the parameters of the show. House is a calculating, self-interested, rational utility maximizer par excellence. His utility is pleasure and his pleasure is avoiding pain... and of course getting more pleasure. He is Bentham's man; he is John Stuart Mill's homo-economicus; he is a neo-liberal fantasy in the flesh. House is not a complete human being by any stretch of the imagination and yet this is the human being idealized by rational choice theorists.

Clearly, this is meant to be a light-hearted post. I get that. And I don't have any against Yadav. I usually find his posts to be interesting and insightful. What follows is most decidedly not intended to be a criticism of him, but of the view he expresses here -- which I think is shared by many. I'm simply using this post as an impetus to beat a horse that, to my continued dismay, seems to still have some life left in it.

First, the "rationality" of "rational choice theory" is not, not, not synonymous with Homo Economicus. Of course, this point has been made many times before, and I'm sure that repeating it yet again won't do anything but temporarily ease my frustration. Whatever prevents people from getting it right, it clearly isn't a lack of readily available information that would dispel such a myth. But I guess I just can't help myself.

I am, after all, only human. Sometimes, I act based on emotions. I'm prone to cognitive biases. My self-control is not what I desire it to be. Each and every day, I make decisions that would be sub-optimal, if my one and only goal was to maximize my material well-being. I even give money away to charity, and not because it drops me down into a lower effective tax bracket or anything like that.

But, then, perhaps this only goes to prove that I'm an apostate. Not only am I not a "neoliberal fantasy made flesh", I do not even aspire to be. So I guess I'm not qualified to speak for "rational choice theorists," as it would appear that I am not actually one myself. It's funny though, because I've always assumed that I am exactly the type of scholar people have in mind when they refer to "rational choice theorists". After all, I'm both a game theorist and a self-appointed "rational choice" apologist. But now that I know what "rational choice theorists" idealize, it's clear that I don't qualify.

Either that, or we need to admit that the view of "rational choice theorists" held by many corresponds so poorly to the population of scholars it presumably describes that we need to seriously reevaluate how we use that term. Or retire it altogether.

I've got three issues with statements such as the one Yadav made, and the broadly shared view that it reflects.

1.) The "rationality" assumption necessary for any model in which actors are assumed to maximize their purely subjective utility says nothing about what people want or how much information they collect about how their actions will translate into outcomes. It assumes that actors have preferences that won't cycle (HPWC) and THAT'S IT.

2.) You can't get very far from the simple starting point of HPWC. To build a model of any use, you clearly must add some specific assumptions about who the actors of interest are, what they want, what they know, etc. But it is not true that one must, or even that all practicing game theorists do, assume that people behave like Homo Economicus. See here, where honor is assumed to be an important determinant of state behavior. Or here, where actors choose how much effort they are willing to exert to ensure that they receive accurate information before making critical decisions. Or here, on why people might exhibit an interest in politics primarily because all the cool kids expect them to. Or here, where two of the most prominent game theorists in economics analyze a model in which a single individual is assumed to be in competition with himself (read: has problems exerting self-control). I could go on, but hopefully you get the point. Either we have to admit that "rational choice" does not mean what most people think it means, or we need to admit that the scholars most people view as "rational choice theorists", or simply "rationalists", do not in fact correspond to the, quite frankly, absurd and inaccurate stereotypes about them.

3.) Of course, while there are many examples of "rational choice theorists"* making assumptions about the actors in theirs models that fail in some way to live up to the demands of Homo Economicus, some models are indeed built upon assumptions that look a whole lot like Homo Economicus. I don't deny that. But even so, I think Yadav, like so many others in this field, is making the mistake of assuming that because a theorist has built a model that contains a certain assumption, this means that they genuinely believe in that assumption. He even seems to think that scholars who are interested in coming to better understand how Homo Economicus would behave in a given situation necessarily think that we all ought to aspire to being more like Homo Economicus.

I would ask Yadav then if he has ever accused the good people of Rand McNally of believing that the Earth is flat. Or whether he thinks that most physicists are firmly convinced that we live on a frictionless plane. After all, they spend a whole lot of time building and analyzing models that contain such assumptions. So they obviously must think such things are true in reality, right?

We use models to help us understand the world better. We do not have to believe their assumptions in order to think they are useful. Yet again, I'm going to plug the work of Clarke and Primo (book, article). If and when the day comes that the typical political scientist is familiar with even the gist of their argument, I may actually stop referring to myself as cranky (the unhealthy obsession with dead bodies is here to stay, though.)

Now, admittedly, Yadav referred to but two scholars by name: Bentham and Mill. I'm not going to get into an argument about whether it is fair to say that House perfectly embodies the normative ideals these two men expressed. It doesn't seem at all clear to me that he does, but whatever. The important point here is that many people hold views about the scholars who are working in the fields of economics and political science today that look quite a bit like the views Yadav expressed concerning "rational choice theorists." Whether Yadav himself meant for his statements to apply to any living, breathing scholars, I don't claim to know. As I said at the outset, my goal here is not to criticize Yadav, whose contributions to the Duck I typically enjoy a great deal. It is to push back against common misperceptions of "rational choice theorists".**

*If it's not clear yet, I of course refer here to people I assume would be perceived as such. I can't actually say with any certainty whether they are or are not, given that those scholars who have been explicitly identified as such rarely possess the qualities that seem to define the term.

**Again, what I mean is "those scholars who seem to be the scholars people have in mind when they refer to rational choice theorists."

How Not to Write Referee Reports

2012-01-24T17:06:00.008-05:00

There's a lot of very good advice out there about how to do it right. See here, and here, as well as the links therein. But I'm a crank, so I'm going to focus on what not to do. (See also here.)

As Marc Bellemare says, you are not deciding the outcome of the process. You are providing advice to the editor. So you should ask yourself what the editor wants to know. I've never edited a journal, but my guess is the answer is some mix of: "Will people cite this paper?" and "Will people wonder what the heck we were thinking, letting this dreck in?"*

There are a few things that reviewers frequently focus upon yet do nothing to help the editor answer these questions.**

To wit:

1. Did the author forget to cite the referee and/or the referee's friends?

Unless your work (or that of your friends) is both relevant and influential, no one but you will notice. If you (or your friends) have some obscure piece that no one but you (or yo---you get by now that I realize no one really does this on behalf of their friends, right?) has ever cited that truly is relevant, you should still ask yourself, has the author overlooked some important point that your poor, neglected publication discussed? If so, go ahead and mention it. Otherwise, keep it to yourself. The author may not see your name, but the editor does. And s/he will likely note that you're that reviewer.

Not only is this behavior shameful, it's also non-strategic. If you want to take umbrage at the fact that the author didn't cite a piece that no one but you would have cited in their shoes anyway, do it with the really really good papers that you've recommended be published without revision. That might actually affect your citation count. How do you benefit from getting a paper rejected on this basis? Your ego isn't really that delicate, is it?

2. Did the author "leave something out"?

This is a huge pet peeve of mine. Most people in this field don't seem to understand what a model is. Or the fact that we are all modelers, even those of us who do not analyze formal theoretical models. (On that point, see Clarke and Primo. I cannot recommend this book highly enough. It ought to be required reading in every scope and methods class.)

Put simply, if the author sets out to persuade you of a specific argument, and you have nothing to say that would undermine the validity of said argument, shut up. Merely pointing out that a model has committed the sin of being a model is a waste of everyone's time.

By this, I mean that if you notice that a theoretical model has made an assumption that is not 100% true, before you say that the paper should be rejected on that basis, you should ask yourself whether there is good reason to believe that the results would be different if the author had not made that assumption. If not, you're just stating the obvious -- that the author's model has simplified reality.

If you are evaluating a quantitative piece, and you notice that the author has not controlled for your favorite variable, ask yourself this: is there a story to be told about how omission of said variable is likely to induce omitted variable bias? A story that would pass the laugh test? A story that goes further than simply asserting that x2 also correlates with y? If not, you are again merely pointing out that the author's model is in fact a model. Congratulations.

If you are evaluating a qualitative submission, the same applies. I have no experience with this work, but I've got to imagine that referees here too have a tendency to recommend rejection because the author forgot some piece of trivia that has no bearing on their argument. Don't do that.

Of course, when I said "shut up", what I really meant was, "recognize that what matters most is whether the author has provided a plausible answer to an interesting research question". If you are recommending that the journal extend the author an opportunity to revise and resubmit the paper, and you think that the author will be more likely to persuade readers of the paper of their argument if they don't neglect z, fine. Mention that they left out z. But don't say, "This idiot left z out! Everyone knows that z is really important for explaining y. Much more so than x! Clearly the author knows nothing about y."

That just makes you look petty and/or stupid.

3. Did the author fail to address some question that you would have addressed?

This one is a bit trickier. No one should need to be told not to insist upon a reference of their work. In a world where social scientists did not have the same visceral reactions to, and unreasonable fears of, "reductionism" that much of the American right has with respect to Islam, people also wouldn't need to be told not to recommend rejection for every manuscript that "leaves something out". But this last point is a tough one, because there's a fine line between recommending that the editor reject a paper that will fail to connect with the target audience versus recommending rejection of a paper that would in fact interest the right people, simply because it does not interest you.

If you feel that the manuscript, along the way to answering q1, in fact raises q2, before you say so in your report, you might want to ask yourself whether q2 is something lots of other people would be interested in, or something you wish more people were interested in. That is, when you say to yourself that the author is asking the wrong question, are you simply saying that because the author has inadvertently stumbled in the general direction of one of your hobby horses? If you gave the manuscript to a talented grad student, would they want to know what the author was thinking in leaving q2 unanswered?

Suppose, for example, that the author is seeking to test the hypothesis that x1 is positively associated with y. The author has, of course, included a laundry list of control variables, even though there's no obvious reason to expect most of them to act as confounders. One of these control variables has a coefficient estimate that is significant but in the "wrong" direction (read: we thought we knew that it had the opposite effect, and apparently we were so confident in that finding that we're willing to call evidence to the contrary "wrong"). If there's a huge literature on the effect of said control variable, then even though said variable often really has no business being there in the first place, it would indeed be rather strange if the author did not mention, at least in passing, why they think they're getting that result.

In other words, yes, there are conditions under which the editor might want to know that the author failed to pay sufficient to attention to a question that you feel the manuscript raises.

But think of it this way -- if it's such an obvious point, the other reviewers will probably raise it. If you're not sure yourself why you're raising this point, maybe you should just let it go.

*Personally, I think most editors are especially focused on that last question. That's why the top 3 journals are filled with papers that use original data and/or clever methods to test really obvious hypotheses. Such work isn't going to change the field or end and debates, but it doesn't really inspire angry rebuttals or cause anyone to question whether the editors know what they're doing.

**At least, I really want to believe these things don't tell the editor much of what s/he wants to know. The system is even more broken than I think it is otherwise.

Is China a Security Threat to the US?

2012-01-24T13:28:00.010-05:00

Beckley's article has sparked an interesting debate about whether China is actually catching up to the US. Drezner provides characteristically insightful comments here. Erik Voeten over at the Monkey Cage weighs in on a few points here. Both are relatively receptive to Beckley's claim (though Voeten, like me, takes issue with some of the specific points Beckley lays out). What really caught my attention though is Voeten's point here:

Yet, this also points me to the biggest problem I have with this article: the intellectual strait-jacket it imposes that forces us to see world politics as a zero-sum game.

So, if we believe that China is catching up, then [according to Beckley], mercantilism is the only answer. This neither reflects the conventional wisdom nor prudent policy advice. China “catching up” does not equal China “taking over the world” nor does it equate a future world where Americans can expect to be catering to their Chinese overlords. This may be the view of some alarmists but mainstream analysts can think of many good reasons to maintain a liberal international economic policy even if China is catching up.

Just yesterday, a former student of mine who is interviewing for a position at the CIA stopped by to ask my thoughts about the future of US relations with China and North Korea. He's writing an essay as part of his application for the CIA. He asked me whether I thought the US was likely to go to war with China in the next twenty years. After a stereotypical amount of hedging, I eventually said something along the lines of, "no, not likely, though it's possible. What I'd be more worried about is proxy wars." He then asked, "So, do you think China is even a security threat to the US? At all?"

I'm not sure why my failure to say "Oh, God, yes, I'm up all night most nights worrying that World War III could start tomorrow" would make him think that I saw no threat at all. I did say that there was some chance of war, and that I was concerned about the possibility of proxy wars. But whatever. Outside the Ivory Tower, there are only two answers to any question: yes and no. And if the person you're speaking to thinks the answer is yes, and you hedge even a little, you obviously are in the "no" camp.

But, anyway, this made me step back and think, "What do we even mean when we refer to security threats?"

When I teach War & International Security, and I present models of commitment problems induced by rapid shifts in power, I emphasize that in these models, what prompts the declining state to attack is not a fear of being attacked themselves if they do not strike first, even though we often refer to the wars that occur in such models as "preventive" or "preemptive". What the declining state is doing is locking in as favorable a distribution of benefits as it can now rather than allowing the rising state to reach a point where it will be able to credibly demand a revised distribution. In short, the "threat" in these models is merely a threat of the loss of bargaining power.

If you listen to the rhetoric about China, the question seems to be whether China will attack the US (or Taiwan, or South Korea, or Japan). There is also fear of economic warfare (read: they hold lots of US debt something something took our jobs something something manufacturing base something something collapse of US economy). The more creative will mention cyber warfare (a threat that is easily overblown). All of these "threats" have something in common. They involve China engaging in a deliberate, intentional effort to do harm.

But China's rise could harm the US even if the Chinese did not intend for that to happen. You don't have to be a Malthusian (and I'm not) to recognized that resources are not infinite. Influence most certainly is not infinite. When crises break out, be they military or economic, the range of options available to the US will be more limited if China becomes as powerful as, or more powerful than, the US. The IMF and other global institutions will be less inclined to choose policies that reflect the interests of the US. The dollar may cease to be the world's reserve currency (which brings the US what Eichengreen calls "exorbitant privilege"). The list goes on.

Of course, Voeten is also right that the world is not strictly zero-sum. The US and China have both profited greatly from mutual economic cooperation. Few non-economists appreciate just how powerful specialization and exchange are for generating wealth. Resources may be finite, but our ability to make the best use of the resources that are available depends in no small part upon our willingness to embrace this idea.

Agreeing to move closer to the Pareto frontier doesn't eliminate the possibility of conflict over the distribution of benefits once there though. Frieden, Lake and Schultz make this point beautifully in chapter two of their brilliant intro text. For too long, IR scholars have seen conflict and cooperation as alternatives to one another. They are not. The question of how much stuff there will be to divide up depends largely upon how much cooperation there is. The question of how that stuff gets divided up has to be answered one way or another though.

