Yes, It Matters that Sweden Could Join NATO

A new wave of academic research confirms that alliances matter

It’s been a dramatic week in NATO politics. After months of stonewalling Sweden’s historic bid to end its neutrality and join the alliance, Turkey’s president, Recep Tayyip Erdogan, agreed to drop his opposition. With Sweden’s way clear to join the alliance, NATO will soon expand again.

Swedish neutrality has long been a big deal (Sweden is a big deal and it used to be the major power of northern Europe). If you’re interested in international relations, however, you might be able to think of reasons why joining an alliance isn’t a big deal. Hey, aren’t treaties just scraps of paper? More to the point, if states have interests, and if NATO’s pledge to come to Sweden’s aid is only credible if defending Sweden is in NATO’s interest anyway, then why would the treaty itself matter? Surely, a certain type of realist would argue, treaties just don’t matter very much.

Quantitative research into alliances has long suggested that they do matter, not least because a treaty engages a state’s reputation in the eyes of its leaders and other states. A recent trend of research, however, suggests that public opinion in countries that sign treaties of alliance also matters, with publics more likely to support interventions on behalf of treaty allies than other countries. In fact, it might matter a lot to how much publics will support action on behalf of countries attacked by Russia.

I happen to be a part of this set of findings through an unexpected path.

A couple of years ago, a very bright undergraduate asked if I would supervise her thesis. She’d just come back from an internship at NATO headquarters, and she was interested in how the public viewed alliance commitments in cyberspace. I knew that NATO had taken an interest in cyber activities but I also knew that not many scholars had examined the credibility of its commitments in cyber—even though it’s obviously not clear how significant an attack would have to be in order to trigger an alliance designed to protect against physical attacks. (There’s some complicated arguments here but they’re lawyerly in the best and worst ways—the sorts of distinctions that make commitments more specific but hardly any more binding.)

Let’s be clear. Alliances + credibility + cyber is an awesome dissertation idea, much less an undergraduate thesis. So I was thrilled by the idea.

Her idea—very good!—was to test this using a survey of all (then) 29 NATO members. I noted that would probably cost a few hundred thousand dollars and require us to learn Macedonian. We compromised on a survey in one country, the United States. Lindsey set about learning about NATO, cyber, alliance theory, survey experiments, and human subjects ethics, and I advised her work. Within a few months of in retrospect intense, COVID-isolation-fueled study, Lindsey had prepared a set of survey experiments and we fielded them online. The result became her senior thesis and then, after some rewriting, additions, and new data analysis, a submission to peer-reviewed journals. Soon after that, and the inclusion of some new experiments suggested by a reviewer, the work that began as Lindsey’s undergraduate thesis was published in the Journal of Global Security Studies. (And the first-author order here is, as I think this makes clear, not alphabetical.)

We found that whether a country had a treaty alliance really does matter. Specifically, we randomized whether a Russian cyber attack took place against a country that was a NATO member or a non-NATO member. It was 2021 when we wrote this (before the Ukraine invasion) and we thought we were dealing with hypotheticals—a reminder of the political scientist’s curse: “may your research be relevant”. We knew that it was likely that we would find some effect of a treaty on public support, not least because earlier work by Michael Tomz and Jessica Weeks had found a large treaty effect.

But we also thought there was more to be done. Tomz and Weeks had randomly varied whether a country that was attacked was a treaty ally of the United States or not, as well as other factors. Their headline finding was that there was a 33 percentage point increase (from 46 to 79 percent support) in the treaty compared to non-treaty condition. This is huge. If anything, it might be too large. Perhaps respondents assumed that there was something different about the country that had a formal treaty compared to the one that wasn’t—having such a treaty, after all, might mean that there’s something special about that country beyond whether it’s a democracy or not.

So we decided to test the influence of an alliance in classic high-school science-fair ways. We’d find a way to test whether NATO alliances mattered by comparing two countries that were as similar as possible on everything but whether they were a NATO ally or not. And it just so happened that there were two such countries: Norway, then a NATO member, and Sweden, then not. They’re both democracies, both rich, and—crucially—both had longstanding defense arrangements with the United States. (During the Cold War, Sweden’s military cooperated with the rest of the West, and U.S.-Swedish defense ties have been deepening for a while.) Recall that our aim was to find out how much a treaty itself mattered, not anything else that could be potentially associated with a treaty. And we crafted our experiment so that respondents would know that Norway was an ally but that Sweden was a non-ally partner.

(I should note here that I told Lindsey that Norwegians and Swedes would scream if we told them that we assumed Americans would view their countries as interchangeable but-for their treaty status. In reality, they’ve mostly shrugged and said they expected nothing less from Americans.)

Notably, we were interested in cyber operations, not full-scale invasions—the sort of (forgive me) grazyone operations people cared about during the pre-2022 invasion era. So our test is somewhat apples-to-oranges compared to Tomz and Weeks. However, we also found that a treaty alliance mattered, with respondents about 10 percentage points more likely to endorse coming to the aid of NATO member Norway than non-NATO member Sweden after a cyber attack.

So, congratulations, Sweden! You’ve now got 10 to 33 percentage points more hypothetical American support, just by joining NATO. (I should note there’s a ton of other findings in this paper, so if you care about NATO, cyber, or U.S. foreign policy, read it—and if you don’t have access, ask me and I’ll send it to you.)

The irony is that Tomz and Weeks (with Kirk Bansak) managed to do a version of the paper I’d told Lindsey was going to be too expensive. In recently (very recently) published work, they carried out survey experiments in 13 NATO countries and found a 23 to 33 percentage point increase in willingness to defend NATO allies from invasion relative to non-NATO allies. This is much bigger than our cyber effect, and interested parties should send me money so I can figure out why. (That’s no joke.) But it’s yet another piece of evidence suggesting that alliances matter.

I hope to share more such research with you soon. In the meantime, just reflect on a couple of things. First, undergraduate research can really matter and can even be on the cutting edge of substantive importance. Second, political science research can really matter and it’s getting incredibly competitive and rigorous. Third, go read my paper with Lindsey.