Are Threats More Credible When They Come from Democracies?

Over the last decade, international security scholars have created a cottage industry investigating the role of “audience costs” in coercion, thanks in large part to a theoretical argument first put forward by Monkey Cage blogger, James Fearon. One implication is that democratic leaders will be more likely to have their threats believed than non-democratic leaders, because they will likely have to pay an electoral price for bluster, in the event that their bluff is called. The audience costs approach has the virtues of being interestingly counter-intuitive (democracies are more frightening when they issue threats than dictatorships), more complicated than it appears at first glance (it is difficult to test, because much of the argument involves moves that actors don’t make rather than moves that they do, creating problems of selection bias in analyzing the available empirical evidence and so on). For just these reasons, it is also a mainstay of IR Ph.D. introduction to theory courses – it works very nicely for teaching incoming students how to reason their way through a variety of the common pitfalls of empirical and theoretical research?

But does the democratic credibility bit of the argument hold true? On the one hand, there is a substantial empirical literature arguing that it does. On the other, a new article by Alex Downes (a colleague at GWU) and Todd Sechser argues that much of this literature uses problematic data, and hence is suspect.

we investigate the quantitative data sets most commonly used in tests of this proposition—namely, the Militarized Interstate Dispute (MID) and the International Crisis Behavior (ICB) data sets—and find that most apparent democratic “victories” in these data sets are not actually successful threats. … most cases in these data sets entail minor military skirmishes, border and airspace violations, fishing boat incidents, and other events in which the participants did not actually make any demands. These cases reveal little about the conditions under which target states are likely to submit to a challenger’s coercive threats. … they do not differentiate crisis victories achieved by brute force from those achieved via coercive diplomacy. This distinction is critical because the democratic credibility hypothesis argues that democracies are better able to prevail in crises without having to resort to decisive force. …In short, the democratic credibility proposition rests on a shaky empirical foundation. … We reassess the hypothesis using the Militarized Compellent Threats (MCT) data set, a new collection of more than 200 compellent threats issued between 1918 and 2001. …This analysis yields no support for the claim that threats by democracies are more effective. …Several recent studies have contested the view that democratic institutions confer a unique bargaining advantage on leaders during disputes … their potency has been limited by the apparent weight of quantitative evidence in favor of the theory. Our analysis contributes to this debate by showing that much of the theory’s empirical support does not, in fact, survive close scrutiny.Our findings also carry broader implications for the study of coercive diplomacy, suggesting that two of the most commonly used data sets in international conflict research are largely inappropriate for studying coercion in international relations.

This finding is likely to produce quite a bit of controversy. Some Monkey Cage readers (e.g. Chris Gelpi) are engaged with these questions – I look forward to seeing how the argument goes …

8 Responses to Are Threats More Credible When They Come from Democracies?

  1. William G. Nomikos July 16, 2012 at 4:31 pm #

    Whatever the result of this ongoing debate–it is important to distinguish between the claim that democratic threats are more credible than those of autocratic states and the theory of audience costs. The former is an indirect implication of the latter, as per Fearon’s formulation. Selection problems, as Schultz and others have pointed out, pose a far more serious threat to the empirical verification of the democracy/credible threats hypothesis than miscoding does. Moreover, Downes and Sechser’s study does not invalidate the bevy of experimental data, most notably Tomz (2007), in favor of the hypothesis.

  2. Henry Farrell July 16, 2012 at 5:30 pm #

    William – thanks – this is right (my original post was written in too much of a hurry).

  3. Henry Farrell July 16, 2012 at 5:37 pm #

    Now fixed (also fixed a spelling error too) …

  4. Ya Nara Ghalaba July 16, 2012 at 11:54 pm #

    I don’t have a comment on the “audience costs” question, but still I have a comment on the headline question “Are Threats More Credible When They Come from Democracies?”

    Governments of democratic countries can’t threaten other countries credibly if they lack democratic support at home for carrying out the threat. On the other hand, if a threat by a democratic government is provably backed up by support from the bulk of the electorate in supporting carrying out the threat, that’s a very credible threat. It is more credible than a threat by a non-democratic government in a country where the general population’s views about the threat are underdetermined and murky. Here’s an example from this year’s current affairs.

    UK foreign minister William Hague on approx 7 Feb 2012 said of the Syrian government: “This is a doomed regime as well as a murdering regime. There is no way it can get its credibility back internationally or with its own people. When you realise that, you see what a mistake Russia is making by backing this regime.”

    In the months since then, the UK’s foreign minister has said at least once that a Western-led military attack was an option that is and should be kept in the deck of cards in playing the game against the Syrian government. That’s a threat, you know.

    However, a UK public opinion survey conducted in mid-June 2012 (sponsored by Chatham House) asked the question: “Thinking about popular uprisings (such as in Libya, Egypt and Syria) in which citizens attempt to overthrow a dictator, which ONE of these statements comes closest to your view?”
    (A) Britain has a moral responsibility to support such uprisings regardless of whether it benefits Britain’s national interests: 23 percent of respondents
    (B) Britain should only support such uprisings if it benefits Britain’s national interests: 20 percent of respondents
    (C) Britain should not involve itself at all in such uprisings: 43 percent of respondents
    (D) Don’t know: 15 percent of respondents

    Among the respondents who said that they vote for the UK Conservative Party in UK elections, 46 percent picked the option that Britain should not involve itself at all in such uprisings even if it would benefit Britain’s national interests.