There is little doubt in my mind that the world will be richer if the US and China cooperate with one another economically. There is also little doubt in my mind that a stronger China means a smaller share of the pie for the US. I don't worry all that much about China being a "threat" in the layman's sense of the word. But that's not the only kind of "threat" worth worrying about.

The million dollar (or yuan) question then is: if China is currently on a path that would allow it to surpass the United States in the next few decades (and I stress the "if"), would the US be better off getting a smaller share of a larger pie, or should it take actions to contain China and thereby ensure that it receives a larger share of what will be a smaller pie?

I don't know.

And neither does anyone who's being honest. Because the answer depends on how much smaller a share of how much larger a pie, as well as the costs associated with whatever policies the US enacted in an effort to contain China's rise. And you ought to be very, very wary of anyone who tells you that they can give you credible estimates of those things. Odds are, anyone who gives you an answer is selling you their ideology, though they may dress it up pretty first.

This is What Bargaining Looks Like

2012-01-22T13:23:00.003-05:00

US-American Indians Edition

The new issue of AJPS contains a fascinating article by Arthur Spirling, "U.S. Treaty Making with American Indians: Institutional Change and Relative Power, 1784-1911."

I saw an earlier version of this at Polmeth. And the paper is indeed primarily a methods paper. But readers of this blog may be interested in what the analysis the data Spirling compiles from the roughly 600 treaties negotiated between the US government and the American Indians tells us about bargaining.

There are precious few papers that seek to offer quantitative tests in support of the intuitive claim that states who enjoy greater bargaining power (either because they have a greater ability to win wars or a greater willingness to incur costs) in fact enjoy better divisions of disputed benefits. When people speak of the empirical evidence in support of observable implications of bargaining models, they typically mean in terms of when wars occur and how they unfold.

Thanks to the incredibly fine-grained data Arthur collects from treaties, we can finally test the claim that the distribution of benefits will change in response to changes in the distribution of power. To be fair, this isn't the most counter-intuitive claim ever to come out of a formal model. Myself, I find it so intuitive that I sometimes forget that there are people who doubt it. But there are.

In particular, Spirling compares the expectations of the bargaining literature to those of an historical-institutional framework that would emphasize the importance of changes in official US policy towards the American Indians as well as changes in the authority or resources granted to the president in order to implement such policy.

To be clear, this does not mean that Arthur takes into account all possible historical factors, or even necessarily all of the important ones. He focuses in particular on 1871, when Congress sought to strip the president of the right to treat with the American Indians directly. A number of historians and legal scholars have argued that this important date marked a fundamental change in US-Native American relations.

Yet Spirling finds no evidence of this.

He does, however, find strong evidence that the treaties negotiated with the American Indians reflected changes in the distribution of power, changes in the likely costs of war that the US would experience, and war outcomes.

There are a few notable "breakpoints" (stats talk for years in which we can say with a great degree of confidence that there was a sharp change in the trend). But, interestingly, 1871 doesn't appear to be one of them.

The y-axis in this figure is an estimate of how generous or harsh are the terms of a treaty, with 1 representing a treaty that was relatively favorable to the American Indian tribe and 0 reflecting a particularly harsh treaty.

(You don't need to know this to see the main point, but if you are curious, circles denote treaties that were ratified, while triangles represent treaties that were rejected, and squares represent treaties that were unratified).

The three change points that Spirling identifies occur in 1825, 1853, and 1866. He suggests that these correspond to a period of growing public support for dislocating the American Indians (culminating in the 1830 Indian Removal Act), which lowered the expected (political) cost of using force for the US government; the population bulge in the 1850s that increased pressure for westward expansion; and the end of the US Civil War freeing up resources that had previously been unavailable for prosecuting wars against the American Indians.

Spirling writes:

[W]e found strong evidence—statistical and substantive—that treaties became harsher over time, though it was not obvious that the 1871 alteration in bargaining rights made much difference in itself. Instead,we noted that the mid-1820s, and then the middle years of the nineteenth century, correspond to a marked shift in the nature of treaty making.We showed how this behavior corresponds to our theoretical priors about bargaining between parties as circumstances change: as the United States became more powerful in the middle of the nineteenth century, it was able to drive harsher bargains with enervated tribes.

It should go without saying, but lest there be any confusion -- none of this is in any way intended to legitimate or apologize for the brutal, inhumane, repulsive treatment of the American Indians by the United States. That a democratic state enacted such policies, with strong support from the public, continues to be a stain on US history, and an indictment of current thinking about democracy in political science, particularly international relations.

My point here is only that the historical record does indeed provide evidence in support of the claim that the distribution of benefits shifts in response to shifts in bargaining power, as reflected both by changes in the distribution of material capabilities as well as the ability to ability to commit said capabilities to battle and the expected (subjective, political) costs that would be incurred as a result. These are not just ideas that formal theorists made up.

Miscellaneous Links

2012-01-21T14:13:00.005-05:00

Below the fold.

1. Faulty arguments about the end of the Iraq War. Excellent post by Scott Wolford, who has been blogging pretty regularly lately. If you're a fan of this blog and you're not following his, shame on you.

Re: Iraq -- I obviously do not know this for sure, but I'd argue that the Bush administration had reason to suspect that when it began "the surge", it would be doing nothing but delaying the inevitable. Sort of like the way the Obama administration is currently delaying the inevitable in Afghanistan. Speaking of which...

2. ISAF's got 99 problems, and the ANSF is definitely one. Q: how do you know you're winning a war? A: when your allies account for more than 5% of your fatalities, and not by accident.

3a. Why we're not going to war with Iran. (H/T Dan Drezner.) I'm ashamed to admit I wasn't aware of how many times we've heard this song before. See also 3b. Risk inflation and Iran. The very first line of Helprin's piece tells you it'll be useless. Brad Smith offers a nice critique of some of the other nonsense contained therein.

4. This is what bargaining looks like: carriers in the Gulf edition. Iran told the US to keep its aircraft carriers out of the Gulf. Even passed a (completely vacuous) law prohibiting such. By returning to 2 carriers, the US has issued a fait accompli that would provoke war if Iran happens to be the type they've been representing themselves to be, but will go largely unanswered otherwise.

5. What do we know about democratic transitions? The always recommended Jay Ulfelder offers some thoughts.

6. The pace at which liberal democracy is taking hold in the Arab world is breathtaking. Exhibits 39, 40, and 41. Of course, none of this means the Arab Spring won't prove to be a very good thing in the long run. Maybe it will. But I continue to be both amused and disappointed by how predictably things have unfolded lately, in contrast to the wildly unrealistic and charmingly romantic commentary we saw early on.

7. Let's check in with US-Pakistani relations, shall we? "Seeking to downplay Pakistan's refusal to allow him to visit the country, US Special Representative for the Afghanistan-Pakistan region Marc Grossman on Friday said he will wait till Islamabad re-evaluates its relationship with Washington." That's nice.

8. Some excellent advice for grad students. I didn't follow most of this, but I wish I had. In fact, most of this continues to be good advice after grad school.

9. Against treating methodological preferences as matters of identity. Your choice of methods (for a given project) should be seen as a choice of the tool that is appropriate for the job, not a way of signaling your allegiance to the Cool Kid Club.

10. Math doesn't suck. You do. (H/T Zach Jones.) Not for the easily offended. And I wish he'd have left out the misogynistic para. But overall, some good points raised. No one ever asks "When will I use this?" of any of the many other things that they're expected to learn -- even those they really will never use. And if you manage not to use math, that just means you've chosen not to take advantage of a powerful tool. It doesn't mean math isn't useful.

If you choose to believe that the only reason I favor the type of work that I do is because I think it puts me in the Cool Kid Club, I'm glad I could make your day by putting links 9 and 10 next to each other.

How Not to Think About China's Rise

2012-01-20T17:35:00.003-05:00

In a new article in International Security, Michael Beckley claims that the Chinese Century is a myth. This has attracted the attention of Andrew Sullivan.

There's a lot of information in the article. I'm not going to engage all of it. Suffice it to say that Beckley's careful argument deserves close attention, and by no means does it rise or fall on the basis of the one point I'm going to hone in on. Let me be clear that my goal here is to nitpick with one particular aspect of his argument: Figure 1, which Sulllivan reproduced in his post. It is not, by any means, the entirety of his argument. But it's perhaps the most eye-catching.

The figure, as you can plainly see, plots GDP per capita for the US and China from 1991 to 2010.

I've put together an analogous graph for the US and Luxembourg.

Anyone prepared to argue that the 21st century will belong to Luxembourg?

Didn't think so.

Okay, okay, that's a little unfair. Let's look at what Beckley was actually trying to say with that graph.

It is, however, important to recognize that GDP is not synonymous with national power, and that countries with larger economies do not necessarily have more resources at their disposal. Half a billion peasants will produce a large volume of output, but most of it will be immediately consumed, leaving little left over for national purposes. As Klaus Knorr argued, what matters for national power is not wealth, but “surplus wealth.”

Knorr is not talking about "surplus wealth" per capita. He means in aggregate. So the fact that Luxembourg's population is so small would be sufficient to explain why they will not dominate the 21st century. The US has a somewhat smaller amount of surplus per capita, but that's multiplied by a much larger number of people, and thus yields a much greater amount of surplus wealth.

But you know what?

China doesn't lack for people. As you may have heard.

Suppose China had, oh, I don't know, say 1/4th the surplus wealth per person compared to the US. Since they've got somewhere on the order of 4 times the population, that would give them roughly the same amount of surplus wealth. Now, I don't know where we're drawing the line for "surplus", but pointing out that Chinese GDP per capita is less than US GDP per capita is fairly misleading, given the difference in population.

Yes, there are lots of other reasons to be skeptical of claims that the future belongs to China. The rest of the essay discusses many of them. Others have as well (see here, here, and here for some briefer takes). I'm not necessarily taking issue with the overall argument Beckley puts forth. For all I know, China will slow down, or break up into a half dozen smaller countries. I have no idea.

What I am saying is that we ought not to be too impressed with the fact that US GDP per capita is so much higher than Chinese GDP per capita. Beckley is right that GDP is not a perfect measure of power. But it's fairly decent. And I'm not sure GDP per capita is any better. In fact, I feel fairly comfortable saying that, as a first approximation, GDP actually does give us a better picture of who's who in international politics than GDP per capita does, though both contain important information. Yes, Chinese GDP per capita remains, to be blunt, rather pitiful. But, China will soon have a real blue water navy, complete with aircraft carriers.

Clearly, there's some "surplus wealth" being generated over there.

Stop Asking "Do Sanctions Work?"

2012-01-19T10:59:00.007-05:00

Whenever someone proposes responding to a crisis with sanctions, the first response is that sanctions don't work. Then there's a flurry of media stories asking the question "Do sanctions work?" or "Can sanctions work in [country x]?".

There are three good reasons why we should break this cycle.

First, sanctions often aren't designed to succeed (ungated). They're sometimes designed to appease those who demand that the US "do something" when the president doesn't actually care enough about the issue to threaten force. They may even be least likely to be chosen when they'd be most likely to succeed. For sanctions to work, they need to have actual bite. They need to come from the right sources (ungated), and they need to be strictly enforced. Enforcing sanctions is not easy, since many industries within the sanctioning state would prefer to go right on doing business (see here and here). The failure of sanctions that were were not designed in such a way as to give them their best chance of success obviously biases our estimate of how well they could work.

Second, the mere threat of sanctions often is successful. When we observe sanction imposition, we're already looking at a truncated sample that includes the most resolved targets. If we observe the low rate of concessions being granted subsequent to sanctions being imposed and infer from this that sanctions are an ineffective tool of statecraft, we'll inappropriately be reluctant to even threaten using them in the first place, and thus forego opportunities to extract easy concessions. Granted, this point doesn't necessarily speak to the current debate about what to do about Iran (which I'll return to below), but it's still worth remembering that we shouldn't be too quick to dismiss economic coercion.

Third, and, most importantly -- no one ever asks "does military coercion work?" Yet the (sometimes) unstated conclusion of most "sanctions don't work" arguments is "therefore we should use force."

You may find it odd that I'd even raise the question of whether military force works. Of course it does, you're thinking. And you're ready to rattle off a set of carefully chosen, yet hardly representative, examples of success.

But bear with me for just a moment.

There are two ways that the application of military force might succeed.* The first is by brute force. If the goal is to conquer territory, repel an enemy invasion, destroy weapons factories etc, then the goal can be accomplished directly through military force.

The second is through coercion. Here, the goal is to inflict enough pain on the opponent to induce voluntary compliance. The logic here is precisely identical to that of sanctions. There is no theoretical difference between the two. In practice, it may be either easier or more cost effective to apply the level of pressure necessary to ensure compliance through military force. In others, it won't be. It depends on the circumstances.

Thus it ought not to surprise us that militarily powerful states tend to achieve their objectives when said objectives can be accomplished through brute force, but in fact have a stunningly poor track record when it comes to using military force to ensure voluntary compliance. Compare "defend regime" and "remove regime" to "peacekeeping" and, most importantly, "policy change", in the figure below (taken from this excellent article by Patricia Sullivan):

Consider the current standoff between the US and Iran. In principle, it's possible to prevent a state from acquiring WMD through brute force alone. Ask the Israelis, who did so in Iraq and (probably) Syria. But many experts believe that will be difficult to impossible with Iran. So if the US and/or Israel are going to stop Iran from acquiring the bomb, they're going to be attempting to achieve a goal of policy change through coercion. And coercion, though possible, is not easy -- whether you're attempting it through sanctions or by military force. Especially when the goal is policy change.

It's not at all clear to me that Iran is in fact going to develop a nuclear weapon. And it's not clear to me that if they did, this would be an outcome the US and Israel could not live with. (The Isrealis seem to be coming around to that idea too.) But I'll leave it to others to sort out what price the US and/or Israel is, or should be, willing to pay in order to coerce Iran. It's not my job to make those kinds of value judgments. What I can tell those of you who are convinced that "something must be done" is that neither our theoretical understanding of coercive bargaining nor the historical record provides much support for the belief that military force will "work" against Iran whilst sanctions will not.

*Regarding the question of threats to use military force, see here and here.

