• Purchase/rental options available:
In lieu of an abstract, here is a brief excerpt of the content:

• Author's response:

I write to offer some clarification regarding Andrew Hinde's review of my book, Birth, Death and Religious Faith in an English Dissenting Community, which appeared in a recent issue of your journal.

First, with regard to Dr. Hinde's objection to my use of event history analysis as a hazard model, I described event-history analysis in the following terms: "The event affecting the life of an individual, or the transition between events, normally serves [in event history analysis] as the unit of analysis. Event history analysis assumes the form of logistic regression in which the probability of a person being at risk of a particular event is determined by a set of explanatory variables. These models are usually termed Hazard models, since the Hazard rate (the probability of individuals being at risk) serves as the unobserved dependent variable; they are logistic because the Hazard rate, as a probability, varies between zero and one." Supporting this assertion, I cite Paul D. Allison, Event History Analysis: Regression for Longitudinal Event Data (Beverley Hills, 1984). I go on to observe, citing historical demographer, George Alter, that it is possible "'to think of Hazard Models as multivariate models of life table processes." 'The results of the logistical regression are slightly more precise than Dr. Hinde describes; they show that the probability of a birth occurring rises significantly following the conversion experience.

Secondly, This finding is explored in greater depth in the chapters on Baptist and Anglican fertility using a technique called "path analysis," which is another form of regression analysis, in which the direct and indirect causal flows emanating from explanatory variables are broken down into their constituent parts. This technique has the virtue of displaying the interplay of explanatory variables with each other as they combine to affect the dependent variable. So in one example I write, "With regard to family size (D[ependent] V[ariable] 'Kidcount'), age at marriage and age at conversion show parity in their [End Page 417] respective direct effects, which are each robust, in terms of the size of coefficients and probability distributions. This remains true of the indirect effect of marriage age, although by the course of mediation by age at conversion, its coefficient size becomes reduced in its total effect, thereby rendering conversion age the dominant variable in the model" (69-70). Dr. Hinde declines to mention my use of path analysis.

The differences between Anglican and Baptist in these models are indeed more ones of degree, yet they are nuanced and deserve comment. A gender difference emerges within each denomination, which bears emphasizing, and there is further an absence of class difference in the analysis of fecund-fertility rates, which bears mentioning.

Third, Dr. Hinde suggests that there is little difference between the fertility of Baptists and Anglicans, "save possibly that late conversions among Baptists gave fertility a slight boost at older ages." Yet a comparison of fecund fertility rates shows a marked difference among selected age-cohorts (fig. 5.7, 72 and appendix 3). He also states that my sample sizes are too small. But there are 499 Anglican families and 97 Baptist families, which is not mean by the standard of early-modern family reconstitution studies and especially in an environment of high geographical mobility induced by industrial revolution, which bears mentioning, and which one hardly finds in other family reconstitution studies; these tend to focus instead on more stable populations. In the path analyses, the data sets were reconfigured into a long form, so that size of the Ns is higher. In fig. 4.7 (49), for example, the N for the pre-marital subset is 151 based on 26 family clusters, while in the post-marital subset, the N equals 355 based on 69 family clusters. The Ns among Anglicans, in the post-marital subset especially, are much higher, e.g., 2424 based on 462 family clusters.

Dr. Hinde may find my presentation dry, but that is in the nature of the material. It is no more so than, say, EA Wrigley's, population history from family reconstitution, and perhaps even less so, as it does not dwell on methodological technicalities.

The main...