Michael Miller’s aptly titled book does an outstanding job crystallizing a long tradition on the effect of shocks on regime change. In addition to advancing a theory with some novel theoretical twists, the book is exemplary of a methodological approach that Gary Goertz and I have called “large-N qualitative analysis.” Although the book includes thoughtful time-series cross-section statistical work and mediation analysis, its core claim rests on a simple observed regularity: that the overwhelming share of all democratic transitions from 1800 to 2014 occurred either in the wake of shocks or through initiatives by incumbent authoritarian elites who believed they could win at the democratic game. Inference comes not only—or even primarily—from the statistical work but by demonstrating through detailed case studies that the regularity is, in fact, causal.

The theoretical insight of the book centers on the observation that regime changes may not result in the fundamental shifts in power often assumed. If authoritarian incumbents are strong, then democracy is possible because they believe they can compete. As this observation has been noted by a number of scholars of democratization, particularly Dan Slater and Joseph Wong (“The Strength to Concede: Ruling Parties and Democratization in Developmental Asia,” Perspectives on Politics 11 [3] [2013]: 717–33), I will pay somewhat less attention to it. By contrast, if incumbent authoritarians are under severe challenge, they have already been weakened in some way and democracy may provide a way out.

As Miller notes, this theoretical insight arose inductively out of the empirical identification of these two causal paths; in combination, Miller claims they account for no less than 91% (127 of 139) of the democratic transitions in the dataset (an updated and slightly modified version of

Carles Boix, Michael Miller, and Sebastian Rosato, “A Complete Data Set of Political Regimes, 1800–2007,” Comparative Political Studies 46 [12] [2013]: 1523–54). Of this total, the “electoral continuity” path accounts for a distinct minority of cases (27 of 139), while the dominant “shock” path accounts for the bulk of them (100 of 139 transitions), with 12 remaining cases as potential anomalies.

The concept of “shock” is defined quite expansively to include no fewer than five subpaths, three domestic (coups, civil wars, and assassinations), and two international (defeat in war and “hegemonic withdrawal”). Although all these are defined in political rather than economic or social terms, it is clear that “disruption” is by no means limited to its political consequences. Defeat in war, for example, typically results in economic as well as political fallout. The domestic shocks Miller identifies trigger social mobilization by oppositions, which Miller treats as endogenous to the process.

Nonetheless, Miller focuses on the political dynamics. Autocracies do not go gently into the night; they reflect an equilibrium between the leadership, supporters—including in the coercive apparatus, and cowed oppositions. That equilibrium only breaks down when leaders recognize they are weakened in the wake of shocks and that clinging to power can result in worse outcomes, including to their personal safety, than ceding it. As Miller summarizes—following Barrington Moore, “no disruption, no democracy.”

Postulating an antecedent condition to an outcome and showing that it holds empirically is exemplary of a regularity approach to causation that can be traced all the way back to Hume. In this case, the claim is a necessary-conditions or “Y-generalization claim”: if the dependent variable Y (democratic transition), then the antecedent condition X (shocks). Yet this approach has run into the obvious criticism that such “constant conjunction” may or may not be causal even if apparently necessary. That is where a much more complex mechanistic account, summarized in a broad causal mechanism figure (p. 9) comes in; it is also the role ultimately played by case analysis, which in Miller’s book encompasses not a sample but the entire population; thus, the moniker “large-N qualitative analysis.”

Note that despite the relatively spare causal process centered on leader calculations about their political and personal future, the actual intervening process between shocks and democratization is, to say the least, significantly more complicated. Miller is rightly concerned about one particular selection problem: that the shocks are not motivated by democratization, which is largely demonstrated to my satisfaction. But that hardly exhausts a host of intervening steps that may also constitute necessary conditions on the way from shocks to transitions.

For example, shocks create leader vulnerability both on the supply and demand side, undermining economic and coercive capabilities and shifting the opportunity structure for opposition to mobilize. Miller even makes reference to the effect of shocks on beliefs (see, e.g., pp. 49–57). This complexity is reflected not only in the delineation of subpaths—for example, using different types of shocks—but also of subsubpaths. These routes exhibit subtly different mechanisms. Miller also draws a distinction between “primary” cases, in which the causal story appears paramount, and “secondary” cases, in which the causal process operates either in conjunction with other factors, through functional equivalents or more partially. To take but one example, the civil war case can be divided into four types: rebel victories, government victories, stalemates, and cases in which civil wars are ongoing. At a general level, the theoretical claim about the relationship between shocks and democratization in all these cases might hold, but does it matter that they in fact transition in quite different circumstances? And does it matter that the sequences may differ in ways that place “shocks” as but one condition in a more complex causal chain?

These problems have preoccupied the qualitative methods literature for some time and have now been treated with some rigor (on sequences, e.g., see Tulia Faletti and James Mahoney, “The Comparative Sequential Method,” in James Mahoney and Kathleen Thelen, Advances in Comparative Historical Analysis, 2015, pp. 211–39). But I am highly sympathetic to the approach, in large measure because of its tacit rejection of the narrow conception of causal inference that now prevail and the use of multiple methods to strengthen confidence in the theoretical priors. A striking feature of the book—and one adopted by Robert Kaufman and me as well (Stephan Haggard and Robert Kaufman, Dictators and Democrats: Masses, Elites and Regime Change, 2016)—is to attempt not simply a coding (what Miller calls a “qualitative coding”) but also what is in effect an abbreviated causal process observation for each case in the sample (again, thus “large-N qualitative analysis”). The appendix from the book provides thumbnails of each case but it is clear that each of those abbreviated “qualitative codings” draws on a much deeper well of literally thousands of sources.

In addition to the case chapters, the book has three sophisticated statistical chapters that start by confirming the core regularity using a more standard average-treatment effects setup. This requires reposing the hypotheses in probabilistic terms (e.g., “democracy is more likely after defeat in foreign war”), with the advantage of testing core claims under ceteris paribus conditions through the inclusion of controls. It only takes relatively simple plots to show the estimated likelihood of democratization in the face of shocks or a durable ruling party is significantly higher than in the country-years in which both were missing.

The more standard statistical setup is augmented by the use of mediation analysis that not only tracks particular features of the paths but also complicates the analysis by placing entire paths within deeper structural contexts (such as level of income). A final substantive chapter takes up the issue of democratic survival using standard hazard models, and opens up an important extension on the project by noting that shocks also have adverse effects on the survival of democratic as well as authoritarian regimes.

There is no such thing as a final word on a topic as expansive as regime change; the past is constantly changing as we continue to turn over the causal building blocks that make up the democratization craft. But this is an extremely impressive book not only for the boldness of its claims but also the nuance in seeking to walk through the intervening steps linking shocks and dominant parties to democratization. The story is not quite as neat as the simple formulation would suggest; shocks appear to be but one—if very important—component of a more complex story. But we make progress not simply by toiling away in the nuance but stepping back and considering the insights that arise from simple but important regularities.