Democracy, Intelligence, and War Outcomes

2012-01-18T05:24:00.008-05:00

One of the primary reasons I am skeptical about the putative link between democracy and victory is that I don't see any compelling reason to expect it to be true in the first place.*

Most of the arguments for why democracies should win their wars more often are actually fairly intuitive at first glance. But when you dig deeper, you realize that there are some problems. The selection effects argument, which R&S put more emphasis on than the war-fighting effect (I may have more to say about the war-fighting effect at some point) ultimately rests upon a decision theoretic view of the world (see this post). To their credit, R&S are upfront about this.**

What happens if we assume, unlike R&S, that politics actually has distributional consequences (remember Lasswell -- politics is about who gets what), and that states may use either war or negotiation to allocate disputed goods?

Before I begin, perhaps I should clarify why I have said that R&S assume that politics does not have distributional consequences.

In a previous post, I assumed that R&S think that if a state does not fight a war, that means they are left with the status quo. But in fact, they don't really even assume that. If you read their essay in this edited volume, which offers an excellent and concise summary of their book, you will notice that they have just about nothing to say about actual threats to security, loss of territory, access to resources, or any of that. Winning wars makes for good political theater, and losing wars bad political theater. R&S argue that democracies are like boxing champions, who refuse most fights because they have little to gain and everything to lose if they are defeated. The implication here is that winning is good because it proves you are great and enhances your popularity, while losing is bad because it tarnishes your reputation and therefore means being unseated (as champion, in the case of boxing, or as leader in the case of leaders of states).

I don't know about boxing champions, but when states decide, "hey, you know what, I don't need this win, and a loss would just be embarrassing, so I'll sit this one out," that is rarely without consequence. George H.W. Bush may or may not truly have believed "I'll win or I'll be impeached" in 1991, but had he decided not to go to war, that would have meant that Iraq would have taken over Kuwait. And maybe Saudi Arabia too. He couldn't simply shrug off the challenge because he didn't need to prove himself, the way a boxer would. And if Ben-Gurion had thought in 1948 that the odds of victory were not very good, he could not have simply said to Egypt and Syria and Jordan and Lebanon and Iraq that he didn't want a fight.*** Again, we can't simply assume that states can opt to ignore a fight without suffering any consequences.

Okay, these are cheap shots. At least, I hope you agree they are cheap shots. I don't think it should really be all that controversial to say that politics has distributional consequences. Right?

Nor do I think it makes sense to assume that, should war be averted, the distribution that will obtain is given exogenously. That's fancy academic talk for saying that who gets what is determined by choices states make. The way many scholars have begun to try to account for that is with bargaining models. As I've discussed before (see in particular here), the "bargaining" of interest in such models may not necessarily entail sitting down at the negotiating table and hashing out terms. It may instead refer to the use of fait accompli. And that's fancy academic talk for saying that if a state chooses to sit this one out, they're going to have to live with whatever outcome their opponent imposes on them instead.

In fact, a critical insight of the application of bargaining theory to the study of international conflict is that whether states choose to fight or not is a function of what outcome their opponents seek to impose.

Crazy idea, huh?

When you put it in plain English, it seems impossible to disagree with. (At least, I hope it does). And yet, the vast majority of IR scholarship focuses only on whether states find the expected payoff from war to be relatively good or relatively bad in absolute terms, without any appreciation for what the alternative is. Or they treat the alternative as fixed, rather than allowing the opposing state to decide how hard they want to push.

What does any of this have to do with intelligence?

R&S argue that democracies will win the wars that they choose to fight not only because they are more risk averse, but also because they are better at estimating risks. They only go to war if they are very likely to win, whereas autocrats don't min taking bigger gambles, and they are better at accurately forecasting whether they are in fact likely to win.

I have no idea whether democracies are in fact better at gathering intelligence, or have a better functioning "marketplace of ideas", or whatever. For the sake of argument, I'll assume that they do. What I want to convince you of is that if this was true, and if we further assume both that international politics has distributional consequences and that the distribution of goods is going to be determined by the states themselves, we have no reason to conclude that democracies will be more likely to win the wars that they fight.

In this paper with Scott Wolford, which will be appearing in ISQ this summer, we analyze a model in which state 1 first allocates a finite budget of resources to intelligence gathering, military expenditures, and domestic consumption, then receives intelligence reports about the strength of 2 before issuing an ultimatum to 2.

Our results indicate that when states find both the marginal cost of improving the quality of their intelligence and the marginal cost of improving their prospects for victory to be relatively palatable, they will tend to make investments in the latter that give them an even chance of victory, and investments in the former that give them a relatively modest chance of receiving inaccurate intelligence. When either the marginal cost of improving the quality of their intelligence or the marginal cost of improving their prospects for victory is relatively high, they will invest nothing in intelligence.

Put differently, the only states that are likely to possess quality intelligence in equilibrium are also states that will have made military investments that give them a roughly even chance at victory -- but no better. If democracies tend to find the cost of investing in their intelligence apparatus to be relatively low compared to autocracies (which seems consistent with R&S), then democracies should be likely to put themselves in a position to have a roughly even chance of victory, but no better.

Why is that?

The intuition is something like this. You are not sure if your opponent has 100 units of military power available or 110. You yourself, though, have only 10 units. Whether your opponent has 100 or 110 almost doesn't matter. The best case scenario and worst case scenario are essentially the same. You're in big trouble. Suppose instead that you have 1000 units. Again, the fact that you don't know whether your opponent has 100 or 110 unites doesn't matter. The best case and worst case scenario for you are both pretty dang attractive.

Suppose, instead, you have 100 units. Now you really care whether your opponent has 100 as well or has 110.

In the latter case, investing in intelligence gathering may make a lot of sense. In the first two cases, it almost certainly does not. States invest in intelligence when they are near parity because the material implications of their inability to accurately determine their opponents' strength are greater when the two are near parity.

Our model may be overly simplistic. We only allow one side to invest in intelligence and arms. We assume there are only two states in the world. We don't worry about domestic politics. There's a lot you can criticize. But I think the logic connecting greater investments in intelligence to parity is pretty strong, and I don't see any reason to expect that this would go away if we complicated the model. I could be wrong, of course.

At any rate, suppose you don't even buy that much. Suppose you think it's possible that democracies will be much better informed than autocracies, regardless of the distribution of capabilities. I'm not sure why that would be true, but I'll play along.

Once we dispense with decision theoretic accounts and allow for the possibility of bargaining, it's very hard to escape the conclusion that states who are both well informed and likely to win any potential war are also states who are unlikely to fight wars in equilibrium. Maybe democracies are, as a general rule, both very well informed and very well prepared for victory. Fine. Those democracies shouldn't be the ones showing up in the data sets we use to analyze war outcomes.

This is where you say, in a tone that suggests you think I may have completely lost sight of this fact, "But, fil, democracies do fight wars. And they win them too."

Indeed they do (if we set aside certain types of wars that get in the way of our nice and neat story). As a purely descriptive matter, that's an empirical fact that I do not contest. The question is why.

I don't know the answer. But I'll tell you this much -- it almost certainly isn't because they have better intelligence. I say that because the logic connecting superior intelligence to a greater propensity to win wars is, as far as I can tell, underwhelming. And R&S present no empirical evidence about the quality of pre-war estimates of victory as a function of regime type, so don't try playing the "you're just being clever, but ultimately all you've got are a bunch of untested claims" card. Because when it comes to explaining why democracies allegedly select themselves into winnable wars, whilst autocracies are more willing to gamble, all anyone else has got is untested claims too. And, in this case, that claim rests upon the assumption that politics is not, in fact, about who gets what, and that strategic interaction is irrelevant.

When we get past that point, let me know.

*For me, that's a problem, because my attitude (which I am well aware is not shared by the modal political scientist) is that when we have identified a correlation in (crude, poorly measured) observational data, but no compelling theoretical story for why there should be a causal relationship there, then we should be reluctant to conclude that there is a causal relationship. Many people in this field say that's getting it all wrong. I think that's largely because they grossly overestimate the power of statistical analysis of observational data to correctly identify causal effects, as I've discussed before and will discuss again, but for today, I'll set that aside. In this post, I'm not going to try to convince you to reject barefoot empiricism. Whether you agree with me that the gap between (the logical implications of) R&S' putative explanations and their correlational results is damning or not, I hope you at least agree that it's something of a problem that their supposed explanations do not necessarily predict their findings.

**Let me be clear that I think that R&S deserve a lot of credit for that. I can think of some prominent arguments that only work if we pretend that states cannot really negotiate in any meaningful sense, who were not upfront about that.

***I don't mean to make that sound like Israel is "the good guy" and the Arab states "the bad guys." The Israelis committed many deplorable atrocities in that war, and in the years to come. My point is only that refusing to fight has consequences. There is more at stake than risking the embarrassment of defeat, and attendant domestic political sanction.

Miscellaneous Links (updated again)

2012-01-17T09:03:00.005-05:00

I've been busier the past few days than I expected. Regular posting to resume soon. In the meantime, some links.

1. Fascinating NYTimes piece on the Guyanas and their contested borders. If your ignorance of this least-talked-about part of South America is anywhere near as bad as mine, you'll learn a lot from this.

2. Exhibit A of strained relationship between US and Israel regarding response to Iran. Exhibit B. My understanding is that there's been a sense of "these are our allies?" in the USFP community for some time, but it was mostly kept behind closed doors. All the public ever heard was how Israel was the closest ally of the US (though, of course, the US has about seven of those). Are we seeing a sea change in US-Israeli relations?

3. And you thought that US politics was unduly influenced by Israelis... wow. Just wow. Can you imagine if something comparable were true in the US? This blew my mind.

4. France accuses Israel of water "apartheid." This continues to be one of the most poorly appreciated (by Americans) factors in the Israeli-Palestinian conflict, IMO. Surprisingly strong language though, and already being called a "mishap" by the French.

5. This is what comes from refusing to distinguish between "Islamic" and "terrorist." I wish I could say that I didn't have to correct half my family members the same way someone needs to correct Rick Perry.

6. If the protests don't bring down Assad, perhaps economic woes will. I've seen a number of people speculate that Assad will survive because they don't expect any significant foreign support to be provided to the rebels, and the rebels can't do it alone. I'm not entirely persuaded by these claims, but I probably haven't thought enough about it yet.

7. Pakistani President Yahya tipped US journalist off to attack on India ahead of 1971 war. Not sure that means all that much, but interesting all the same.

8. NYTimes story gushing about Taiwan's recent presidential election and China's response. Of course, it doesn't even mention the KMT, the DPP, or the pan-blue and pan-green coalitions. Can you imagine a NYTimes story about a European national election that failed to identify the parties of any of the candidates, including the winner? I don't know about you, but my understanding of (modern, representative) democracy is that it has a little something to do with parties. But it's not like we expect much from the nation's most prestigious paper or anything.

9. Evaluating the impact of policies using RDD. Great post. Too few political scientists are familiar with this method

10. Romney, Obama, and the Taliban. It's Scott Wolford. Go read it.

11. This is what bargaining looks like: Gulf War edition. A literal ultimatum. I continue to be baffled by the number of IR scholars who already express their arguments in the language of strategic behavior and responding to incentives, yet resist the insights of bargaining models and instead rely on (implicit, informal) decision theoretic models of war onset.

12. Best song about stealing heroin from the Taliban that you will hear today. (H/T Jake Wobig)

13. Layman's response to academic complaints about Publish or Perish. Sounds about right to me. Though ask me again when I'm denied tenure...

Do We Know that the American Dream is Dead?

2012-01-13T20:21:00.013-05:00

I try hard not to write about economic issues. I have no expertise on them. But sometimes I can't help myself.

You might want to come back tomorrow and see if I have something to say about international politics. (Hint: I probably will.) If I were you, that's probably what I would do.

But, if you're curious, look below the fold for some thoughts on why recent studies proclaiming the death of the American Dream are (so far as I can tell) incapable of establishing such a conclusion.

Suppose we have two hypothetical countries. In country 1, the average person's income in generation 1 is 45,000, and the average income of that person's children will be 20% greater than their own. In country 2, the average person's income in generation 1 is 35,000, and the average income of that person's children will be 10% greater than their own. That is, country 1 is richer, and their children tend to do better than them by a greater margin than is true in country 2.

But that's just the averages. Suppose parental incomes are also more variable in country 1, but the variance in children's advancement over their parents is the same in both countries.

Let's get more detailed. Here's my STATA code:

set obs 100000

gen country = 1 if _n<=50000replace country = 2 if _n>=50001 & _n<=100000

drawnorm parinc1 parinc2, means(45000, 35000) sds(7500,5000)
drawnorm growth1 growth2, means(1.2, 1.1) sds(0.05, 0.05)
gen parinc = .replace parinc = parinc1 if country==1replace parinc = parinc2 if country==2
gen inc = .replace inc = growth1*parinc if country==1replace inc = growth2*parinc if country==2

That is, I created an empty data set with 100,000 observations. Half the observations belong to country 1, and half to country 2. As mentioned above, country 1, relative to 2, is richer, with more variable incomes for the first generation, and the children in country 1 tend to do better than their parents to a greater degree than the children in country 2.

I next sort parental and child income into quartiles.

Let's take a look at some cross-tabs, starting with country 1:

| quartile

parquart | 1 2 3 4 | Total

-----------+--------------------------------------------+----------

1 | 11,099 1,400 1 0 | 12,500

2 | 1,393 9,170 1,936 1 | 12,500

3 | 8 1,915 8,841 1,736 | 12,500

4 | 0 15 1,722 10,763 | 12,500

-----------+--------------------------------------------+----------

Total | 12,500 12,500 12,500 12,500 | 50,000

And for country 2:

| quartile
parquart | 1 2 3 4 | Total
-----------+--------------------------------------------+----------
1 | 10,698 1,781 21 0 | 12,500
2 | 1,778 8,275 2,408 39 | 12,500
3 | 24 2,393 8,045 2,038 | 12,500
4 | 0 51 2,026 10,423 | 12,500
-----------+--------------------------------------------+----------
Total | 12,500 12,500 12,500 12,500 | 50,000

There's a lot of information there. Let's just ask a simple question -- what's the average probability of moving up by at least one quartile?

In country 1, the richer country where children tend to see their incomes increase over that of their parents by a larger margin, the average probability of jumping up by at least one quartile is 10%. In country 2, it is 12.5%. For those of you who only care about relative measures, that means country 2 is 25% more upwardly mobile than country 1!

That's right. The poorer country, where children's incomes increase from a smaller base by a slower rate, is the one with greater "upward mobility", as it is commonly measured. How can that be?

Because relative measures don't tell you what you think they do. As I've said before.

The average income for the second generation in country 1, btw, is roughly 54,000. On average, then, the children in country 1 earn 9,000 more than their parents. In country 2, the average income is 38,500. On average, then, the children in country 2 earn 3500 more than their parents.

