Two Statistical Problems in the Princeton Project on the European Fertility Transition

The Princeton Project on the Decline of Fertility in Europe (or European Fertility Project, hereafter EFP) was carried out at Princeton University’s Office of Population Research in the 1960s and 1970s. This project aimed to characterize the decline of fertility that took place in Europe during the nineteenth and early twentieth centuries. The project’s summary statements argued that social and economic forces played little role in bringing about the fertility transition. The statement stresses instead a process of innovation and diffusion. A central feature of the EFP argument is a series of statistical exercises that purport to show that changes in economic and social conditions exerted little influence on fertility. Two recent papers on Germany for this period have used similar data and methods to draw different conclusions. These findings echo those of researchers working in other contexts, who increasingly find that economic and social factors play a strong role in fertility. We show that one reason for the new findings is some general statistical problems in the Princeton methodology. These are reason to temper acceptance of the Princeton project’s larger message. Keyword: fertility transition JEL codes: J13, N33, O15 The Princeton Project on the Decline of Fertility in Europe was a large-scale research project undertaken by the late Ansley Coale and his collaborators at Princeton’s Office of Population Research in the 1960s and 1970s. The project compiled measures defined at the level of administrative areas for most western European countries and used this data to study the patterns of fertility decline and its correlation with possible explanatory factors. This research has been extremely influential,because of the project’s scope and the skill and ingenuity of the individual studies. The project’s overall conclusion, often called the “Princeton view,” downplayed the importance of economic and social change in causing the fertility transition in Europe, and instead stressed a process of innovation and diffusion, driven by similar attitudes and communication networks. Two recent studies of Germany in the late nineteenth and early twentieth centuries come to quite different conclusions. Patrick Galloway, Eugene Hammel, and Ronald Lee (hereafter GHL) studied Prussia, the largest of the German states, while our own research (hereafter BG) focused on Bavaria, which was the next-largest state. Both projects find a clear role for the economic and social forces that the Princeton project (hereafter “EFP” ) downplayed. These new results for Germany build on a much earlier paper by Toni Richards, who used the data underlying John Knodel’s EFP monograph on Germany to come to conclusions at variance with his own. The difference in results is surprising because both of the more recent studies bear strong similarities to the EFP approach, while Richards’ study is based directly on the EFP data. This paper argues that much of the difference can be attributed to two problems in the statistical methods used by the Princeton authors. These problems are quite general and would affect studies other than those for Germany.1 Our argument suggests caution in accepting the EFP conclusions. But our critique should not be exaggerated; we cannot show that the other EFP studies were wrong, we can only demonstrate the problems in the German case and note that related problems might affect other studies. On a similar note, we want to stress at the outset that the weaknesses we identify here reflect in a real way the EFP’s strengths. The first problem we dis1 Galloway et al (1998b, pp.195-208) surveys the methods used in recent research on the fertility transition. A recent paper by Potter et al (2002) uses methods similar to those advocated here, and also comes to the conclusion that social and economic forces have not been given due weight in explaining the fertility transition. The context for that paper is modern Brazil.