The Measurement of Segregation in the Labor Force

Citation

Chew, R. (2002), "The Measurement of Segregation in the Labor Force", International Journal of Manpower, Vol. 23 No. 1, pp. 105-108. https://doi.org/10.1108/ijm.2002.23.1.105.4

This book is a well‐written survey of the literature on the gender and occupational segregation of the labor force. Chapter One gives an introduction to the subject. It emphasizes the fact that gender segregation arises because women are more likely to be inclined towards specific occupations and industries, or because discrimination exists even before they enter the labor market. It also emphasizes the parallelism that exists between the measurement of income inequality and of labor force segregation.

Chapter Two provides a clear and concise presentation of various methods of classifying occupations and industries as “male”, “female” or “mixed” based on various ways of defining the degree of feminization of an occupation or industry. The first is the Oppenheimer approach, which is based on the percentage of female workers employed. The second involves classifying industries or occupations by comparing their gender ratio with the overall gender ratio in the workforce. This approach is an improvement over the Oppenheimer approach, as it takes into consideration the fact that the proportion of females in a given occupation or industry is likely to be determined by the overall proportion of females in the workforce. The third is Blackburn et al.’s (1993) marginal matching approach, which is based on the ranking of occupations or industries by decreasing gender ratio, drawing the line at the ratio where the total number of workers (both male and female) in the occupations (industries) at and above that ratio equals the total number of female workers in the labor force, and classifying as “female” those occupations (industries) which fall into this group of occupations (industries). The fourth is the bootstrap approach, which states that an occupation (industry) is “female‐intensive” when its gender ratio is greater than the overall gender ratio, and as “male‐intensive” otherwise. It stops short of identifying the advantages and disadvantages of each of the methods.

Chapter Three discusses occupational (industrial) segregation by gender. It identifies two methods of analysis. The first method involves the use of the relative dispersion of the gender ratios across the different occupations (industries), of which the Duncan and Duncan (1955) index of dissimilarity is the most popular index of occupational segregation. The second method is in terms of the absence of an overlap between the distribution of men and women among the different occupations (industries), which involves the use of the concept of “Transvariazione”. In discussing this method, the authors show the linkage between the Duncans’ dissimilarity index and the Gini‐segregation index, both of which are weighted measures of the dispersion of the gender ratios.

Chapter Four illustrates clearly the linkages between various sets of indices of occupational (industrial) segregation in the literature. One set of studies is based on intuitive interpretations of the Duncans’ dissimilarity index. All the studies in this set (Moir and Selby Smith, 1979; Lewis, 1982; and Karmel and Maclachlan, 1988) are based on the principle adopted by Cortese et al. (1976) in their extension of the dissimilarity index, with the exception of Zoloth (1976). The other set of studies is based on the Gini‐segregation index. This set includes Silber’s (1989) interpretation of Berrebi and Silber’s (1987) G‐matrix and Lieberson’s (1975) index of net difference. A third set stresses the degree of feminization and masculinization of the various occupations and industries. The first approach in this series is the Hakim (1981) SR index, an alternative version (SR’) of which is proposed by Siltanen (1990). Blackburn et al.’s (1993) marginal matching (MM) approach is parallel to SR’; the only difference between MM and SR’ is that SR’ is based on the traditional dividing line determined by the overall gender ratio. A fourth set is related to the concept of entropy. Included in this set are Fuchs’ (1975) index and Hutchens’ (1991) index, which measures the inequality of the gender ratios. A related index is Charles’ (1995) Association index. Finally, a new class of segregation indices is proposed by Kakwani (1994).

The contribution of Chapter Five, which lists the various properties that are considered desirable of segregation indices, is identifying those indices that satisfy each respective property. The authors state that there is no reason to assume that a measure of segregation should be occupational composition invariant, since the occupational structure determines the shape of the distribution of the gender ratios from which a measure of occupational segregation by gender is derived. The authors next look at Chakravarty and Silber’s (1994) axiomatically derived index. They, however, note that it is far from ideal because it does not satisfy all the characteristics regarded as desirable of a segregation index.

Chapter Six defines the concept of a segregation curve, as presented by Duncan and Duncan (1955). The authors illustrate clearly the parallelism between the Duncans’ index and the Pietra index for which a similar result holds with the Lorenz curve. They also prove that GS is equal to twice the area lying between the segregation curve and the diagonal, a result similar to that linking the Gini index and the Lorenz curve in the income inequality literature. They then go on to show that the segregation curve, which is an application of the Lorenz curve, allows an ordinal rather than a cardinal comparison of segregation across countries or over time. This may be done using an axiomatic approach, as taken by Hutchens (1991). However, if the segregation curves cross, one cannot say whether one distribution is more equal than the other. The authors then go on to apply Basmann et al.’s (1991) idea of parametrization to the segregation curve.

The authors next present the various functional forms which have appeared in the income inequality literature and have been applied by Deutsch et al. (1994) to the study of occupational segregation by gender: the Pareto distribution, Kakwani and Podder’s (1973) general functional form and Kakwani and Podder’s (1976) alternative functional form. They then go on to compare the results obtained on the basis of these equations with those derived from a non‐parametric approach. They find that Kakwani and Podder’s alternative functional form gives the best fit to the data. This is also true of the estimation of the index GS derived from the various functional forms.

Chapter Seven is devoted to proving that it is possible to decompose some of the segregation by population subgroups, as in the two approaches to the decomposition of income inequality: entropy indices that measure inequality between population subgroups and within each subgroup (see Shorrocks, 1984), and indices that include in addition the degree of overlap between the distributions of incomes corresponding to the various subgroups. Their empirical results indicate that studying occupational segregation at higher levels of aggregation causes one to ignore not only the segregation within each occupational group but also the overlap between groups. The authors use the Gini segregation index to show that such a decomposition allows one to decide whether segregation by gender in the labor force is stronger with respect of occupations or to industries.

Chapter Eight analyzes changes over time in the level of occupation (industrial) segregation. It decomposes the change over time into two components, one reflecting the change in the importance of occupations (industries), referred to as the mix effect, and the other reflecting the change in the gender ratio within each occupation (industry), referred to as the composition effect. This can be applied to both the Duncan and the Gini segregation indices.

Chapter Nine extends the analysis to the case where a multi‐dimensional classification of the labor force is given. Furthermore, these indices may be further decomposed to indicate the occupational mix and gender composition mix. It notes that the Karmel and Maclachlan (1988) technique is most useful for analyzing the components of a change in overall segregation.

Chapter Ten incorporates the analysis of occupational (industrial) segregation into that of wage differentials by gender by taking Oaxaca’s (1973) equation for income differential by gender, correcting for selectivity bias by applying Heckman’s (1979) procedure, using the bootstrap technique to identify the male and female and mixed occupations, estimating separate earnings functions for the male and female occupations and suggesting a method to decompose this wage gap into three components, a first one reflecting differences in human capital elements, a second one measuring wage discrimination in the labour market and a third one which is a consequence of occupational (industrial) segregation by gender.

This presentation in this book is thorough, comprehensive and concise. It provides all the information necessary for an in‐depth study of labor segregation. It is a must for the graduate student, and useful not only to the economist, but to the sociologist as well.