I don't know about you, but if I had to choose between being born into country 1 and country 2, I'd go with country 1. I get that not everyone feels the same. Some people think that equality for its own sake is super important.

I could be snarky and say that those people presumably are rejoicing about the significant decline in inequality the US experienced during the recent recession, a decline brought about by most people losing a lot of money but those at the top tending to lose more. But we don't even need to go there.

Suppose I set aside my own biases and say that reasonable people could well disagree about whether country 1 is a better place to live in than country 2. But unless we're trying to be deliberately misleading, we should all agree that if the United States looked like country 1 and most Western European states looked like country 2, that would hardly constitute evidence that the American dream has failed. It would only mean that America is a less equal, but more prosperous, place. A place where rags to riches stories are by far the exception rather than the rule, yet where the difference in absolute terms between a child's income and their parent's income is pretty promising. Far more so than elsewhere.

I don't know if that is actually what we're seeing. It may well not be. To the extent that I know much about what's going on in the US economically right now, I see plenty to be upset about. I'm not trying to say that I have any particular reason to believe that Horatio Alger's America has ever existed anywhere but in fairy tales.

But if someone told you that US foreign policy had gone completely off the rails lately and cited as evidence the high rate of incarceration in the US, you might want to tell that person to back up a few steps. The conclusion (USFP on the wrong track) may well be appropriate. And it is indeed factually correct that the United States imprisons a large share of its population. Many people, myself included, consider that a bad thing. But it's got nothing to do with whether USFP is off-track.

Similarly, the question of whether the American Dream is alive and kicking cannot be answered by asking whether the United States is doing a good job of producing outcomes that the European model was designed to achieve (i.e., promoting equality at the expense of growth). You might like the European model. That's cool. I'm not telling you that you shouldn't. It's a swell model in all sorts of ways. But I am telling you that evaluating the probability of moving into a different quartile than your parents is about as relevant for establishing the conclusion you've claimed to establish as incarceration rates are to the evaluation of the relative success of USFP. We as a scholarly community can do better than that.

At least, I'd like to think so.

Re-Upping My Predictions for Afghanistan

2012-01-12T02:23:00.008-05:00

I've made some predictions in the past on this blog, and my track record is, admittedly, mixed at best. So take this with a grain of salt. But, for what it's worth, I'm repeating, and clarifying, my expectations for Afghanistan.

1.) There will be no agreement between the US, Karzai, and the Taliban before January 2013.

2.) If there is, some or all of the actors will defect from its terms by the end of December 2015 (probably sooner).

3.) If the US leaves Afghanistan as scheduled in 2014 (reasonably likely), the Taliban will either regain control of the central government or acquire de facto control of broad swathes of Afghanistan within a year.

4.) If the US leaves Afghanistan as scheduled in 2014 (reasonably likely), Afghanistan will be neither peaceful nor democratic in any of the three years immediately following US withdrawal.

Rather bold of me, no?

Actually, I don't think so. I bet most of the key players would agree with just about all of the above, at least in broad strokes. Perhaps they'd hedge a bit more, and opt for looser time intervals. But overall, I think it's been clear for quite some time how this all will end. And the National Intelligence Estimate has just released a report that seems to envision a fairly similar scenario. Though, predictably enough, the administration is pushing back against it.

In fact, the administration is in the middle of a huge push to get talks going. And the NYTimes, ever a critical watchdo, has a front page story that, until you get to the very end, reads like a White House press release. We're told that, according to the Secretary of State (who has no incentive at all to put the best possible spin on things), "for the first time there appear[s] to be support for a political resolution that include[s] [the Taliban]."

Yes, yes, the US is offering the government that it once declared responsible for the 9/11 attacks just about everything it wants. But if that sounds to you like the US is admitting that it is not going to achieve the objectives it has set for itself*, consider this:

“The reality is we never have the luxury of negotiating for peace with our friends,” Mrs. Clinton, who has pressed the initiative within the administration, said at the State Department.

“If you’re sitting across the table discussing a peaceful resolution to a conflict, you are sitting across from people who you by definition don’t agree with and who you may previously have been across a battlefield from.”

Plus, the US still considers "a peaceful, democratic government that preserves the gains made over the last decade" nonnegotiable. And it will ask the Taliban to renounce Al Qaeda. So it's all good. I mean, as long as the US says that's what it expects to get, that's what it will get, right? If you think that an agreement whose terms involve the US giving the Taliban virtually everything it wants in exchange for (cheap talk) promises to disarm, renounce Al Qaeda, and respect democracy will become utterly unenforceable the moment the US withdraws, well, you must hate America. Or have taken some game theory classes and learned about such crazy ideas as backwards induction and commitment problems. One of the two.

Of course, even the NYTimes eventually (three paragraphs from the end of a three page article, where no one will read it) gets around to admitting that

Even now, the officials said, much remains uncertain, including the role of Pakistan in any negotiations, as well as the willingness of any of the sides to come to terms on meaningful, lasting reconciliation.

So again I ask, why should I not conclude that the only reason the US is still in Afghanistan is to keep Obama's hopes for a second term alive? On what basis might one conclude that the administration sincerely believes that it is going to achieve it's "nonnegotiable" goals of ensuring a peaceful, democratic Afghanistan? What credible evidence is there that the Obama administration's strategy for Afghanistan is anything other than the Nixon Strategy (i.e., don't lose the war until your second term)?

*Random and meaningless geek note -- it also sounds a helluva lot like what Littlefinger told Eddard Stark here.

Everything You Never Wanted to Know About Diversionary Conflict

2012-01-11T00:42:00.001-05:00

Nothing brings a people together like a common enemy. One of the most consistent findings in the literature on presidential approval is the "rally-round-the-flag" effect, or the tendency for the public to suddenly decide that their leader has been doing a wonderful job during times of crisis. The greatest example of this is 9/11, but it's a general phenomenon.

Ever since this was discovered, scholars have worried that presidents might have an incentive to go out looking for a crisis when things aren't going so well for them domestically. Instances where leaders enter into crises in order to divert the public's attention away from domestic problems are known as diversionary conflicts.

Sometime this week, I hope to finally wrap up what I hope will be my last set of revisions on the last paper I will ever write about the topic. (When I do, I'll post a link.) In anticipation of that, this post offer's a layman's version of the arguments my co-author, Daehee Bak, and I present in attempt to demonstrate that there's nothing to see here.

There are two very good reasons that previous scholars have identified for why we might not expect US presidents to go out there looking for someone to beat up on when things are going bad for them domestically.

First, if there really was an incentive to do so, traditional whipping boys of the US -- at various points in time, the Saddam Husseins, Moammar Qaddafis, Fidel Castros, and Kim Jong Ils of the world -- would be able to see it coming, and would do everything they could to deny the US an incentive to start trouble.*

Notice how quick Iran was to accuse the US of making up an excuse to attack them, and thereby distract the US public's attention from the ailing economy, back in October?

Second, the US public is a bit more sophisticated than this literature gives them credit for. When an unpopular president goes out looking for trouble when they'd really like people to be talking about something else, it's not hard to see through. Recall, for example, that when Clinton lobbed missiles at Sudan and Afghanistan in 1998, at the height of the Lewinsky scandal, he was accused of trying to "Wag the Dog". More than one-third of the US public believed that his primary motivation for these strikes was to divert attention away from the Lewinsky scandal. As a general rule, the more it seems that the president has political motives for using force, the less likely he is to see a significant rally effect (see here).

To illustrate why we think that this dog won't hunt, we discuss an incident in 1984, when Regan's approval ratings were just below 50%. Sudan (at that time a US ally) suffered attacks that they alleged (with pretty good reason) were conducted by Libya. The president of Sudan begged the US to intercede. And the US did -- sending a few AWACs (reconnaissance aircraft). But Sudan wasn't satisfied, and Qaddafi continued his aggression towards Sudan, supporting rebels in south Sudan. He also continued issuing bellicose statements targeted not just at "that bad man in Khartoum" but also against the US. Nonetheless, the US lost all interest in the incident after that.

If traditional diversionary arguments had any real bite to them, this makes no sense. What was Qaddafi thinking, attacking a US ally and threatening the US after it (sort of) came to the defense of its ally? Yet, maybe Qaddafi knew something that IR scholars who are obsessed with the diversionary hypothesis don't, because the US didn't both to do more than sending some spy planes, despite the fact that you couldn't ask for a better opportunity for a diversionary war.

Daehee and I argue that if we're going to tell any kind of story linking diversionary incentives to the incidence of international conflict, we've got to explain why 1) the foreign actor would play along, and 2) why the public would not suspect that they were being duped.

The best we can come up with is an argument that, while logically consistent, occurs under conditions we consider wildly implausible. Therefore, we conclude that there's really very little reason to expect that conflict is more likely to occur when the president is in desperate need of a bump in the polls.

Our crazy story, that we ourselves think is implausible, is that rivals of the US would initiate a low-level dispute against the US, rather than the president himself being the initiator of the conflict. They would do so, however, only when they believe that giving the president the rally effect he so desperately needs would prevent the president from going to war. That is, if there is a legitimate disagreement between the US and the foreign state, and the issue is just important enough that the US would be willing to go to war under some circumstances, but not so important that the president can't be persuaded to leave things by if the foreign leader orchestrates a sham crisis, purely as political theater, in order to help the president get reelected, then and only then could you get a (low-level) conflict occurring due to the domestic problems facing the US president. The public would still rally behind the president in such a case because it wouldn't look like the president went out looking for trouble, given that the conflict was initiated by the other side.

So we've got a story that circumvents the two obstacles preventing traditional diversionary conflicts from occurring. But the conditions under which it would occur are quite demanding. The US needs to have a credible threat to go to war if the foreign actor doesn't play ball, yet also have a credible promise not to go to war if the other side does indeed serve up the kind of crisis that would create a rally around the flag. Hard to see that actually happening.

In the second half of the paper, we demonstrate that the conditions that are most ripe for diversionary conflict do not seem to be associated with an elevated likelihood either of US rivals initiating conflicts, a decreased likelihood thereof, nor an elevated likelihood of the US initiating conflicts against its rivals. Past studies have found a general link between domestic turmoil and US uses of force, but they use very crude measures. We argue that if you can't find clear evidence of any pattern at all when the president's ability to benefit from a rally effect is at its very greatest, then it's pretty unlikely that whatever past studies have found is really diversion -- particularly given that there's no theoretical reason to expect it to have been in the first place.

Specifically, we look at the behavior of the US and its rivals in the months immediately preceding a presidential election in which the incumbent is eligible for reelection, and check to see if conflicts are more likely when the president's approval is in the range where a bump of a few points could spell the difference between victory and defeat.**

None of this is definitive. And b/c of data concerns, we haven't looked at other democracies. I have no real expectation that any of the dedicated believers in diversion will be persuaded. But hopefully we can convince some of the people who have no real dog in this fight that there just isn't much going on here. Neither the US nor its rivals seem to behave very differently when the president is in a position to benefit in a huge way from a rally effect than when he is not. It just might be time to admit that diversionary conflict was an interesting idea, but it simply doesn't seem to tell us much about US conflict involvement.

*This is the "strategic conflict avoidance" claim attributed to Smith (1996).

**Actually, we test for a quadratic relationship, and assume that if diversionary behavior, or strategic conflict avoidance, or whatever, was taking place, then there'd be a real difference between the experience of president's with approval near the middle of the spectrum than presidents with very high approval (who are all but certain to be reeelcted anyway and thus have no need for a rally) or those with very low approval (who are all but certain to be tossed out no matter what and thus cannot benefit from a rally unless it is implausibly large).

What Does That Mean?

2012-01-09T20:44:00.010-05:00

I'll be teaching American Foreign Policy again this semester for the first time in a few years. And though I've already decided not to use any textbooks, I'm looking through some of the ones that publishers have sent me, and fighting the urge to vomit.

Am I teaching political science, or some mix of high school civics, diplomatic history, and current events?

Of course, that's the poli sci textbook market in general. Nothing unique to AFP. The one thing I am finding that resonates through most AFP texts that I don't find in other undergrad texts (at least in IR -- I can't speak for American politics) is the emphasis on American exceptionalism. It's not just a topic most feel the need to cover, which would be perfectly understandable. Rather, it seems that most texts on the market have a lurking, implicit assumption in favor of American exceptionalism.

And the more I think about it, the more I find that I'm not even entirely sure what the word means.

Of course, when I say that I don't know what it means, I don't mean to feign utter ignorance. What I mean is that the word seems to mean at least three different things. If we could disentangle the different meanings, maybe we could have a real conversation about how appropriate is for our textbooks to be pushing them on our students. We might, perhaps, decide that students out to be exposed to all of the component ideas, but that our textbooks should implicitly assume the inherent correctness of each of those ideas. Or maybe we'd drop some entirely. I don't know. But I'm pretty sure it's doing our students a disservice to fail to distinguish between the different components.

So what are the different component ideas to which I'm referring?

First, there is the historical/sociological argument that the United States is unique because it was "conceived in liberty", embodies a frontier spirit, and to this day defines itself by its ideals rather than ethnolinguistic nationalistic identity.

Second, there is the view that the US occupies a unique position in the international system currently, and as such, is best positioned to provide global public goods like fostering a system of open economic exchange, keeping sea lanes open and safe from threats posed by piracy or rogue states, combating transnational terrorism, combating nuclear proliferation, etc.

Third, there is the view that the US is destined to lead the world towards liberty and democracy.

My guess is that, if you surveyed IR scholars, you'd find that a pretty large number agree with number two. And not just as a matter of ideology. A number of well-developed theories of international cooperation emphasize that global public goods are best provided by the system leader. The thing here, though, is that there is nothing mystical or magical or unique about the US. This isn't about destiny. Historically, Great Britain played a similar role. In the future, China, or India, or a United Europe, or who knows who else, may possibly play the same role. At this moment in history, it just so happens that the US has greater military, economic, and institutional capabilities to play such a role. But that's all. I have no particular objection to our textbooks teaching this argument. But I strongly prefer that they use slightly more objective language in so doing, that they point out that what is exceptional about the US is its capabilities. One need make no moral presumption about the quality of the character of its people to believe this argument.

The first, on its face, does not appear to be pulling at anyone's heartstrings. It walks and talks like sensible scholarship. But I have serious doubts about it. First of all, while it may be true that the modal form of nationalism is based on ethnolinguistic identity, it is not true that the United States is the only state that exhibits what some scholars refer to as civic nationalism. So, uh, where's the exception?

