Leaving Science: Occupational Exit from Scientific Careers (review)

Concern over science and engineering attrition is not new, but most research has focused on early interest in science, choice of a science major, and the subsequent decision to persist or to switch fields. Preston has focused on a more difficult population to study longitudinally, namely, career scientists.

The introduction and first chapter present published statistics on attrition from the science and engineering pipeline, interpreting those statistics within the cultural and economic context of science in the United States. In Chapter 1, Preston describes the sources and limitations of each of her three research databases: (1) the SSE national survey data collected by the NSF from 1982 to 1989, (2) a "university database" consisting of work history surveys sent to samples of science and engineering alumni from one large public university in the northeast, and (3) interview data from a subsample of the university sample.

Chapter 2 addresses the magnitude of the exit from science and engineering careers. Preston uses univariate and multivariate analyses of the national and university data to determine workplace or worker characteristics that increase or decrease the probability of exit. Important gender differences are noted and discussed. From the interview data, for each gender, Preston compiles four general categories of reasons respondents give for leaving science and engineering. In chapters 3 through 6, she elaborates on those reasons, drawing on her statistical findings, interviews, and references to the literature, especially in economics.

Chapter 3 focuses on leaving science for income and opportunity, the issue on which males and females differ the most dramatically. To attempt to explain the relatively large number of men leaving science in search of higher salaries, Preston applies human capital theory as well as alternative theories of job and occupational mobility. She notes from the literature those variables that influence occupational choice and exit, and links the literature with her data. In this chapter she contrasts salary profiles of science graduates who choose science careers with those who choose other careers. Finally, this chapter compares the salary profiles of two types of science leavers: income seekers and amenity seekers.

Chapter 4 studies the effects of family responsibilities on a scientific career, contrasting males and females. It examines economic explanations, results of previous research, and results derived from the three databases.

Mentoring is identified in Chapter 5 as a factor having an important impact on a woman's probability of continuing in science. Preston reviews the relevant literature on mentoring, by gender. For her interview sample, she computes the probability of finishing a graduate program and the probability of successful employment in science as a function of mentoring.

Chapter 6 addresses discontent with science itself as a major reason for exiting science careers, especially for women. It compares experiences that attract students to science with experiences that drive scientists out of their jobs, and it shows that there are sharp contrasts between men and women.

One defining characteristic of the field of science is rapid change in its body of knowledge. Chapter 7 explores the effects of skill depreciation and the stress induced by having to keep up with new discoveries on the exit from science occupations. Especially interesting is an analysis of 22 fields of science ranked according to their rates of knowledge growth, and the relationship of rates of growth to frequency of occupational choice and subsequent career attrition.

Chapter 8 examines the evidence for gender discrimination and its possible relationship to exit from science.

Finally, in Chapter 9, Preston uses the four major reasons for occupational exit identified in the previous chapters to develop thoughtful policy recommendations.

Critical readers will have some problems with the book. There is not enough information about the samples. Sample sizes should have been revealed in all of the tables and graphs as well as in the technical appendix. It would have been useful if, in the first chapter, we had seen tables summarizing each of the three databases, showing cross-tabulations by gender of the number of sample members in each...