According to Wacziarg (italics mine for emphasis):
The book's substantive contribution is to bring ... intellectual traditions together under a consistent theoretical framework, delivering rich empirical predictions on the factors leading to democratization and institutional stability. These factors are the economic, social, and institutional determinants of the costs and benefits of repressing versus conceding, such as the size of the middle class, the structure of production, economic inequality, the prevalence of economic shocks and crises, and the degree of globalization. Each is found to have sometimes complex, but always empirically testable, relationships with the political regime.
Yet providing testable theories is not the book's only contribution to scientific knowledge. Another, perhaps more important, contribution is the manner in which its central questions are approached. Economic Origins of Dictatorship and Democracy contributes to social science by addressing issues traditionally studied by political scientists with the rigorous tools of economic analysis. Acemoglu and Robinson use formal game theoretic models, proceeding from assumptions to empirically falsifiable predictions, in keeping with the Popperian tradition. Such an approach is relatively new to political science, in spite of that discipline's name. This is particularly true for the subdiscipline of comparative politics, which still largely relies on rhetoric and anecdotes--rather than mathematically rigorous proofs and large-sample statistical evidence--to explain social phenomena. The authors' work is an admirable illustration of a growing trend toward formal reasoning and the derivation of empirically testable propositions from internally consistent, stylized models. This is a trend that revolutionized economics starting in the middle of the last century. It is now sweeping political science, and the experience of economics suggests it is the way of the future. As it takes hold, the social sciences are bound toward greater and greater consilience.
One feature of scholarship my post suggested that law ought to value more than it traditionally has is greater emphasis on developing testable legal theories as well as actually doing the testing. While we may never get to the point of regularly producing and working with "mathematically rigorous proofs and large-scale statistical evidence," would it hurt us so much to move farther away from relying as much as we do on "rhetoric and anecdotes"?