For that matter, one might take issue with the idea that nationalism in the US is defined solely in terms of commitment to certain principles. Most politicians are careful to stick to such language, and perhaps that's what matters. But there seem to be significant elements of the US public (including some of the intelligentsia) who feel that you ain't American if you ain't white, mono-lingual, God-fearing, and preferably Protestant. Whether the legacy of the frontier explains why Americans have a different attitude towards guns and individualism and all sorts of other cultural matters than are typically found in the OECD I'll leave to the sociologists and people who actually care about those things. But at minimum, I find it problematic to claim that the US is different from all other states (that is what exceptionalism is all about, right?) because it rejects identity politics and embraces all those who share its liberal values. Sorry, but I've got to call bullshit on that.

Finally, while religious leaders and talk radio hosts have every right to promote the notion that the US has a destiny to do this, that, or the other thing, for the life of me, I can't see how that has any place in an undergraduate textbook in a discipline that is attempting to be objective and scientific. I won't claim to know that it is not the US's destiny to bring freedom to the world...but I'll be damned if I'm going to stand in the front of the room, with the full weight and authority of my position behind me (which I sometimes deride, but let's not kid ourselves -- when the people with PhDs behave as if any idea deserves credence, impressionable youths take notice), and tell them that it does.

One might even go so far as to say that we have an obligation to point out to our students how frequently the US has behaved in an entirely unexceptional matter. Has behaved as though, just like every other state out there, it is willing to compromise its principles when it believes doing so serves the national interest. Or when specific politicians believe that it will advance their individual careers. And that is indeed how I teach my classes. But even if we don't go that far, we could at least try to be a little more neutral, no?

Now, I should also say that there are some very good textbooks out there already. For example, there is this laudable text by James Lee Ray. There are various reasons I chose not to use Ray's text, but a subtle endorsement of American exceptionalism isn't one of them. I don't mean to suggest that I'm under the impression that there are no alternatives. I'm just bemoaning the fact that there are so few of them. And that so many students across the country are enrolled in courses in which the instructor has adopted a textbook that advocates American exceptionalism. And, most of all, the fact that most of those textbooks probably do not even mean to endorse all the different connotations of that term. Our students are so used to hearing it thrown around by people who use it in an almost mystical sense that I think we ought not even use the term at all if we're not willing to step back and discuss what exactly we mean, and what we don't mean, by it.

Bargaining and Conflict Over Natural Resources

2012-01-09T15:17:00.002-05:00

To judge by what I've seen in some recent working papers and published articles, there's still a lot of confusion out there about what "rationalist" explanations for war (read: simple crisis bargaining models) do or do not predict empirically.* It seems a lot of people think that all Fearon told us is that war is more likely when there is more uncertainty or when power is shifting rapidly, and that all other factors lie outside the scope of "rationalist" explanations for war.

An easy mistake to make. I'm not judging anyone here. As a grad student, I believed something very much like this, and told a fellow grad student of mine that bargaining theory told us that natural resources shouldn't affect the likelihood of war. But I couldn't have been more wrong.

The mistake I made, and others continue to make, was in failing to grasp the difference between fundamental causes of conflict and the potential for various conditions to alter the probability of one, should one be introduced by said fundamental causes. That is, even if we don't think of natural resources as issues that influence power (in which case, they could influence the likelihood of conflict: see here for a general take and here for an argument specifically about natural resources and conflict), the presence/absence of natural resources speaks to the issue of contention and the relative likelihood of war given the presence of information or commitment problems, though the latter would still constitute the root cause of conflict. In a much more abstract way, I tried to tackle this issue here. In this post, I'm going to try to do so more concretely.

But first, some modeling.

For arbitrary reasons, there's a convention among bargaining theorists to fix the size of the pie over which actors are negotiation to 1 while allowing their subjective evaluation of the costs incurred in war to vary. This causes unnecessary confusion, as people often think the cost terms are objective measures rather than subjective ones. It also makes it harder to see that the very simple bargaining models that now lie a the heart of so much theoretical work in IR are perfectly capable of accounting for variation in the stakes.

So let's step back from that convention and instead arbitrarily fix the costs of war to 1. I do this for each side, but that doesn't actually mean I'm assuming that each side will suffer the same cost in a war. What it means is that all of the other parameters of the model are implicitly expressed in relation to the cost of war each actor would incur. Let v_C be the challenger's value for the issue in dispute, relative to the cost of war. So v_C equal to 2 would mean that the disputed good is so valuable to the challenger, that acquiring full control of it would be worth twice as much as the value that is lost by fighting a war, while v_C of 0.5 would mean that the issue is of so little importance to C that even if C were to acquire full control of it, that would only offset half the loss of value of fighting a war. Let v_D, which is interpreted similarly, denote the defender's value.

As is standard in this literature, assume that C issues an ultimatum to D, denoted x. If D accepts, C receives v_Cx and D receives v_D(pi - x), where pi is the size of the pie (clever, no?). Ordinarily this is fixed to 1, but since we've already fixed the cost terms, let's let this vary. As you'll see, it makes no difference anyway.

If D rejects, C receives v_Cw - 1 and D receives v_D(pi - w) - 1, where w is C's expected share of the good following a war.

D accepts any x less than x-hat, where x-hat is a threshold created precisely to have this property and is equal to w + 1/v_D.

Since C can get v_Cw + v_C/v_D by demanding x=x-hat, and receives a payoff of v_Cw - 1 by setting x equal to anything greater than x-hat, it follows that -- under complete information -- C will always set x = x-hat, which by construction is a demand that D will accept, and war will never occur.

Nothing new here. This is just Fearon's proof about the inefficiency of war in a slightly different form.

Now suppose that C does not know how well she will do in a war against D. All she knows is that with probability q, w will be relatively large (denoted w-large) and with probability 1 - q, w will be relatively small (denoted, cleverly enough, w-small).

I'll skip the proofs, but we can readily establish that C will either issue a relatively small demand, which I'll denote x-small (because I'm imaginative), or a relatively large one (x-large). When C sets x equal to x-small, D is certain to accept, and war will be averted. When C sets x equal to x-large, D accepts with probability q and rejects with probability 1 - q.

Um, fil, what does any of this have to do with natural resources?

Well, whether C chooses x-small or x-large depends upon the size of q relative to some threshold q-hat (which is derived precisely in order to have this property), and q-hat depends on how important the issue at stake is to each side. Provided we think natural resources influence the stakes (does anyone doubt that?), we've got a clear story for how natural resources would play a critical role in determining whether war occurs or not.

However, the precise way in which they do so is perhaps a little more nuanced than you might expect, so let's go through the details.

Specifically, q-hat is equal to (1 + v_C/v_D)/(v_C(w-large - w-small) + 1 + v_C/v_D).

The proper way to sort out the impact of a change in v_C or v_D on q-hat is through comparative statics analysis (i.e., take the partial derivative). I suspect that anyone still reading at this point is perfectly comfortable with that, but in the vain hope that I've still got the (partial?) attention of some others, I'm going to illustrate the key effects simply by plugging in some numbers.

Let's consider four cases. In the first, the issue is unimportant to both actors (v_C = v_D = 0.5), and there's a substantial amount of uncertainty (w-large - w-small is 1). In the second, it is again unimportant to both, and there's only a modest amount of uncertainty (difference b/w the two w's of only 0.2). In the third, let's assume that there are significant amounts of some natural resource at stake, and so the issue is fairly important to both sides (v_C = v_D = 2), and there is a substantial amount of uncertainty. Finally, in the fourth case, let's again assume the issue is important to both sides, but let there only be a modest amount of uncertainty.

The values of q-hat (which q must exceed for there to a chance of war, meaning that larger values make it less likely that C will gamble and thus promote peace) that we get in these cases are approximately: 0.67, 0.95, 0.5, and 0.83, respectively. Meaning, there threshold that q must cross before C is willing to risk war is lower when there is more uncertainty, regardless of how important the issue is, but the stakes still mater quite a bit. The difference between q-hat with substantial versus modest uncertainty is roughly 0.18 when the stakes are low, but roughly 0.33 when the stakes are high. When both sides consider the issue to be really important, as we might expect when there are significant amounts of natural resources at stake, the conditions under which war can occur can be considerably more permissive, particularly if there is a substantial amount of uncertainty. It is still uncertainty that is creating the problem here. But that doesn't mean that "rationalist" explanations for war can't account for the importance of natural resources.

Now let's take things a step further to see why bargaining theory not only can be reconciled with, but can tell us something novel about, the relationship between natural resources and conflict.

What happens when the issue is important to only one side?

Even if we're talking about natural resources, that might be the case, provided one actor has greater demand for that resource or has a more limited supply. I'm not going to provide numerical examples of that, but it turns out that the effects are not symmetric. An increase in v_D strictly lowers q-hat, while an increase in v_C may either increase or decrease q-hat, depending, in part, upon the gap between w-large and w-small. (Feel free to check this for yourself if you don't believe me.) Since the gap between w-large and w-small is likely to be greater when the states are near parity, that's something we can work with.

That tells us that, even though the simple little model I've set up largely confirms our intuition, it does more than simply serve as an elaborate exercise in telling us what we already know. It gives us a more nuanced understanding of the relationship between natural resources and conflict.

If we could come up with empirical measures of each state's need for natural resources, rather than simply measuring their presence, and we could be confident about who the defender is and who the challenger is (which may or may not be as simple as initiator and target in EUGene -- see here), we'd have two novel hypotheses.

H1. The likelihood of conflict increases as the challenger's resource-sensitivity increases, particularly for dyads near parity. The more imbalanced is the distribution of power, the more attenuated this relationship will be.

H2: The likelihood of conflict increases as the defender's resource-sensitivity increases. There's no particular reason to expect this effect to depend upon the distribution of capabilities.

So, any grad students out there in search of a dissertation topic, all you've got to do is come up with a way to measure resource-sensitivity and you're good to go.

*I won't link to specific papers, because I don't want this to sound like I'm trying to name and shame people. Some people I consider friends, and whose work I really respect, have made some misleading claims about this. I don't blame them. I blame bargaining theorists for being so bad at outreach. One of the (admittedly not so modest) goals of this blog is to start to correct that.

Miscellaneous Links

2012-01-08T20:08:00.003-05:00

Below the fold.

1. Musharraf to return to Pakistan, stage a political comeback. Should go over as well as a Cheney presidential run in the US.

2a. Again I say, al Shabab is losing the war in Somalia, not that anyone seems to have noticed. 2b. Kenya has made a rather costly commitment to the AU PKO, which was already doing well.

3. UN High Commission on Refugees's report on 2010. Afghanistan and Iraq account for nearly 50% of global refugees. About 75% of refugees reside in countries that neighbor their country of origin. Developing countries host roughly 80% of refugees. I wonder why the US isn't more popular in the developing world...

4. Israeli Navy will soon be focusing most of its effort on protecting offshore oil rigs. Most of these rigs are not within Israel's territorial waters (12 nautical miles from land), but only within its "economic" waters (70 nautical miles). That makes them highly sensitive to attacks. Not that the Arab-Israeli conflict has anything to do with resources off course. No, it's irreconcilable religious differences and historical grievances. I keep forgetting.

5. Anton Strezhnev has an insightful (and inaugural) post on Libya and the ICC. Punchline: the NTC's lack of commitment to the ICC need not be sign as a lack of commitment to the rule of law. I remain less than optimistic about the NTC, but this is a good point, and well argued.

6. US, Saudi Arabia move towards regime change on Syria. But check the date of publication. The more things change...

7. Detecting corruption in India. (H/T Brendan Nyhan) Clever research design. Sadly predictable findings.

8. Some good advice about blogging. Mostly common sense, but there's a lot of good ideas here that you might not have considered.

9a. Some meditations on the relationship between goals and victory from Dan Trombly and 9b. Adam Elkus. See also 9c. this 2007 JCR article by Patricia Sullivan.

10. You probably already heard about the US Navy rescue, but ICYMI. Putting hard power resources to work exercising soft power?

11. Muslim Brotherhood offered guarantees to US about peace with Israel. Just days after promising to put those treaties to a national referendum. I just don't get why the US would insist on guarantees thought. We all know joint democracy promotes peace, right?

12. "What is your definition of 'against'?'' Today's Republican Party seems to think that the one thing the Bush administration got wrong about the War on Terror was refusing to make it a war against Islam. That's an...interesting way of looking at it.

Obama's Pakistani Gamble?

2012-01-08T15:39:00.001-05:00

In a lot of ways, I think the US-Pakistani relationship is easy enough to understand, in broad strokes, that I find most media coverage of it to be remarkably frustrating (epitomized here).

But I will freely admit that there are some things I still don't understand. Specifically, it looks to me like Obama gambled that killing Osama bin Laden would do more to improve his electoral prospects than maintaining a semi-functioning relationship with Pakistan. And that seems to me to be a bit nuts. But my general philosophy is that whenever it seems like a leader has made an incredibly stupid move, that's almost certainly a sign that there's more to the story than I understand. So I'm going to assume I'm wrong in some way. What I'm wondering is: is that because it's wrong to think he made such a gamble, or wrong to think that (he should have known that) it was a bad bet?

On the one hand, I'm not sure it's safe to conclude that the raid that killed OBL is responsible for the recent breakdown of US-Pakistani cooperation. It might simply have been just a low point from which the relationship could have otherwise recovered, if not for the recent (and largely accidental?) border incident. After all, Bush and Musharraf apparently had an agreement that the US would be allowed to conduct such a raid, while the Pakistanis would feign outrage after the fact. Granted, neither Bush nor Musharraf were in power at the time, so who knows if a similar agreement was operating at the time, but still. I think the exact cause of the breakdown remains unknown. And, as I mentioned once before, the ISI itself seems to think I was right about the civilian government having been partly complicit in that raid. So maybe Obama (correctly) thought that it was possible to get OBL while maintaining (what passes for) cooperation with the Pakistani government, even if things did later fall apart for other reasons.

For the sake of argument, though, let's assume otherwise. That is, let's assume that Obama thought (correctly?) that he faced a tradeoff between getting OBL and maintaining the SQ vis-a-vis Pakistan, and he choose to give up the latter in favor of the former.

It was pretty easy to predict at the time that killing OBL would only provide a temporary boost in approval ratings (see also here). Indeed, Obama's approval ratings followed a fairly predictable trajectory in the months after. Granted, we don't get to observe the counter-factual. It's possible that Obama would be doing worse right now if GOP criticisms of his "appeasement" foreign policy were actually sticking, and perhaps the reason they are not is OBL. We don't want to judge the wisdom of an action strictly by evaluating the desirability of outcomes observed along the equilibrium path after said action, after all.

