Making History Count: A Primer in Quantitative Methods for Historians

Published by EH.NET (March 2004)

Charles H. Feinstein and Mark Thomas, Making History Count: A Primer in Quantitative Methods for Historians. Cambridge: Cambridge University Press, 2002. xxi + 547 pp. $85 (hardback), ISBN: 0-521-80663-1; $30 (paperback), ISBN: 0-521-00137-4.

Reviewed for EH.NET by Walter D. Kamphoefner, Department of History, Texas A&M University.

Assume a Data Set

The two co-authors, Charles H. Feinstein, now Emeritus at Oxford University, and Mark Thomas of the University of Virginia, are both specialists in British economic history, and they draw upon their experience teaching methodology courses at their respective institutions. With characteristic British understatement, this 500-odd-page tome (plus appendices, bibliography, and indexes) is designated a primer. The characterization is apt inasmuch as their approach was “designed primarily for students who had no prior knowledge of statistics, and who were in many cases initially hostile to, or intimidated by, quantitative procedures” (p. xxi). However, the degree to which the book succeeds in banishing the hostility or intimidation remains, in my opinion, open to question and depends to some extent on the mathematical facility of the reader. Although it is the authors’ professed assumption that readers “have no retained knowledge of statistics, and very little of mathematics beyond simple arithmetic” (p. 3), if so, that “little” has to go a long way.

In the course of fifteen chapters, the reader is taken from elementary concepts such as levels of measurement and convention of notation up through multiple regression with many of its technical considerations for both cross-sectional and time-series analysis, plus a chapter dealing with logit, probit, and tobit models appropriate to limited dependent variables. Some of this material, particularly in the later chapters, can be quite technical. Still, a tripartite division allows the text to work effectively for readers of various levels of statistical competence. Essential definitions and concepts that need to be studied carefully by all readers are set off in shaded boxes. Otherwise set apart are 22 “panels” discussing more technical material that may be skipped over by neophytes. Overall, one of the strengths of the book is its rich illustrative material, comprising some fifty tables and eighty-five figures or sub-figures. Nine pages of bibliography and both a name index and a subject index round out the book and help make it accessible. Data sets are provided for access at no charge from the Cambridge University Press website (p. 7). Both the authors of the text and the scholars who originally assembled these data sets are to be commended for making them available. Each chapter concludes with several pages of helpful study exercises, based in part on these four supplemental data sets

The bulk of the illustrative material is drawn from the authors’ own research specializations: British economic history. As a result, whole fields of quantitative history are left out, along with the particular applications and examples that might be useful for readers. For example, the set of ecological inference techniques applied to voting analysis, both the older work associated with J. Morgan Kousser and the newer developments by Gary King, are totally absent. In fact, one looks in vain for any citations of the leading American quantitative journal, Historical Methods, in the bibliography. Three of the four case studies used as examples in the book are based in Britain or Ireland; only one is from the United States.

There are two serious shortcomings of the book that should be noted. Although largely limited to economic history and British examples, the book does indeed, as promised in the introduction, “relate the quantitative techniques studied to examples of their use by historians and social scientists” (p. 3), but only in a very bloodless way, devoid of drama or personality. From reading it, one would hardly suspect that there are any scholarly controversies arising from quantitative analysis, much less learn about competing hypotheses or paradigm shifts, and the ways in which quantitative techniques or approaches have contributed to them. The contrasts with Edward Tufte, The Visual Display of Quantitative Information (Cheshire, CT, 1984); and Envisioning Information (Cheshire, CT, 1994), could hardly be greater.

Secondly, one is reminded of the joke about the economist discussing the problem of being trapped in a well, who begins his solution by assuming a ladder. Here, what is assumed is a data set. Unless one happens upon the discussion of index numbers in Appendix B, one might come away with the impression that data sets fall from the sky ready to be analyzed. Paradoxically, the way the book deals with data is for my taste both insufficiently theoretical and insufficiently practical. Although the skeptical Sir Josiah Stamp does make a token appearance (p. 319), the question of epistemology, of what is knowable or measurable, never comes up in the book. That economic historians hate social constructionism is perhaps understandable; that they refuse to acknowledge its existence or address any of the objections it raises, less so.

Also at the other end of the spectrum, on the concrete, practical level, very little attention is given to the raw material to which quantitative analysis is applied, where it is obtained from, and what intermediate steps need to be taken or decisions have to be made before multiple regression can be applied to it. This is by no means a nuts-and bolts handbook. The last decade has seen a breakthrough in the provision of public use samples. The most important such venture has produced what is known as the IPUMS (Integrated Public Use Microdata Series) at University of Minnesota, which at last count includes 27 samples covering the period 1850-2000 (www.ipums.umn.edu/). Originally restricted to the United States, it has now advanced to an international phase, applying similar techniques (nationwide random household-based samples of population censuses encompassing individual and household characteristics, uniform coding of variables) to various other census data around the world. Research opportunities, especially for cross-sectional analysis, have grown exponentially. Yet there is no mention in the entire book of these data projects or any of the scholars involved.

Within these limitations, this is very sound handbook; I did not detect a single statement that was methodologically questionable. It may suit the needs of the core constituency of this list rather well. Among the relatively short list of competitors still in print, Hudson is much more basic and similarly “insular.”[1] Jarausch/Hardy, while also less technical, includes a section on cross-tabulation statistics that goes much beyond that in the work under review.[2] Darcy/Rohrs provides excellent guidance in how one gets from sources to data sets, and has much to offer in the realm of political history and voting analysis. It mentions logit only in passing, but gives a competent overview of techniques up to and including multiple regression.[3] But for anyone teaching, say, time series analysis and its special problems and diagnostics, Feinstein/Thomas is probably the logical choice.

Notes:

[1] Pat Hudson, History by the Numbers: An Introduction to Quantitative Approaches (London: Arnold, 2000; New York: Oxford University Press, 2000).

[2] Konrad H. Jarausch and Kenneth A. Hardy, Quantitative Methods for Historians: A Guide to Research, Data and Statistics (Chapel Hill: University of North Carolina Press, 1991).

[3] Robert Darcy and Richard C. Rohrs, A Guide to Quantitative History (Westport, CT, 1995).

Walter Kamphoefner teaches Quantitative Methods in Historical Research and Problems in Quantitative Methods in the Department of History at Texas A&M University.