ALAIN Desrosieres, Administrator of France's Institut National de la Statistique et des Etudes Economiques, had the attractive idea of combining a history of the mathematical development of statistics with a history of their use by governments. The result, The Politics of Large Numbers: A History of Statistical Reasoning,t is the kind of book that I would ordinarily recommend as professional reading for today's quantitative policy analysts. Among its other uses, such a history reminds us how new and immature our profession is. Startlingly few national data were supplied on an annual basis until recently. In the United States, a statistic as basic as the unemployment rate wasn't invented until the 1930s. Analysts could not track annual changes in family income until the late 1940s. National breakdowns of socioeconomic indicators by race did not become available until the 1950s.

The statistical techniques policy analysts now take for granted are also recent creations. The correlation coefficient was invented little more than a century ago. Survey researchers and pollsters were still learning how to take a halfway decent sample in the 1930s. The staple of contemporary policy analysis, multivariate regression, was first used for a major policy analysis in the mid 1960s, in the famous Coleman Report.

The Politics of Large Numbers thus could have filled an important role. Several good books have been written on the history of mathematics and statistics, but few on the use of statistics by governments and, to my knowledge, none that combines both narratives. Another virtue of The Politics of Large Numbers is its international sweep, encompassing the experiences of France, Germany, Britain, and the United States. Without question, the book has a lot of useful information stuffed into it. The question prospective readers must ask themselves is how much work they are prepared to devote to extracting it-for the book has one, but significant, shortcoming. It is nearly unreadable.

THERE are moments of relief. Sometimes Desrosieres presents a straightforward narrative that lasts for several pages. But these are moments punctuating a wearying struggle. I do not know how to apportion the blame between a French academic tradition that seems to value prolixity, a transnational tradition of bad writing in the social sciences, or the failings of the translator. Whoever or whatever is to blame, the writing is dreadfully convoluted even when the underlying message is simple. It is one thing to wade through difficult prose if one is rewarded with thoughts so exquisite that they could be expressed in no other way. It is maddening to do it for commonplaces.

How am I to document this point in a brief review, when everyone knows that a disgruntled reviewer can find a few opaque sentences in the most crystalline of books? Let us go to that part of a policy book where we can be sure the author wants to be clear: the last chapter, which he knows is the only part of the book that his readers are sure to read (or at least look at). Here are the opening two sentences:

Among the traits characteristic of the historical line of research begun during the 1930s by the Annales school, reference to statistical objectifications has been significant. From this point of view quantitative history has inherited, via Simiand, Halbwachs, and Labrousse, elements of the Durkheimian school and, even closer to the source, of the mode of thinking centered on averages engendered by Quetelet, who opposed macrosocial regularities to the random, unpredictable, and always different accidents of particular events.

Let me assure you that familiarity with the Annales school, Simiand, Halbwachs, Labrousse, the Durkheimian school, and Quetelet does not make decoding these sentences a whole lot easier. Now let us go to the place within the last chapter where the author has the maximum incentive to be not just clear but compelling, the very end of the book. Here, in its entirety, is the closing paragraph of The Politics of Large Numbers:

A public debate that employs statistical reason, either in support of its investments or to discuss them, is thus circumscribed by contradictory constraints. On the one hand, controversy can lead to a questioning of the equivalence and permanence of the qualities of the objects. But on the other hand, the institution of other conventions is very costly. I would like the reflection offered in this book on the relationships between statistics and the public sphere to help clarify and analyze these spaces of durably solidified forms, which must simultaneously remain undebated so that life may follow its course, and debateable, so that life can change its course.

I REST my case. The problem is not that the sentences are completely uninterpretable-here, for example, Desrosieres is making a reasonable point about the pragmatic utility of social statistics and their imperfections as representations of reality. But what a morass of words he uses to convey a simple thought. If you can discern subtleties I missed, The Politics of Large Numbers may be the book for you. Otherwise, readers may wish to wait for a more accessible history on this important topic.