But we also don't want to lose sight of the downside of deteriorating relations with Pakistan. The US ceased drone strikes in November, and the effects are already making themselves known. At least, it would seem so. Again, we want to be careful about drawing inferences, because we don't know the counterfactual. But in this case, the change in US policy seems to have been caused by factors other than expectations about the future strength of militant groups (to put it mildly), so I feel reasonably confident that the post hoc ergo propter hoc fallacy is less of a concern here than it normally is.

So, if Obama actually made the gamble I've suggested, is it wrong for me to think that he bought himself a short-term boost in the polls (coupled with the ability to deflect criticism of his foreign policy that probably wouldn't persuade moderates anyway) at the price of having 12 months or so of worse headlines coming out of Afghanistan than would otherwise be the case? Am I wrong to think that without (half-hearted) Pakistani cooperation, the US public just might realize in 2012 what it otherwise wouldn't learn until 2014 -- that the US is not going to achieve its objectives in Afghanistan?

Like I said above, I'm always suspicious of simple "what a dummy!" arguments. Odds are, I am indeed missing something. I'm just not yet sure which potential flaw in my reason is the one I should be focused on.

Perhaps all of them?

Proposal Power, Revisionism, and Initiation

2012-01-05T19:13:00.002-05:00

The ultimatum has become the workhorse of theoretical analysis of crisis bargaining. Partly, this is just because it is the simplest bargaining protocol, and we've come to believe that it's important to allow for the possibility of bargaining. But some of us would go so far as to say that it is justifiable to assume away the possibility of counteroffers, for various reasons (more on that below).

If you're willing to buy that, though, you might nonetheless ask: does it matter who the proposer is? Under the ultimatum protocol, one side has all the bargaining power. Not just in the literal sense that they choose the terms of the agreement, but in the more meaningful sense that this ensures that the responder will never receive an outcome that is even the tiniest bit better than their outside option.

Usually, we put no thought into this. Depending upon the question you want to ask, it may not actually make any difference. But, other times, it does. So how do we decide who plays what role? Should/does proposal power correspond to revisionist aims (i.e., dissatisfaction with the status quo), as I assume explicitly here? Should/does proposal power correspond to being the side that will be the initiator of armed conflict, as Slantchev explicitly assumes here?

It is for precisely this reason that I struggle with how to refer to what state 1 does when they set x. Are they issuing a demand? Are they offering concessions? Are they merely proposing terms? Or perhaps something else altogether?

Depending upon the assumptions we make, any of those might be appropriate. And that's no small matter. Even if it is one that I gloss over in every single paper that I write (all of which, for some time now, have included an ultimatum crisis bargaining model.)

It has only been in the past few months that I have started to realize just how profound an insight it was that Fearon mentioned, in passing, that player 1's act of choosing a value for x can be thought of as issuing a fait accompli.

Let's try to get a bit more concrete.

In legal settings, trials are ex post inefficient in precisely the same sense that wars are inefficient, and fully informed actors that faced no commitment problems or principal-agent problems would be expected to reach settlements or plea bargains. Labor strikes are also quite similar. But there is often a fundamental difference between the first two and interstate crisis. And that difference is that fait accompli are typically impossible in legal and workplace settings.

That is, the prosecutor cannot sentence the defendant to ten years in jail in hopes that the defendant will prefer that to a trial. The status quo (defendant, being innocent until proven guilty, will serve no time in jail) can only be altered by either an agreement being reached or a trial. And, if we're talking about situations where there is a labor union engaged in collective bargaining, management cannot unilaterally revise the status quo. Wages are negotiated, and contracts are signed, or everything breaks down.

Now, at this point, someone is probably thinking, "but most workers are employed in settings where their wages can be unilaterally cut without their consent." Yes, I know. But how often do we see labor strikes in those settings? I'm not saying that I love The Man, and that those of you who love to stick it to him have nothing to be upset about. I'm saying that bargaining theorists in economics who study labor disputes are interested in a fundamentally different problem than bargaining theorists in political science who study crises.

In international politics, actors typically do have the ability to unilaterally alter the status quo. The options are not simply costly dispute resolution (war, trial, labor strike) or mutual agreement, as I believe is the case in legal and labor settings.

Example: the US and Iran are currently engaged in crisis bargaining over a variety of issues. Iran has the power the unilaterally disrupt the flow of oil out of the Gulf by closing the Strait of Hormuz. It has the power to unilaterally sever ties with Hezbollah and other non-state actors. It also has the power to unilaterally develop a nuclear weapon. The United States has the power to unilaterally alter the number of military bases immediately bordering Iran (see here if you have doubts about whether Iran might have cause to care about this). The United States has the power to remove sanctions on Iran, or to take Iran off the list of state sponsors of terror, or to remove its support for Saudi Arabia and Israel. I can go on, but I think you get the point.

Now, as this example illustrates, it is often (perhaps always) true that both sides have the ability to unilaterally alter the status quo. That's a problem for simple models like the ultimatum.

Might we have cause to think that one side has greater incentive to do so?

I would argue that indeed we do.

Sticking with the example of the US and Iran, the US is largely satisfied with the status quo, while Iran has revisionist aims. Both sides have undertaken policies they would like the other to abandon, but until Iran ramped up its development of its nuclear program, that was far more true for Iran than for the US. Many of the policies I listed above were adopted instrumentally in order to exert leverage over the other actor. They are not what is being bargained over so much as how the states are going about the business of bargaining. At root, the US likes the economic, cultural, and political status quo in the region a good deal more than Iran does, though neither is exactly at their ideal point.

If we justify modeling crisis bargaining via the ultimatum, rather than a Rubinstein model or whatever, and we base this justification on the possibility of unilateral revisions of the status quo (i.e., fait accompli), then it seems to me that we should indeed adopt the convention of assuming that the proposer is the revisionist state.

What does that mean about initiation?

On a certain level, it makes sense to me that we might equate dissatisfaction with the status quo with initiation. That tends to be the case, empirically. Something like 75% of dyadic MIDs are initiated by a revisionist state, and the same is true if you limit the analysis to wars. That's pretty reassuring. But it still means that a non-trivial number of disputes are not initiated by revisionist states. And though Powell has proven that only one state can be dissatisfied with the status quo at any given time (in the sense that they'd be willing to go to war rather than let the status quo stand), there are plenty of MIDs where both sides are revisionist. And more than a few where neither are (even though our theoretical models tell us that war cannot occur when both sides are satisfied).

That leaves us with two options. One, we need to admit that the MID codings of revisionist aims don't quite map onto our theoretical notion of dissatisfaction with the status quo. Or, two, we need to consider the possibility that status quo powers may be likely to engage in the first use of force in some cases.

I think both are likely. In fact, we have an entire class of theoretical models dedicated to establishing explanations for the latter -- models of preemptive or preventive war brought about by commitment problems that result from rapid shifts in power.

How common are such wars? I don't know. Dan Reiter thinks they're pretty rare. But his use of the term "preemptive" is far more limited. Or the use of the term by bargaining theorists is too broad. I don't mean to make it sound like it was unreasonable of him to define it the way he did. Just that our theoretical models lead us to predict satisfied states will initiate wars under conditions that do not neatly map onto the conditions Reiter had in mind in his article.

Okay, but what does all this mean?

There are two takeaway points.

First, I think bargaining theorists should be more willing to justify the choice of bargaining protocol. It's not just a useful simplification. There's actually good reason to think that states issue ultimata. Unlike in other contexts, agreement is not necessary for the status quo to be altered. And there's typically going to be one side that has a stronger interest in doing so. Yes, this bargaining protocol implies that one actor has much less bargaining power, and will be held down to their reservation point (i.e., will do no better than their outside option). But I don't think that's a random quirk of a stylized model that we only use because we can. I think that may actually be how international crisis bargaining really works -- even though I'm not prepared to say the same about legal or labor disputes.

Second, I think empiricists need to be more careful about the interpretation of "initiation". I'm hardly the first to point this out. My adviser, Glenn Palmer, once had the students in his proseminar write a paper about the problems with the way the MID project codes initiators, motivated by the observation that Poland is the initiator of the MID that ultimately expands into WWII. One response is, well, mistakes happen. The point I'm making is that the problem is deeper than that. We have good theoretical reason to believe that, in some conflicts, the state that will use force first will be the state with revisionist aims, but in other conflicts, it will be the one seeking to uphold the status quo.

In fact, an enterprising young empiricist might consider it worthwhile to look into testing the hypothesis that initiators of MIDs that occur in the midst of a rapid shift in power will be less likely to have revisionist aims than initiators of other MIDs. Myself, I don't care enough to check just now...though I might look into this when I start working on the book I'm allegedly writing...

Are Democracies Better at Maximizing Security?

2012-01-04T23:46:00.010-05:00

As regular readers have no doubt noticed by now, while I do very much believe that domestic politics plays an important role in shaping international outcomes, I'm a skeptic when it comes to the many glorious and wonderful things IR scholarship has allegedly demonstrated about democracy (see here, here, here, here, here, here, here, here, and here). But as I've done many times in the past, I'm going to, for the duration of this post, take these claims at face value and focus on an unappreciated logical implication of even the most optimistic of accounts. I won't even question whether democracies really do select themselves into conflicts that can be won quickly and easily (really, I won't, not even parenthetically).

In their book, Reiter and Stam claim not only that democracies win the wars they select themselves into, but that "the attributes associated with democratic institutions, those that provide for personal liberty, freedom of expression, and collective material growth of common citizens, [are] also the same attributes that, in the worst of times, allow states to provide for their national security as well." That is, "[i]t would appear that democratic nations not only might enjoy the good life of peace, prosperity and freedom; they can also defend themselves against outside threats from tyrants and despots" (page 2).

We can have our cake and eat it too. What's not to love?

Like I said, I'm going to take R&S at face value here. If you object to the implications of the argument I'm going to develop, keep in mind that this means you are, per force, objecting to R&S's argument. Unless you want to admit that hold logic in contempt, that is. You can't just pick and choose which implications of a given argument you are going to accept. If you accept the argument, you accept everything it implies.

Myself, I have serious concerns about their argument. So I don't necessarily expect the conclusion to which this post is building to hold. Again, let me be as clear as I can -- the goal of this post is not to derive a testable hypothesis with the implicit claim that it is likely to survive empirical investigation. The point is to demonstrate to you that if you believe R&S, you necessarily must also believe the conclusion I'm about to put forth. It's a package deal.

So, obligatory scholarly throat-clearing out of the way, let's get to it.

R&S's theoretical argument is best captured by a decision theoretic model. They claim that there is no real need to concern ourselves with the possibility of strategic interaction, because "one important and explicit assumption we make here is that targets must fight when attacked" (page 14).

Implicitly, then, their model is something like this:

State A initiates a war against state B if and only if the gains from war, which I'll denote w, less the loss of utility associated with incurring the costs of war, which I'll denote c, exceed A's value for the status quo, denoted q. That is, if w - c > q, we get a war, and if w - c < q, then state A stands pat.

More or less, anyway. Because of domestic politics, it's just a touch more complicated.

In contrast to some scholars, R&S do not argue that democracies are necessarily more sensitive to the costs of war. Rather, they argue that democratic leaders are more sensitive to the outcomes of war (see here for why we shouldn't believe that, and here on how those two assumptions yield quite different conclusions in a bargaining context).

So rather than focusing on w - c > q, let's say that no leader, of any regime, goes to war unless w - d - c > q, where d is the risk of domestic punishment (i.e., removal from office), where d is itself a function of w. Let us further assume that d is larger for democratic states as well as being more responsive to w.

What I'd like to show can be readily proven analytically. But not everyone is as comfortable with algebra, so I'm going to do some very simple simulations and linear regressions.

Here's the code for creating the simulated data (in STATA):

set obs 10000

gen w = uniform()
gen q = uniform()
gen c = uniform()
replace c = c/10

gen dem = 0
gen draw = uniform()
replace dem = 1 if draw<0.5
gen d = .
replace d = (0.5)/(5+w) if dem==0
replace d = (0.5)/(1 +(4*w)) if dem==1

gen war = 0
replace war = 1 if (w - d - c >= q)

First, I create an empty data set with 10,000 observations. Then I create the three basic variables, all uniformly distributed (on the unit interval for w and q, but only ranging up to 0.1 for c, so as to ensure that the inequality will be satisfied for a sizable subset of the data).

The next cluster of code generates d. I begin with a baseline variable that is also uniformly distributed form 0 to 1. I then create a binary indicator for whether State A is democratic, which takes on a value of 1 with probability 0.5. I then set the values for d, making sure that d is decreasing in w (i.e., assuming that there is less risk of domestic punishment for fighting a war that ends well); that d is more rapidly decreasing in w for democratic states; and that d is, on average, relatively small (again, simply to ensure that I get a good number of wars).

Now, admittedly, the way I went about specifying d is arbitrary. Feel free to play around with the functional form if you want. The only important thing is that d be decreasing in w, more sensitive to w for democracies, and neither too small (in which case you won't get any real effect of regime type) or too large (in which case you won't get many instances of democracies fighting wars).

Finally, I create a binary indicator for whether State A goes to war against B or not, using the inequality from above.

Now for the analysis.

First, let me prove to you that this theoretical model captures the R&S argument.

reg w dem if war==1

Your results will vary some due to random variation, but I have 3423 observations where war is equal to 1, and within that subset, the coefficient estimate on democracy is about 0.03 (p < 0.001). So democracies enjoy better outcomes from the wars they choose to fight. Good news.

Now for the bad news.

gen outcome = .
replace outcome = w if war==1
replace outcome = q if war==0
reg outcome dem

That is, I create a new variable that records the distributive outcome of the crisis. That is reflects how much of the disputed benefits will be under State A's control at the end of the day, which is simply the outcome of the war in the event that one should occur, and the status quo otherwise.

What happens when I regress outcome on democracy?

I get a coefficient estimate of -0.01 (p < 0.001). That is, the same theoretical model that tells me democracies win the wars they fight because of selection effects necessarily implies that democracies are systematically achieving worse outcomes in the international arena.

The intuition is simple. If democracies allow the status quo to stand unless they are likely to do very well in war, they will of course do well in the wars they fight, but they will also be more willing to accept marginal status quos that other actors would be willing to use force to overturn.

The following analogy may help (if you play poker -- as I assume many IR scholars do, given the frequency with which others use poker analogies). The "rock", who folds every hand unless he has a great hand, is pretty likely to win those few pots that he contests. But that doesn't mean that he's going to walk away with the most money at the end of the night. There are in fact very few poker stars who play a very tight game. Phil Ivey, widely considered the best player alive, is extremely aggressive.

