U.S. presidents frequently employ unilateral action, for instance in the form of executive orders. But why don’t they do more? Why stop at orders banning Muslim immigrants from our shores? Why not direct the IRS and FBI to harass and arrest political enemies? Why not shut down hostile newspapers, TV networks, and social media companies? Why not command the military to fire on civilian demonstrators? Why not close Congress and shutter the Supreme Court? Why not halt presidential elections?

In The Myth of the Imperial Presidency, Dino Christenson and Douglas Kriner use the tools of modern behavioral social science to address these questions—at least, to a degree—and they offer a sophisticated and nuanced answer. In their account, the real brake on presidential tyranny—for let’s call a spade a spade—is public opinion. But not unmediated public opinion. They present strong evidence that many Americans have no principled objection whatever to presidential unilateralism; in fact, most citizens welcome vigorous executive action if it comes from a fellow partisan and delivers palatable results. To a first order of approximation, partisanship rules all. However, they also show that citizens respond sensitively to cues from Congress and the courts. When those bodies sound fire alarms about executive overreach, citizens respond—or so their evidence suggests. In some sense, then, the ultimate judge of the president’s conduct must be the public. But the public-as-magistrate condemns autocracy only if the other branches first act as grand jury and prosecutor. Failure of those branches to do their jobs implies failure of public opinion to cabin presidential tyranny.

This argument seems plausible. Seeing the right answer and proving the answer right, however, are two different things. How do the authors fare as they muster logic and evidence in support of a sophisticated defense of American democracy?

The problem is twofold. The first difficulty is nearly intractable selection bias. Vigorous use of presidential unilateralism—say, suspending habeas corpus, deploying the military in violation of the Insurrection Act, placing judges under house arrest to negate adverse judgments—lies off the recent equilibrium path. Modern presidents have not done such things, yet. (I pick these examples because they occurred during the Civil War.) Critically, presidents take such actions only in extraordinary circumstances. When those circumstances transpire, many citizens will support the president’s actions. We know this because citizens apparently did before. Conversely, absent those circumstances, presidents do not act as American Caesars: they self-censor, believing that extreme unilateralism will bring harsh consequences unto themselves. Given self-censoring, we do not have the data we need to assess public responses to Caesarism, whether warranted or not.

The self-censoring of presidential actions is only half the problem. The other half is well captured by the sign at Black Lives Matters demonstrations to the effect that if you ever wondered what you would have done during the civil rights movement or the Holocaust, you are doing it now. Putting aside quibbles about the equivalency of events, the point is psychologically acute. We all imagine ourselves moral. But in fact, history shows that most people are not moral heroes: most people are moral cowards, and some are dark villains under the right circumstances. There is only one way to cut through completely sincere self-deception and find out who a person truly is: turn the screws to put them under genuine moral pressure and then see what they do. That is why the greatest of all social psychological experiments, the Milgram Electric Shock Experiment, is more revealing about the off-the-equilibrium-path nature of Americans than any survey can be—including top-of-the-head answers to hypothetical vignettes in survey experiments. Because of sincere self-deception and a lack of imagination, responses to vignettes about presidential tyranny have essentially zero external validity. They predict nothing about actual behavior when the going gets rough. They only predict answers to other surveys in happy times.

How do Christenson and Kriner deal with these formidable problems? They appreciate the problem of presidential self-censoring, which mostly robs us of the natural experiments we would want—as social scientists, if not as citizens. Their response is to rely almost exclusively on survey experiments, not observational data (with an exception I discuss shortly). In this sense, their methodology is radical.

They deal with the “zero external validity” critique in the following way. They do not confront subjects with vignettes about extreme executive overreach. Instead, they gently prime subjects about actual executive actions, like Obama’s executive order lowering student loan payments. Then, they randomly offer anodyne vignettes, such as “Many members of Congress from both parties, however, oppose the President’s decision. They argue that Present Obama’s actions have overstepped his constitutional authority and that a major change in student loan policies requires new legislation from Congress” (p. 81). These vignettes stay extremely close to the lived experience of the subjects, so close that, plausibly, self-deception and a lack of imagination do not eviscerate response validity. The responses probably do tell us something about how the public might evaluate presidential unilateral action on student loans in the face of institutional fire alarms. Though absent a real-life treatment (headlines, news programs, tweets, talk radio harangues, arguments with crazy uncles over Thanksgiving dinner, a punch in the face), one cannot be entirely sure. Nonetheless, as the authors imaginatively cycle through variant after variant on this procedure, they accumulate little tesserae that, when assembled into a mosaic, create a rough but arguably fairly credible portrait of the US public’s response to normal executive orders under normal circumstances.

