August 19, 2021

I value Twitter because it helps me stumble upon interesting people and information that I might not otherwise encounter. But I’ve also had a propensity to doomscroll when breaking news feels overwhelming.

About a year ago, I came across a research finding that going down an internet rabbit hole can actually stimulate creative thinking, so long as you start with a specific question. So now, when I find myself starting to doomscroll, I try to come up with a specific question related to the news — preferably one that can make me a more informed citizen — and then veer off to investigate that. Dear reader, you will not be surprised to learn that I have found this more enriching than doomscrolling.

On Monday night, amid the tragic images coming out of Afghanistan, I felt myself starting to doomscroll, and decided to come up with a question I could investigate instead. I had already watched several clips in which pundits proclaimed a U.S. “credibility” crisis (apparently a long-running hobbyhorse from some of those pundits). Breaking news articles declared that “Western deterrence now looks hollow,” and described “the erosion of American credibility and deterrence.”

Each credibility-crisis hot take was a little different, but the theme of what I saw was that U.S. credibility is now damaged among allies, which will lead to more aggression in the long run from adversaries who perceive both U.S. resolve and alliances to be weak. Former Vice President Pence encapsulated the theme pretty well in writing that this week’s news has “caused allies to doubt our dependability, and emboldened enemies to test our resolve.” (This same theme made headlines when President Trump withdrew troops from northern Syria in 2019.)

The question I alighted on: What usually happens? More specifically, do past military actions impact a country’s ability to deter violence against allies in the future?

That question quickly led me to an interesting paper: “What Makes Deterrence Work?” (You can register for a free account to read it.) The paper examined 54 instances over the twentieth century in which one country (the “defender”) sought to deter another country (the “attacker”) from attacking a third country (the “protégé”).

The researchers examined all sorts of variables to see what influenced deterrence, from a defender’s prior willingness to fight, to the existence of formal alliances between defender and protégé, to the relative strength of defender and attacker militaries, to the possession of nuclear weapons. And their conclusions are, I think, somewhat surprising.

Formal alliances did not predict deterrence success, and neither did the overall strength of a defender’s military. Three factors predicted a defender’s likelihood of forestalling an imminent attack on a third party: 1) Robust trade between the defender and protégé 2) The protégé getting most of its weapons from the defender 3) Local military might — the strength of the military force on the ground in the country being defended right now. That’s especially interesting because it means that if, say, the U.S. is the defender, the fact that the U.S. military is far mightier than the attacker did not matter unless that might was presently apparent on the protégé’s ground. Military potential, in other words (including possessing nuclear weapons), did not matter, whereas present military commitment on the ground did.

I’m cramming a lengthy paper into a newsletter-nutshell, but the overall picture I took away was that attackers are usually deterred when a defender has an enormous economic interest in the protégé country, and has shown how eager they are to defend it; defenders express that eagerness through providing the protégé with loads of sophisticated weaponry and boots on the ground.

The importance of economic interest, while intuitive, really jumped out. The greater the share of a defender country’s imports and exports that a protégé country accounted for, the more likely the defender country was to successfully deter the onset of violence. According to the model the researchers constructed, when a defender and protégé did not have a business relationship, the chance of the defender halting a threat from an attacker before violence ensued was a coin flip. If a protégé accounted for 6% of a defender’s imports and exports, the model predicted an 86% chance of successful deterrence. Defenders, it seems, put their money where their money is.

So, ya know, follow the money — what else is new? But it’s interesting that would-be attackers are clearly gathering strong signals about the potential consequences of violence based on how economically linked the defender and protégé are.

I found two other conclusions of the paper surprising and interesting. First, based on this study, the “credibility” argument I saw while doomscrolling doesn’t look convincing. The researchers, Yale professors Paul Huth and Bruce Russett, note that “the ‘need’ to demonstrate resolve was a key rationalization for America’s experience in Vietnam and throughout the post-Munich era,” but they conclude that “the defender’s past behavior against a threat does not predict whether the defender will be able to succeed in deterring a later attack.” In fact, “the defender’s past behavior in crises seems to make no systematic difference.” Past performance — apparently not a predictor of future results when it comes to deterrence. (Once I latched on to this credibility question, I saw that journalists and political scientists have been writing about the “credibility trap” for years.)

The second conclusion that ran contrary to my intuition was that formal alliances actually had a slight negative influence on deterrence success. Attackers, according to the study, do not appear to value formal alliances as a signal of a defender’s commitment. France famously signed off on Hitler’s annexation of part of Czechoslovakia (in return for a promise of peace), despite a formal military alliance between France and Czechoslovakia meant specifically to deter German aggression. As the paper authors note, alliances didn’t deter the onset of violence in either world war, “although the alliance system did insure that war, once begun, would cascade to include most of the great powers.”

Then again…

…this paper was merely a starting point in a search for something more thoughtful than what doomscrolling was getting me. (Low bar, I know.) I’m not entirely convinced. This kind of mathematical modeling of the past in order to predict the future is fraught with statistical pitfalls. It is very important to ask what usually happens — it is, in fact, known as “reference class forecasting” (a topic in chapter five of Range) and has been shown to improve judgment — but there are many devils in details of any specific model.

The fact that, in this paper, formal alliances had a small negative influence on successful deterrence (as opposed to just no effect) should raise a flag. Does a formal alliance really contribute to deterrence failure? More likely is that attackers — who are aware of the alliance — only threaten an allied protégé when they’re really serious about attacking. (Huth and Russett, to their credit, acknowledge this likelihood.) We shouldn’t conclude from this paper that not having alliances would make for better deterrence, even though that’s what you might think from a context-free interpretation of the study results. Models are tools to aid thinking, not license to turn off your brain. So after reading Huth and Russett, I went looking for contrary studies.

One that took them on directly (you can read it with a free account) found that the characteristics that Huth and Russett highlighted do matter for deterrence, but that overall military strength and alliances also matter. Another contrary study found that reputation from past behavior does matter for future deterrence, but most strongly in repeat interactions with the same attacker as part of the same larger conflict, and less strongly when subsequent conflicts aren’t identical to past ones.

Maybe the overall picture, as a 2013 Washington Post article about Syria policy put it, is that attackers are “smart enough to know that different situations are, well, different.” Reputation may matter, but as part of a complex stew of deterrence that is much more nuanced than headlines would have me believe.

That might be an unsatisfactory conclusion, but the inquiry that led me to these papers has me thinking more critically about the news. I ended up seeing more complexity than simplicity, which does feel a bit confusing, but I also think it’s more representative of reality.

Next time you’re stuck in a late-night doomscroll, zoom out for a moment. Come up with a question that might prompt productive searching instead. You might be surprised by what you find, just like I was.

Thanks for reading. Until next time…

David

P.s. Just a reminder: in coming weeks, this newsletter will relaunch on another platform. Your subscription will automatically transfer, but you will receive a notice allowing you to opt out. And if a friend sent you this newsletter, you can subscribe here.