Put differently, if we believe R&S, we must necessarily conclude that democracies are indeed insecure, particularly if the international system is filled with aggressive actors who will chip away at them, knowing that they won't fight back unless they know they can win.

Granted, democracies that do not find themselves surrounded by autocracies (do not find themselves saddled with an unattractive status quo that they'd be unwilling to use force to revise unless they were assured of victory) may not have to worry about this. Suppose you are a relatively weak state that historically shared a border with an aggressive autocratic neighbor, but today enjoy peaceful relations with all of your neighbors (think Belgium). In that case, there's no real downside to being democratic. Perhaps you can have your cake and eat it too. But was democratic governance a "luxury", one that came at a price of heightened insecurity, when Germany was still autocratic? If we believe R&S, yes, it was. They themselves claim otherwise, but the logic of their argument is merciless. A world full of democracies might be something worth aspiring to, but along the way there, democratic governance very much would be a liability. At least for minor powers. A democratic state that never expected to be at a material advantage might not have to worry about this. But all other democracies would.

Is it not curious that scholars were most concerned about democracies being insecure during periods of history when democracy was rare, and have only started to believe that there are no tradeoffs now that most of the world's democracies have few, if any, threats in their immediate neighborhood?

But I won't push too hard on that. As I said above, I don't know that I think that's actually the world we live in. There are a number of things about the R&S story that I don't buy. To be perfectly honest, I'm just not sure there's much to be gained by focusing strictly on regime type, even though I do think that domestic politics matters. My goal, once again, is not to derive a testable hypothesis. I'm just trying to demonstrate that the R&S story, on its own terms, should leave us feeling less sanguine about democracy than people realize.

Miscellaneous Links (updated again)

2012-01-04T21:43:00.006-05:00

A longer list than usual, but it's been a while since I posted one of these. As you can see, I've been thinking a lot about Iran lately. I'll probably have more to say about that soon. For now, I'm still trying to sort out my thoughts.

1. The sanctions regime is starting to get serious. And more may be coming. In my mind, there's really no sense in asking whether sanctions "can" work. The question is whether the West has the stomach to keep up the sanctions if it means an Iranian attempt to shut down the Strait of Hormuz.

2a. Speaking of the Strait of Hormuz, Iran seeks to outlaw foreign presence in the Gulf. Not that this law would have any bite, of course. 2b. Iran warns US not to return aircraft carrier to Gulf. 2c. Iran steps up military patrols. 2d. US intends to maintain presence despite warnings. I continue to think the probability of a war b/w the US and Iran is relatively low, but I'm becoming less confident in that prediction by the day.

3. So what would a US-Iran Naval showdown look like? Punchline: US has huge technological advantage, Iranians huge geographic advantage, and the outcome would likely be victory for the US, but at a very high price.

4. Can the US still direct the Middle East peace process? Not that there's much peace process to direct, but AJE makes a decent argument that the answer is no.

5. How not to support popular uprisings. Jay Ulfelder is always worth reading. This isn't what people want to hear, but I think he's on to something.

6. More on the costs of the Iraq war. Key passage, IMO: "At the peak of spending on Iraq in 2008, direct costs represented 1 percent of U.S. gross domestic product. The portion of GDP devoted to war expenditures during the peak year of spending on Vietnam was 2.3 percent, 4.2 percent in Korea and 35.8 percent in World War II, according to the Congressional Research Service."

7. Denying demographic destinies. The Israeli right claims that the size of the Palestinian population has been overestimated.

8a. Provocative essay on the status of AQ. Some of this strikes me as right on. Too much bold assertion of causal relationships w/o any justification for my taste though. If your ultimate goal is to say that current USFP is wrong, I guess it makes sense to simultaneously claim that USFP is strengthening AQ and that AQ is all but dead. But internal consistency is the goal... 8b. See also this rejoined from Daveed Gartenstein-Ross.

9. Defense Secretary Panetta to unveil defense cuts in a speech tomorrow. Keep the carriers, cut the troops. Advocates of off-shore balancing should like this plan, while COINistas will probably be outraged.

10. Excellent post on the role of blogging in academia, among other things. (H/T Brendan Nyhan)

11. According to the new TRIP survey, IR scholars value the contribution of blogs. I'm actually pretty surprised at some of the findings here. Good news.

12a. Foreign Policy on the relationship b/w policymakers and the academy. 12b. My advisee, Nick Nicoletti, on the same. Can you tell that he was trained by one of the 3% of scholars who think academics should play no role in the policy process?

13. Primo and Clarke's book is finally out. I will have more to say about this in a few weeks. I've said that before about other books, and still haven't gotten around to reading them. Hopefully I'll start catching up on my reading list this semester...

14. PM's question time on what quallys know. Well worth reading. A followup post on what quantoids know soon to follow. I look forward to that.

15. I wish this was the worst problem I saw in my students' writing.

Iran and 9/11

2012-01-04T14:17:00.003-05:00

Many of you may already be aware of this. I somehow was not, until a friend mentioned it on Twitter.

I'm no legal expert, so I won't comment on whether the court's ruling was appropriate. But my sense from looking through the documents is that the level of complicity established by the plaintiff's evidence is a) considerably greater than is widely understood, yet b) well short of "the attacks wouldn't have been possible without their support".

I'm not trying to score points here. Make of that what you will. Maybe you think any level of complicity justifies a decision by the US to use force. Maybe you think that someone who donated $100 to Obama's election campaign can be said to have "provided material support" to the party responsible for hundreds of civilian deaths in Pakistan, though would hardly bear much responsibility for that. If so, you might rightly wonder what exactly the court means when they conclude that Iran "provided material support" to AQ. I doubt anyone who thought a war with Iran was a great idea before will fail to find confirming evidence here, and I similarly doubt that anyone who strongly opposes a war with Iran will change their minds after looking over these documents. Still, something to be aware of.

You can expect those beating the drums for war with Iran to have a whole lot more to say about this case in the coming months.

December Roundup

2011-12-30T13:17:00.000-05:00

So I ended up taking most of December off. Sorry about that. I'll be back in force in January.

Here's the five posts from December that received the most traffic.

1. Frequency of Insurgency and US Funding Priorities.

2. How Did the Iraq War Happen?

3. Insurgency and Terrorism.

4. A Question about Grading.

5. What Does a Nuclear Iran Mean?

Insurgency and Terrorism

2011-12-17T00:36:00.003-05:00

I've received a few comments on this light-hearted post in which I expressed surprise at how much trouble my students were having with what was intended to be a gimme MC question. I thought I'd take the opportunity to explain further what I actually had to say about the topic in class, in case anyone is interested.

FWIW, you can find the lecture slides here.

The (overly) simple definitions I provided in class were:

Insurgency is a form of medium-large scale violence directed towards government targets.

Terrorism is a form of small-medium scale violence directed towards government or civilian targets for political purposes.

I have seen many other definitions of each. I have provided other definitions, in other classes, when seeking to emphasize other points. There are more definitions of terrorism out there than you can shake a stick at. I mentioned this in class. I did not, and would not, claim to have a definitive account of either. Nor do I claim that the distinction between the two is a simple one.

The point I was trying to make in lecture is that both of these are tactics, rather than identities, and that we might expect some set of disaffected elites within the population to seek to achieve their goals by adopting terrorist tactics if they lack the support of the broader population, but resort to insurgency if they enjoy a greater level of popular support.

Yes, that's a stylized argument. It clearly overlooks the fact that many organizations use both tactics at the same time. It implicitly precludes the possibility of governments conducting terror attacks. As do some, but not all, definitions of terrorism. As I said, I make no claims to be offering a definitive and exhaustive account of political violence. Rather, I'm trying to demonstrate to my students that if we start from a set of what I hope others would consider reasonable simplifying assumptions, we might arrive at some surprising conclusions about political violence.

I would note, though, that we're not alone in making such an argument. Many people have made similar points. One such example, of many, is this great paper by Ethan Bueno de Mesquita on rebel tactics.

These lectures build on a paper I'm writing with Brian Hardt, which is currently R&R at JPR. In the paper, we focus on popular versus limited insurgencies, but we might use the same theoretical framework to contrast insurgency and terrorism -- if you're willing to go along with the assumption that terror campaigns are typically marked by low levels of support from the population, whereas insurgencies typically enjoy somewhat more support (though of course there is important variation within each category).

Our argument is essentially that the government negotiates with a set of disaffected elites that we might think of as extremists (though this term is not entirely unproblematic), knowing that this same set of elites may subsequently negotiate with the broader population in attempt to gain their support. We think it's important to think about both processes at once.

The primary insight of our analysis is that those factors that encourage cooperation between the disaffected elites and the broader population, thereby ensuring that the violence directed towards the government would be larger in scale if indeed such violence occurs, should in fact be associated with a lower probability of violence occurring in the first place.

If we think of the two categories of violence as (domestic) terror campaigns and insurgencies (though, as I said, in the paper with Brian, we focus on the distinction between limited and popular insurgencies), that means that population grievances (such as those stemming from poverty) should be associated with a lower probability of political violence, but should violence occur, it will be more likely to take the form of insurgency. That is, I argued in lecture that we might think that poor countries may be no more prone to violence overall -- in fact, they may be less prone to violence -- but it would nonetheless be true that civil wars are a problem of poor states rather than rich ones.

I don't know if this argument is correct. Brian and I doing some empirical work to determine whether it is in fact true that rich countries are more prone to political violence (if we include the full spectrum of what that entails), but less prone to seeing such violence take the form of an insurgency. At the moment, I consider this an intriguing possibility, one that tells us that we might arrive at misleading conclusions w/r/t political violence if we study civil wars independently from terrorism, but I wouldn't claim much more than that just yet. I think we might have something more than a mere possibility proof here, but at this stage I'm not prepared to make any strong empirical claims.

At any rate, I hope that clarifies why I chose to emphasize differences in scale when contrasting insurgency from terrorism. There are surely many other important differences. I just wanted to expose my students to the argument that different tactics act as substitutes, and that population grievances may be a better predictor of the form of political violence a country will experience than the probability that it will.

What More Can I Do?

2011-12-16T19:49:00.002-05:00

I gave a take home MC final in my War & International Security class. There were 30 questions, and the students had to answer 25 of their choosing. Of the 30 questions, 6 were designed to be very challenging, 11 moderately difficult, and 13 somewhere between easy and absurdly easy.

The question below is one of the questions that was meant to be painfully easy. Yet, to my surprise, I have received more emails about this question than every other question combined. And, yes, those emails seemed to be from students who were sincerely confused as to what the best answer was.

Miscellaneous Links (updated)

2011-12-16T15:40:00.003-05:00

You know the deal by now.

1. My advisee Nick Nicoletti has a new blog. Nick is a very bright young scholar who has taken it upon himself to learn about experimental methods. I encourage you all to follow his blog.

2. Foreign Affairs symposium on Afghanistan. I haven't had time to read these yet but they look really interesting. Plus, anything with Jason Lyall's name next to it is worth your time.

3. Why COIN was doomed to fail in Afghanistan. (H/T Jay Ulfelder). Because no matter how much bad news you've read about Afghanistan, there's always room for more. Best quote: "If the COINdanistas had read these five pages–five fucking pages–they’d have had serious doubts about the efficacy of counterinsurgency in Afghanistan." Do read the whole thing though.

4. Israel forms a new special ops command for operating in the "third circle." The first circle, of course, being the Palestinians, and the second being the Arab states immediately bordering Israel. The third, which would include Iran, is everything past that.

5. WTO grants Russia membership. If Kantian Peace arguments are correct, that further reduces the probability of conflict between Western states and Russia.

6. Can Europe be saved? I haven't read this either, but the lineup commands attention.

7. Lazy writing about American empire. I feel strange linking to a blog as widely read as Drezner's, but I'm doing so on the off chance that this one slipped past you.

8. How not to write about African elections. (H/T Jay Ulfelder). Snark : blogging :: gasoline : automobiles.

9. The futility of blogs.

How Did the Iraq War Happen? (updated)

2011-12-15T16:52:00.004-05:00

As you all know, today marks the day that the US officially declares and end to the Iraq War. With any luck, this actually means more or less what it sounds like it means, unlike the official declaration of the end of major combat operations.*

Now seems like a good time then to step back and ask what we know about how this happened.

As I've discussed before, I think there's a lot that we still don't know. Some of that, we may be better able to answer in the future. But a lot of it is probably unanswerable. Therefore, let me be as clear as I can that this post is in no way meant to be a definitive account of what happened. (That probably goes without saying for blog posts, but just in case there's any doubt.) Rather, this is my personal attempt to sketch the outlines of two plausible explanations that recent IR scholarship offers. No more and no less.

The basic point of this post is not that we know why the war happened, but that we're doing better than some voices within the discipline would have you believe.

The Enforcement Problem Story

I don't know whether the Bush administration sincerely believed Saddam had WMD. The Downing Street Memo, the infamous yellow-cake allegation, and other incidents seem to strongly suggest that, at minimum, the administration was extremely selective in its interpretation of available evidence.

But I think it needs to be said that critics of the administration are a bit too quick to assume any level-headed observer should have known that Saddam had no significant stockpiles of WMD. Nevermind the fact that many of the critics of the war believed he did have such stockpiles in 1998 and 2002 (see here and here for a few glaring examples of how the Democratic Party felt about a war with Iraq back before it was politically expedient to be against it).

It's not hard to understand how reasonable people might have mistakenly believed Saddam had significant stockpiles of WMD. He deliberately induced uncertainty.

If the story was as simple as your average critic of the war would have it, the obvious question is why Saddam didn't just let the world know he had nothing. Why was he acting like he had something to hide? Why did he let inspectors in to this palace, but not that one? Don't tell me it was because he found it insulting to have to prove what he knew was true. He knew that the US was gearing up for an invasion. Whatever loss of dignity might have come with allowing inspectors full access to every site, it can't compare to losing power.

I think Saddam was playing two games at once, and the winning strategy in one was the opposite of what it was in the other. So he adopted a mixed strategy. What do I mean by that? He knew that the greater the US's belief that he had WMD was, the more likely he was to be invaded by the US. He also believed that the lower Iran's belief that he head WMD was, the more likely he was to be invaded by Iran. So he had incentives to convince one actor that he had nothing, and another actor that he had significant stockpiles. It was no accident that no one knew for sure what the truth was.

Schultz recently demonstrated that when A and B bargain over some policy of B's, a policy that B can unilaterally alter without A being able to perfectly observe B's choice, war is often unavoidable. What did we have with the US and Iraq? The US was bargaining with Iraq over a policy that that Iraq could unilaterally alter without the US being able to perfectly observe Iraq's behavior.