    The same opinion survey asked: “Would you say that the [current UK] coalition government has changed UK foreign policy for the better, for the worse, or has made no difference?” Only 6 percent of respondents said it had changed it for the better while 32 percent said it had changed it for the worse.

    That makes Hague’s threat to attack Syria much less credible. He couldn’t bomb without more public support. But if his government were to up its propaganda to try to up public support, it’d likely backfire on them, and weaken them in the next UK general election contest, which they earnestly desire to win, and so they can’t really up their propaganda either.

    The same goes for the current democratic government of Turkey with regard to its stance on Syria;

  5. Dan Nexon July 17, 2012 at 8:23 am #

    Tomz’s experimental findings are creative, fascinating, and all that, but I really don’t think they do much for the hypothesis.

  6. Christopher Gelpi July 17, 2012 at 12:34 pm #

    Hi All,

    I think the Downes and Sechser article is interesting, but you may not be surprised to find out that I am not persuaded.

    But first let me say something about the experimental work on this topic. I am a big fan of experimental work and I think the experiments by Tomz and others do a nice job of demonstrating the existence of a basic individual level mechanism regarding the punishment of leaders for failed or inconsistent policies. So I would disagree with Dan that they don’t do much for the hypothesis, but I would agree that they don’t end this debate for at least two reasons. First, we need more experiments that allow for competitive partisan framing of leaders’ behavior to know if this individual level effect is robust to political spin. And second, even if the individual level mechanism is robust, we need to analyze the behavior of state leaders to see whether they actually respond to these incentives from their constituents. So I don’t think that Tomz can get the audience costs argument off the hook from Downes and Secher’s critique.

    As for Alex and Todd’s article, I have lots of minor quibbles and complaints that I would raise about coding issues here and there, but I think my central response would be that they are wrong about what is the appropriate unit of analysis for studying this question. They want to use the individual threat (and the response to it) as the unit of analysis. I think that is the wrong way to test this argument. I think that the crisis rather than the threat is the appropriate unit of analysis. The difference becomes clear in some of the examples they reference, such as the crisis between France and Hungary in 1919. France issued 4 threats, and the fourth was successful. They want to code that as 3 failures and a success. That strikes me as fundamentally wrong. Imagine a state that prefers to begin with relatively low level threats and escalates as necessary until it wins. If it wins ALL of the crises that it enters but uses many threats to do so, should we code them as “not credible?” Quite the contrary. The audience costs argument is about who wins the overall interaction.

    A secondary issue I would raise is the criteria that they use for coding the presence of a threat. Here they seem to take the same approach as Lebow and Stein in the deterrence debate, and so they find the similarly implausible result that more than a third of the ICB dataset involved exclusively non-militarized episodes. This result occurs despite the fact that the rules for identifying an ICB crisis are “a threat to basic values, time pressure for response and heightened probability of involvement in military hostilities.” I think what is happening is that they miss the fact that the nature many threats and the desired response are implicitly understood by both parties.

    I want to be clear in saying that I think there may be issues and questions for which the MCT dataset is entirely appropriate. But I do not think their use of these coding rules to answer this question is appropriate.

  7. Dan Nexon July 17, 2012 at 2:53 pm #

    I think Chris and I probably agree more than it seems from my offhand remark.

    (apologies in advance for the awkward prose; just having one of those days)

    Tomz demonstrates nicely that, under experimental conditions, voters reduce their approval of leaders who back down. This, as Chris notes, establishes some validity to the individual-level mechanism. But the translation of that finding into the political world is far from frictionless; partisan framing is only one reason among many. Others include expectations about the decay rate of the effect, etc. In short, there are a variety of factors shaping whether or not the other side concludes that a threat issued by a leader is more credible because he or she is concerned about suffering significant audience costs. IIRC, Tomz is fairly open about these kinds of issues.

    There’s also an interesting conundrum that Tomz raises. As he points out, prior to his experiments we didn’t actually know if such an effect existed. Thus, if the “audience cost” mechanism worked in the past, that’s likely a function of leadership beliefs, folk wisdom, and intuitions. Even if such wisdom had a “real” basis, it was at least quasi-independent of the mechanism itself.

    I submit that this issue is worth thinking through–particularly in the context of a blog that spends a majority of its time trying to get journalists, campaign managers, and politicians to pay attention to political science research that debunks their conventional wisdom about, for example, the impact of campaigns on election.

  8. Joey July 17, 2012 at 9:06 pm #

    That’s a big “if” in “if the audience cost mechanism worked in the past”, according to Snyder and Borghard’s APSR article last year. According to that piece, not only have post-1945 democratic leaders rarely issued unambiguous threats (backing down from which would be embarrassing/costly) but leaders of authoritarian states have not perceived potential audience costs in democracies.