On the relationship between information problems and commitment problems

Most political scientists that study conflict are familiar with a pair of bargaining frictions that, on their own, can cause negotiations to break down in war: (a) information problems and (b) commitment problems. Fearon’s famous IO piece distinguishes them from one another as solutions to the inefficiency puzzle of war—i.e., why fight when peaceful negotiation can produce the same outcome without all the waste? But in scholarly practice there seems to be widespread confusion about the relationship between these two mechanisms, both in terms of where one ends and the other begins and whether/how they interact. In this post, I’m going to (try to) clear it up.

tl:dr at the bottom.

Before we go further, let’s define some terms.1

First, an information problem exists when at least one informed actor (I’ll use states as my example) knows something about its war payoffs, which shape the bargains it’ll accept in lieu of war, that the other uninformed state doesn’t know (i.e., private information). And when there are disincentives (like the desire to get a better deal) to reveal that information, uninformed states face a tradeoff: the more they demand, the better off they are in a peaceful bargain, but the more likely is that demand to be rejected in favor of war. And sometimes an uninformed state resolves that tradeoff by running a risk of war.

Second, a commitment problem exists when a rising state will experience an increase in bargaining power, typically its chances of winning a war, if a declining state allows it to—that is, if the declining side doesn’t launch a war designed to prevent, stall, or reverse the impending shift in power. When a rising state can’t pre-commit not to use newfound strength, which we think of as pretty typical in international relations, a declining state may launch a war to lock in the status quo rather than see its share of the pie steadily negotiated away in the future. What’s more—what’s critical, even—is that commitment problems can cause war under complete information.2

In the first case, uncertainty causes war that wouldn’t occur in its absence, and in the second, war occurs in the absence of uncertainty, when states agree that one side will grow stronger in the future.

Simple, right? So what’s the problem?

A confusion about the line between these two ideal-typical mechanisms manifests in two claims,

  1. Commitment problems are really just information problems.
  2. Information and commitment problems exacerbate each other.

though I tend these days to see the latter more often than the former.

The first claim is just false. And whether the second claim is true or false depends on the value of variables that the claim pretty much always omits.

Maintaining these confusions leads to faulty explanations (which means faulty policy prescriptions, if you care about that kind of thing) and biased empirical analyses…both of which we should do as much as possible to avoid.

So let’s do some clarifying.

the first claim

I’ll focus on two prominent versions of the claim. I’m sure there are others (like the ones I hear in grad seminars), but I think addressing these two will do the trick.

First, Kirshner’s (2000) asserts that “most of what is distinct” about Fearon’s characterization of the commitment problem “can be traced to the question of private information” (page 143). That’s simply false: Fearon’s models of the commitment problem strip away private information and show that shifting power can cause war when (in fact, because) states agree on everything about the strategic setting. Without that agreement, war might occur, but it need not be due to the commitment problem—and it wouldn’t have anything to do with the moving parts of the model.

In fact, as I’ll discuss more below, when we introduce private information to a game of complete information with shifting power, it can mitigate the commitment problem. Still, that isn’t necessary to dispense with Kirshner’s version of the claim: the commitment problem takes hold most firmly when states know perfectly well what other states intend to do.

Second, Gartzke (1999, 571-573) argues that commitment problems must be a special case of information problems. He reasons that it’s not clear how fighting could solve a commitment problem, since the act of fighting itself shifts relative power. By his line of reasoning, this creates an endless sequence of further commitment problems; therefore, only uncertainty about the outcome of what would otherwise have to be a total war can cause conflict. His line of reasoning might hold up if all shifts on power were large enough to cause fighting—that is, if belligerents have no control over what they fight over and for how long they do it. But of course not all power shifts are large enough to cause further fighting.3 When fighting can arrest, stop, or simply slow shifting power to an acceptable rate, then it makes peace possible where it wasn’t before (because war kills people and destroys things leaders care about), and only sometimes does it require either disarming or destroying the other side.4

If the commitment problem story has run into information-based challenges, they don’t identify it as an information problem masquerading as something else. The useful challenges note that some sources of shifting power, like armaments, are choice variables, raising the further question of why states would arm in the first place up to a level that would provoke preventive war.5 This has inspired work on (a) the distinction between endogenous and exogenous changes in power,6 and (b) why states might keep rising power secret, etc. But that relates more to the second problematic claim, which entails the interaction of information and commitment problems.

the second claim

Addressing this one requires a bit more work, but the claim itself is easy to summarize: if you put information problems and commitment problems together, then bargaining is more likely to fail than it would be with only one such problem. The idea is widespread enough not to need to cite any piece in particular, and to be fair, Fearon does say that private information “might exacerbate” commitment problems (page 401). But, and this is key, he follows up by noting that such a claim would have to rest on an exploration of the interaction of these two bargaining problems. As it happens, that exploration has occurred! It’s shown how uncertainty can mitigate the commitment problem, and whether or not it does rests on—you guessed it—a variable that the second claim tends to omit.