I'm not saying that's the story for what happened. I don't believe it is. But we should at least be clear about how difficult the war would or would not be to explain if the world were as simple as bargaining theorists believe it to be. And the answer is -- not all that difficult.

The Domestic Politics Story

An alternative account, drawing on my own work, puts domestic politics front and center.

As I demonstrate in a working paper (a new version of which I'll be posting in the next few weeks), democratic states can be expected to, at times, fight wars that everyone knows will be prolonged and relatively unsuccessful. If we look at the conditions under which this result obtains, I think there's an argument to be made that they were satisfied in this case.

For this tragic outcome to occur, a few critical things must be true:

1. The public votes on the basis of outcomes that are observed, not those that could have been achieved off-the-equilibrium-path.
2. The public is more likely to punish bad outcomes than good outcomes if the primary legislative opposition advocated a different policy than the one adopted by the government.
3. The outcome of a long war, though less than favorable for the democratic state, would be better than the outcomes of either relatively short or moderately long wars.
4. The per period cost of fighting is relatively low, but not too low, for the democratic state.

There are of course many other assumptions, many of which are standard in the bargaining literature. But these are, I think, the only ones that are both essential to the result and in need of further justification.

Basically, I am assuming that the public does not think as carefully about policy as elites do. Rather, they rely on heuristics.** The simplest is whether policy outcomes are good (in which case, most voters will support the incumbent) or bad (in which case, they will not). Another is the presence or absence of elite disagreement (see here and here). Taken together, I assume that taking your nation to war results in an elevated probability of removal from office if and only if that war ends poorly for your side and the primary legislative opposition politicized the war. This assumption of the model is pretty consistent with what I myself have found in related work. So, personally, I'm pretty comfortable assuming these things as a general rule.

With regards to Iraq, I think it's reasonable to argue that the administration knew before the war that the US would not be greeted as liberators, but would indeed face a messy insurgency after toppling Saddam. At least, Cheney clearly thought so in 1992. Now, maybe he didn't believe what he said in 1992 and did in 2002, even though 92 Cheney turned out to be right, but I suspect that's not the case.

Moreover, I think it's reasonable to argue that the administration believed that the outcome of the post-Saddam insurgency phase of the war would be worse if they withdrew early on compared to staying longer.

Finally, though the cost of the war is deplorable in a human sense, if you consider the fact that the US fought in Iraq for more than ten years, it's pretty hard to argue that the per-period cost, either in terms of casualties or finances, was very high. The US has suffered roughly 4500 fatalities in Iraq. That's 450 per year. In contrast, the US suffered 200 in less than 100 hours in 1991 (you don't really need me to convert that into an annual rate, do you?). I won't even bother comparing that to Vietnam or Korea, because I know someone will (rightly) object that those wars aren't comparable. But even so. Don't get me wrong -- I count myself among those who think the US suffered 4500 too many fatalities in Iraq. I'm not trying to justify the war. I'm trying to justify my assumption, analytically, that the per-period costs of war, though hardly negligible, were, in purely relative terms, not very high. If we assume that to be the case, the story I develop with my formal model looks plausible. If we don't, it doesn't. I happen to think we should be comfortable with such an assumption. But that doesn't mean I think we should be comfortable as human beings with the cost that was incurred.

Anyway, if you're willing to believe that the public, writ large, behaves more or less as I've assumed, and you're willing to believe my assumptions about the likely outcome and costs of the war, then we've got what I think is a plausible story. And it goes like this.

The Democrats valued not only holding office, but also actual policy outcomes. Expressing opposition to the war in late 2002, early 2003 would have, according to my model, prevented war with Iraq. But it also would have enable Iraq to make demands that it otherwise wouldn't have been in a position to make. Perhaps that would have meant easing sanctions. This would have been unattractive on policy grounds -- the primary reason Saddam wasn't the threat the Bush administration made him out to be isn't that he was a nice guy, it's that sanctions and containment were working. It also would have been devastating on political grounds. In the wake of the 9/11 attacks, questioning the president's foreign policy agenda was worse than un-patriotic - - it was verboten. The Democrats would have helped to produce a bad outcome, and they'd have shared the blame for that outcome with the Bush administration and Congressional Republicans. So there were strong incentives to support the war, even knowing that they'd later turn against it.

Why would a party that initially supports a war later turn against it, given that I've assumed that there was no uncertainty on behalf of elites within the US about the likely outcome or costs of war?

Given what I've assumed about the likely outcomes available from short, medium, and long war, and given the assumptions I've made about the per-period costs of war, we might expect the government to choose to fight a long war regardless of what the opposition says -- once it makes the decision to go to war at all. That is, I think the Democrats might well have been able to prevent the war by expressing clear opposition to it beforehand, but I think once it began, there was no real question that the Bush administration was in it for the long haul. Given that, the obvious strategy is to oppose it once things turned ugly.

Now, I don't think this story is perfect either. It leaves out a great many factors that I think might well have been relevant (not least of which being the influence of the oil industry, the Israel lobby, and defense contractors). And I'll freely admit that some of the key assumptions of the model, and my particular application thereof to this case, invite criticism. I think I'm pretty comfortable with the assumptions I've made, but I won't pretend that I think reasonable people can't disagree about some of those points.

What I would say is that this constitutes a plausible explanation of the Iraq War. I'm only claiming to offer one potential way to make sense of it.

All in all, I think there's a lot of important questions left unanswered about the Iraq War, but I think we've done a better job explaining what happened so far than some people may realize.

UPDATE: I recently came across this essay, which similarly takes issue with Lake's assertion that bargaining theory fails to explain the Iraq War. McKoy makes many of the same points I made above.***

*Given the non-trivial risk that Iraq will return to civil war, I do think there's a very real chance that the US will be fighting in Iraq again soon. I don't think that's the most likely outcome, but it shouldn't be dismissed out of hand either.

**This kind of behavior is sometimes called irrational, unfairly. Given a probability of being pivotal that approaches zero and non-trivial costs associated with gathering information, it is far from clear what we'd have to assume in order to conclude that expected-utility maximizing voters would do anything other than rely on heuristics.

***Lake makes some other interesting and unsubstantiated claims. For example, it is far from clear that information problems become more acute when more actors are involved, as Lake asserts. In fact, Scott Wolford, who has done more work on multi-actor bargaining models than anyone I know, has got some work in progress that shows this may be true, but it need not be.

The US and Pakistan Do Not Need Couple's Therapy

2011-12-14T16:59:00.016-05:00

This weekend's edition of the NYTimes magazine has a piece about the US-Pakistani relationship that perfectly illustrates everything I love and hate about journalism at the same time.

There's a lot of great information in here. In terms of relaying relevant facts that more people should be aware of, the article more than delivers. There isn't enough mainstream news coverage that distinguishes, appropriately, between the Afghan Taliban, the Paksitani Taliban, the Haqqani network, and Al Qaeda. There's a real appreciation for history, and for the complexity of Pakistan's relations not just with the US, but with Afghanistan and India. The essay points out the relevance of US trade policy to its relationship with Pakistan, which definitely does not get brought up often enough. All of that is to be highly commended. I'd love to live in a world where more Americans were familiar with the basic facts that Keller lays out so nicely here.

But when Keller tries to explain why the relationship is so dysfunctional, and, worse, when he offers advice about what is needed to repair it, I can feel my brain liquefying, pouring out my nose and ears, and making a mess of my beard.

His central argument is that the US and Pakistan are like a troubled romantic couple, and the first step towards reconciliation is to understand where each side is coming from.

Now, if it was clear that Keller was making a joke, that would be one thing. I myself have analogized the US's relationship with various different countries to romantic relationships before. Many people have. It works. As a joke.* I'm not knocking Keller for drawing the parallel. I'm criticizing him for seriously arguing that the key to repairing ties between the US and Pakistan is to validate Pakistan's feelings. That may well be excellent advice for addressing marital discord. I wouldn't know. But I have a hard time imagining it matters much in this context.

In fact, Keller goes on to argue that US-Pakistani relations have been deteriorating recently because the US no longer has an ambassador who understands where the Pakistanis are coming from the way Richard Holbrooke did. I won't comment on the talent of past or current US diplomats, but by Keller's own terms, this is nonsense. What made Holbrooke so special, according to Keller himself?

While he was characteristically blunt about the divergent U.S. and Pakistani views, he understood that they were a result of different, calculated national interests, not malevolence or mere orneriness (emphasis added).

You read that right. The two sides have different interests.

That is, Pakistan fears a post-American Afghanistan that is allied with India, since that would allow India to bargain with Pakistan with the threat of a two-front war lurking in the background. Even if such a war never occurred, the mutual knowledge that another war between Pakistan and India would be even less likely to produce an outcome favorable to Pakistan would require Pakistan to make considerable concessions. As a result, (some elements of) the Pakistani government feel they have no choice but to resort to the low cost but high risk strategy of sponsoring non-state actors that will destabilize Afghanistan. After all, Pakistan doesn't have the resources to achieve its goals in a conventional manner. In fact, neither does Pakistan have the resources to crack down on all of the militant groups that threaten to destabilize the region even if it wanted to. It's not even clear that Pakistan could be successful in that regard if it had the resources of the United States.

In contrast, the US wants to see such groups defeated/disarmed/dismantled both so it can rest assured that Pakistan's nuclear weapons will not fall in the hands of extremists and also so that the American public wiil feel that the 14 years the US will have spent in Afghanistan (assuming it actually withdraws in 2014) were not a total waste.

Yes, that's fairly complicated. But it's not that complicated, and Keller's got the space to develop that argument. In fact, at various points in the essay, he touches upon just about all of the above.

So tell me this, Keller: if the US and Pakistan have different national interests -- and it seems that you, me, and Holbrook all agree that they do -- aren't they likely to be at odds with one another regardless of whether they validate one another's emotions?

Excuse me while I let loose a string of excessive profanity.

Okay, I'm better now.

Setting aside the asinine metaphor of couples therapy, might we at least argue that some less colorful yet more reasonable version of Keller's thesis is plausible? Might the US and Pakistan see their relationship improve if the US tried harder to be persuade Pakistan that they are a valued ally? To combat the sense that "with the Americans, it’s never a commitment, it’s always a transaction"?

I don't know. I'm not sure any of us knows. But we're certainly socialized to overestimate the importance of the human element. We're inundated with stories that play that up because it's easier to get people to read the news if you dumb everything down, sensationalize it, and emphasize the human element.

I'm obviously not advocating that the US try to be as obnoxious and condescending as possible towards Pakistan. If the primary reason that the US-Pakistani relationship is so troubled is that the two sides have fundamentally different interests, as I rather suspect, there's still no harm to be done in being more diplomatic. But let's not kid ourselves. If the US finds the second coming of Richard Holbrooke and sends him to Islamabad, the US and Pakistan are still going to have problems. My apologies to all the students out there who are convinced that they will someday be the second coming of Richard Holbrooke and solve all of the world's problems as a result.

As far as I can tell, we would do well to assume that, however much the Kellers (and Steve Walts) of the world might like to convince us otherwise, the reason states don't get along the way we'd like them to is often that it simply isn't in their interests to do so. It's almost certainly not because they don't try hard enough to validate each other's emotions, or aren't imaginative enough to gamble on bold game changers, or whatever other easy fix we're being sold today.

When Keller says

If you look at reconciliation as a route to peace, it requires a huge leap of faith. Surely the Taliban have marked our withdrawal date on their calendars. The idea that they are so deeply weary of war — – let alone watching YouTube and yearning to join the world they see on their laptops — feels like wishful thinking.

he's exactly right. When he goes on to say

But if you look at reconciliation as a step in couples therapy — a shared project in managing a highly problematic, ultimately critical relationship — it makes more sense.

my head explodes. No, Keller, it does not make more sense. It really doesn't. You should have stopped as "wishful thinking."

*Of course, it doesn't work as well even on comedic grounds as any of us think it does, but it's hardly news that most academics have a terrible sense of humor.

Which Military Option Does Walt Want Off The Table?

2011-12-13T20:48:00.005-05:00

Steve Walt discusses 5 potential "game changers" here. Number 1 is for the US to take the military option off the table.

I think Walt is on to something, but he's got the details of the argument he's trying to make confused.

The fundamental source of confusion, I think, is that Walt fails to distinguish between the threat of the US using military force against Iran at some point down the road in response to the various substantive disagreements between the two and the threat of the US using force against Iran in the short term in order to prevent Iran from acquiring nuclear weapons.

Call these threats TL (long) and TS (short).

The larger Iran estimates TL to be, the more attractive nuclear weapons acquisition becomes, since possession of nuclear weapons grants Tehran the ability to deter this threat.

The larger Iran estimates TS to be, the less attractive nuclear weapons acquisition becomes, since further moves towards developing nuclear weapons only provoke the US. Until the day that Iran actually possesses nuclear weapon, the path towards acquiring them increases their risk of being attacked.

So when Walt says:

But constantly harping on the possibility of military action is not a good way to build trust -- especially when the opponent is already deeply suspicious. It is also a very good way to convince an adversary that it ought to acquire some means of deterring a serious attack, such as acquiring a nuclear weapon, which is precisely what we don't want Iran to do.

He's partially right. The more the US convinces Iran that TL is large, the more the US pushes Iran towards acquiring nukes.

But it absolutely does not follow from this argument that if the US were to somehow persuade Iran that "we were not contemplating any sort of preventive attack on Iran itself" that this would "clear the air somewhat and strengthen the hand of Iranians who recognize that crossing the nuclear threshold may not be in their own interest."

In fact, if the US was trying to maximize the probability that Iran chooses not to pursue nukes, and was not constrained in any way by domestic political pressure to pursue certain strategies or by the desire to preserve its long term potential to coerce Iran over substantive issues, the US would want to convey to Iran that it would never consider invading Iran in response to regional policy disagreements, but absolutely and unequivocally would be willing to invade in order to prevent Iran from acquiring nuclear weapons. That is, it would want Iran to believe the long term military option is off the table, while the short term military option is very much on the table. Not just on the table, in fact, but a veritable certainty.

The very last thing the US would want to do is persuade Iran that the short term military option was off the table.

Now, there's a lot more going on here. And, as I said the other day, I'm far from persuaded that the US should be willing to use force in order to prevent Iran from going nuclear. But if we're simply asking the question of what strategy the US could adopt if the goal was to minimize the incentive for Iran to develop nuclear weapons, I think Walt gets it exactly wrong. There's a certain sense to his argument, to be sure. But his actual conclusion is that the US should be doing the opposite of what it should be doing if it was trying to achieve the goal he has in mind.