Yet sticking so close to the equilibrium path comes at a cost. Although it is suggestive about the operation of quotidian politics, the method says nothing about what we really want to know—namely, how Americans would respond to genuine Caesarism and whether it is anticipation of those responses that keeps presidents on the far side of our Rubicon.

Chapter 5 offers a departure from the authors’ parade of careful, imaginative, and incremental survey experiments. Here, they interrogate observational data. In particular, they treat the monthly count of significant executive orders as a dependent variable, and presidential approval as a key independent variable, along with some confounders. They estimate vector auto-regression models and argue that presidential approval causes (in Granger’s sense of causality) the number of executive orders, rather than vice versa. In other words, presidents appear to issue many executive orders when they have a bulging account in the bank of public opinion. The reader will recall that Granger causality means, “If A comes before B, A causes B.” So, in the infamous quip, Christmas cards Granger-cause Christmas. Scholars drawn to the credibility revolution in the social sciences will not find this approach to causality very persuasive. They will see only correlation, not causation. Such scholars would look instead to natural experiments involving executive unilateralism, creating a degree of randomized executive action with subsequent public response (Joshua D. Angrist and Jörn-Steffen Pischke, Mastering ‘Metrics’: The Path from Cause to Effect, 2014). Some of President Trump’s executive orders seem tailor-made for this approach.

Let me offer a different caveat. The authors appreciate that executive unilateralism involves severe selection bias, but they do not offer a carefully articulated theory of unilateralism in the shadow of public opinion. From a theoretical perspective, their implicit foil is Howell and Moe’s famous model of executive orders (Terry M. Moe and William G. Howell, “The Presidential Power of Unilateral Action,” Journal of Law, Economics, and Organization, 1999; William G. Howell, Power without Persuasion: The Politics of Direct Presidential Action, 2003), which rapidly became the standard theory of executive unilateralism in American politics. In that model, the president plays a bargaining game with Congress and the courts, exploiting a first-mover advantage to place a policy in the most favorable region of a gridlock interval. This action blocks opponents from subsequently overturning the order. Critically, the Howell-Moe model affords no role whatever for public opinion, voters, challengers, or fire alarms. I am not aware of any model of unilateralism that does, although some recent work on Eastern Europe is suggestive (Monika Nalepa, Georg Vanberg, and Caterina Chiopris, “Authoritarian Backsliding,” 2019, https://www.monikanalepa.com/uploads/6/6/3/1/66318923/paper_june_2019.pdf).

So, we have no actual theory of public opinion, executive unilateralism, and institutional fire alarms, and The Myth of the Imperial Presidency does not create one. This lacuna has an unfortunate consequence: absent a model of unilateralism in the shadow of electoral accountability, the authors can only speculate about what they should see in observational data and whether it is compatible with the (imaginary) model. To put it another way, they do not model the selection mechanism and therefore cannot draw inferences from observational data that account for selection bias.

I have emphasized the limits of The Myth of the Imperial Presidency; I should also make clear my admiration for Christenson and Kriner’s achievement. Their survey experiments show convincingly that unmediated public opinion is a weak reed on which to hang American liberty. They sketch an alternative, sophisticated conception of institutional fire alarms in a separation-of-powers system. What a good idea! It is deep, novel, and plausible. Their batteries of survey experiments demonstrate that institutional fire alarms have the potential to mobilize public opposition against an overreaching executive. This accomplishment should engage serious students of democracy, both empiricists and theorists, positive and normative.

Of course, whether the threat of congressional and judicial fire alarms will prove sufficient to check an aspiring autocrat and whether Congress and the courts will actually sound the alarms on the critical day remain open questions. Concretely, were President Trump’s more extreme moments flukes or fire bells in the night? Was Watergate’s lesson “All’s well!” since the system worked? Or, was the lesson “Modern American democracy skates on thin ice,” so that past luck is a poor predictor of future good fortune?

From this perspective, The Myth of the Imperial Presidency offers scant reassurance to those who fear presidential tyranny is a real dragon, not just a mythical one. But even if The Myth of the Imperial Presidency lays to rest no myths, it opens new doors for thinking about the public, democracy, and the American presidency. This fine book will immediately find a place on graduate reading lists, and it deserves to.