So, how do information and commitment problems interact?

I worked on this question with Dan Reiter and Cliff Carrubba a few years ago, and though the article has mostly garnered attention for what it says about the duration and outcome of war, it says a lot about war onset as well. Our model allow R’s military power to rise while D is uncertain over R’s willingness to use that growing power (I’ll shorthand it as resolve).7

Substantively, this means that D is uncertain over whether a preventive attack on R is worth the costs. Negotiating may leave D vulnerable to a risen R willing to leverage newfound strength, but fighting may prove wasteful if R’s newfound strength wouldn’t have been used. And R, of course, has no incentive to admit that it’s willing to leverage that newfound strength—otherwise, it might provoke a preventive war.

Commitment problem? Check. Information problem? Check.

Now, what happens?

Let’s start by adding an information problem to a commitment problem. Suppose that power is shifting enough for a preventive war to be attractive if D knows R to be resolute but not if R’s known to be irresolute. In that case, then the effect of adding uncertainty depends on D’s knowledge under complete information; that is, whether war would occur or not if there were no uncertainty. If D otherwise knows that R is irresolute, then introducing uncertainty can make preventive war more likely. I suspect that’s what Fearon’s page-401 aside was about. But if D otherwise knows that R is irresolute, then introducing uncertainty plainly discourages preventive war—increasing D’s confidence that it might be able to risk allowing R to rise. So, yes, private information can encourage preventive war, but it can also discourage it. And whether it does so depends on the complete-information knowledge to which we’re comparing the introduction of uncertainty.

Now, let’s introduce a commitment problem to an information problem. Suppose that, without shifting power, D will risk war against R when the latter’s resolve is private information, but it won’t risk war under complete information. Easy enough. But now suppose that R is also rising. This gives us two cases to consider.

  1. If D would attack a resolute R but not an irresolute one under complete information, then adding a commitment problem of this size can increase R’s willingness to risk war, because ending up in a war with the resolute type isn’t so bad.8
  2. If D would attack R regardless of its type (that is, for very large shifts in power), then the information problem is simply irrelevant. Uncertainty exists, but war would occur whether or not it was present.

This is all to say that whether information problems exacerbate or mitigate commitment problems depends on the declining side’s strategies under complete information; and whether commitment problems exacerbate information problems or make them irrelevant depends on just how far power is shifting. Yet only one of these interactions—between information problems and moderately-shifting power—gives us the “add these problems together and get more war” story. Still, if we begin from commitment problems rather than information problems, we can see the latter mitigating the former.

Is this a mess? You bet it is.

But we gain nothing, nothing, by sweeping said mess under the rug and giving our regression coefficients a potentially incorrect interpretation. How these bargaining problems interact depends on variables that, if we’re to settle on useful explanations and good interpretations of empirical models, shouldn’t be omitted.


Information and commitment problems are analytically distinct bargaining frictions. The latter isn’t a special case of the former, however tempting that argument can seem. What matters is how the two problems interact. Information problems can exacerbate or mitigate comment problems, depending on whether war would’ve happened under complete information. And commitment problems either exacerbate information problems or make them irrelevant despite their presence, depending on the size of the shift in power. Unconditional claims about their interaction are thus prone to omitted variable bias—a problem which needs no adjective like “serious” or “pernicious” to be, you know, serious and/or pernicious.

  1. Phil Arena has an excellent post focusing on these issues in more depth. ↩︎
  2. Fearon identifies a third bargaining friction, indivisible issues (pages 389-390), which he notes are likely rare given the possibility of side payments; yet Goddard (2006) shows that indivisibility can be socially constructed, and Powell (2006) saves the idea, though, by showing that indivisible issues are really special cases of the commitment problem. ↩︎
  3. Fearon (1995) makes this point explicitly on pages 403 and 406; Powell (2006) systematizes the insight across several different substantive contexts. ↩︎
  4. For formal treatments, see Leventoglu and Slantchev (2007), Reiter (2009), Powell (2012); and for an informal discussion around a simpler formal model, see Wolford (2019 Chapter 11). ↩︎
  5. See Chadefaux (2011), Debs and Monteiro (2014). ↩︎
  6. Leadership turnover is a good example, because changes in preferences from one leader to the next are discontinuous, even if the probability of turnover can be manipulated (see Wolford 2012, 2018). ↩︎
  7. This is only one way that the problems can interact, but the results aren’t substantively different if the declining side is uncertain over, say, the size of the shift in power. Furthermore, it’s sufficient here to show that a coherent exception to the claim exists. ↩︎
  8. This also leads to the most interesting results in the WRC paper (pages 568-571), because it leads to wars in which the revelation of information over time leads to the continuation, not the termination, of war. ↩︎


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s