Identifying and estimating “treatment effects” is central to modern statistics. In the experimental sciences, randomizing treatment and controls groups is the gold standard of identification, and estimation is accomplished by comparing the mean change in an outcome across treatment and control. Randomization is not possible in the observational sciences, of which history could be said to be a prime example. However, a key insight of recent econometric thought is that one might mimic randomization through carefully analyzed “natural experiments.” In a natural experiment, the researcher makes a case that treatment and control is “as good as” randomly assigned. In economics, typically this is accomplished by a difference-in-difference regression analysis with careful attention to preexisting trends and balancing of covariates so that the treatment and control groups look as similar as possible (“observationally equivalent”) prior to treatment. Difference-in-difference is often supplemented by an instrumental variable—a variable that is correlated with the treatment but uncorrelated with the outcome of interest.

Although economic historians were initially slow on the uptake, the natural experiments approach has become established in cliometric circles in recent years. The Jared Diamond and James A. Robinson volume is intended to publicize the use of natural experiments in history to a wider audience. The authors are an interdisciplinary bunch, individually prominent in their respective fields.

The book begins with an Introduction extolling the virtues of the approach, followed by seven substantive chapters and an Afterword. The first four substantive chapters are by noneconomists. Patrick Kirch investigates differences in social and political structure across three Pacific Islands that were originally populated by a single group (early Polynesians). James Belich reviews frontier experiences across the nineteenth-century world, arguing that the experiences shared a common three-pronged growth-decline-resurgence pattern—rapid population growth followed by bust followed by exports linking the frontier to a distant urban area. Steve Haber surveys the history of banking in the United States, Brazil, and Mexico, concluding that democracy enhanced access to credit. The first part of Diamond's chapter puzzles why the Dominican Republic is (relatively speaking) well off compared with Haiti while the second half conducts a statistical analysis of 81 Pacific Island societies to explore factors that might explain high rates of deforestation on Easter Island.

The final three chapters, all by economists, are probably of greatest interest to readers of this JOURNAL. Nathan Nunn asks if poverty in Africa today can be traced back to the slave trade. Nunn's instrument for the slave trade measures the distance from a country by sea to the major slave markets in the New World and the Indian Ocean or the overland distance to the closest port of export for the Trans-Saharan slave trade. The IV treatment effect is large and negative, suggesting that the slave trade was bad for Africa. Abhijit Banerjee and Lakshmi Iyer's chapter asks whether institutional differences in land tenure established during British colonial rule in India still have effects on public goods provision and political participation. Under one system, landlords functioned as tax farmers, collecting taxes, paying a portion to the British and keeping the rest. Under another system, individual cultivators were responsible for their tax payments. The village system was somewhere in the middle. There was considerable variation across nineteenth-century India in the use of the three systems, which the authors argue was largely random. The principal result is that in areas of India today that were formerly under the landlord system, public goods provision is lower, possibly because of less political competition and lower levels of political participation. Daron Acemoglu, Davide Cantoni, Simon Johnson, and James Robinson attempt to ferret out treatment effects of the French Revolution. The claim is Napolean's decisions to invade portions of Europe were not predicated on future growth prospects, and so one can measure the treatment effect of the Revolution through a difference-in-difference analysis. The outcome is urbanization, which the authors interpret as a proxy for per capita income. The treatment effect is positive—the French Revolution contributed to modern growth. The Afterword concludes with a reprise of major themes and prospects for future work.

Natural Experiments is a mixed bag. I agree that natural experiments methods are of immense potential value in economic history. I use this approach more or less exclusively in my own work and teach it to my graduate (and undergraduate) students. The book is clearly useful in showing that the methods can be applied to a wide variety of historical settings by researchers in different fields.

At the same time, Natural Experiments has significant downsides. From the very beginning its aims are too grandiose. In their “Prologue,” Diamond and Robinson appeal to historians to return to quantitative work, arguing that the natural experiments method is an effective anecdote to traditional cliometrics. Speaking as someone who witnessed them firsthand at the tail end, the debates over Time on the Cross had little or nothing to do with identification strategies or omitted variable bias. I also find it difficult to believe that cultural historians will fall over themselves to embrace instrumental variables in their daily life, eschewing postmodernism. It is telling (and more than a little sad) that none of the authors is employed in a traditional academic history department.

The book is very uneven in its discussion of statistics. On the one hand, the reader is expected to be quantitatively sophisticated enough to comprehend the use of instrumental variables if not their practical implementation. At other junctures, readers are subjected to excessively elementary treatments of basic statistics. At no point is the discussion remotely sufficient to enable a reader to figure out how to analyze her own natural experiment. Such a reader, I suppose, could read Nunn or Banerjee and Iyer in the original (the Quarterly Journal of Economics and the American Economic Review, respectively) or many working papers of Acemoglu and coauthors but then why buy this book?

Natural Experiments purports to sell a method but like the salesperson knocking at your door, there is bait-and-switch. Overall, the noneconomist chapters are disappointing. Kirch and Diamond are entertaining but at best tangentially related to natural experiments methods as practiced in economics while Belich and Haber are pretty much conventional comparative history. The three economics chapters belong to a branch of the “big think” school of economic history investigating whether “institutions matter” in the very long run. Much of the work in this branch, such as the chapters by Nunn or Acemoglu, Cantoni, Johnson, and Robinson, is fundamentally cross-country in design. The book gives short shrift to two serious issues with this line of work—namely that sample sizes are usually very small and data problems can be so severe that robustness is easily compromised. Some of the best natural experiments work in economic history is not about whether institutions matter in the very long run and, as such does not appear in the book, thereby shortchanging the reader.

Criticisms aside, Natural Experiments can be useful supplementary reading in undergraduate courses in world economic history, or graduate courses in anthropology, sociology, and political science with a quantitative, historical focus. Perhaps some historians who read the book will be lead to reinvestigate the use of quantitative methods and that, too, would be a very good thing.