Labour market segmentation and union wage gaps.

Abstract There has been a great deal of research regard the effects of unions on union--non-union wage gap. Most of the studies regarding the impact of unions on wages have assumed that apart from the division between union and non-union workers, the labour market is relatively homogeneous. A number of economists, however, have argued that the labour market is segmented, implying that there are distinct labour markets and that some workers employment opportunities are concentrated in "bad jobs" while other workers employment opportunities are concentrated in "good jobs" which are rationed.

This paper will explore whether the relative wage differential between union and non-union workers differs between the independent primary, subordinate primary and secondary labour markets. Labour market segments are defined using "job zones". "Job zones" are distinct groups defined by the level of specific vocational preparation necessary for a particular occupation, allowing for the comparison of skill levels and training for each occupation. The data on "job zones" comes from the Occupational Information Network database (O*Net). We estimate separate equations for union and non-union workers in each segment using data from the Current Population Survey and calculate union non-union differentials for each labour market segment. The findings of this paper suggest that the greatest differentials are in secondary labour markets followed by differentials in the subordinate primary labour market and that the smallest wage differentials are in the independent primary labour market.

Keywords: wages, unions, segmented labor markets

INTRODUCTION

There have been numerous studies of the effect of unions on the wages of union and non-union workers. Most of these studies concentrate on the relative impact of unions on the union-non-union differential (Parsley 1980; Freeman and Medoff 1984; Lewis 1986; Hirsch and Addison 1986; Blanchflower and Bryson 2004; Hirsch 2004). Most of the studies regarding the impact of unions on wages have assumed that apart from the division between union and non-union workers, the labour market is relatively homogeneous. A number of economists, however, have argued that the labour market is segmented, implying that there are distinct labour markets and that some workers employment opportunities are concentrated in "bad jobs" while other workers' employment opportunities are concentrated in "good jobs" which are rationed. In recent years, a number of empirical studies that have found strong support for the segmented labour market model (Dickens and Lang 1985; Osberg et al. 1987; Boston 1990; Fichtenbaum et al. 1994; Orr 1997; Kalleberg 2003; Reid and Rubin 2003; Davia and Hernanz 2004). There has also been important research along similar lines dealing with wage differentials between contingent and full-time workers (Belman and Golden 2000). One important question that has not been adequately addressed in the segmented labour market literature is whether unions affect wages differentially between sectors. This paper attempts to fill this gap in the literature. Specifically, this paper will explore whether the relative wage differential between union and non-union workers differs between primary and secondary labour markets.

The next section will discuss the problems of defining labour market segments and the data used in this paper. The third section will present a number of models to estimate the union--non-union wage gap in both the primary and secondary sectors. The models will then be estimated and the results compared with previous findings in the literature. Finally, the last section will present a summary and conclusion.

DEFINING LABOUR MARKET SEGMENTS

A central concern in defining labour market segments has been the issue of truncation bias. One of the central predictions of segmented labour market theory is that returns to education and work experience will be smaller or non-existent is secondary labour markets. However, Cain (1976) has argued that studies, which use income to divide labour markets into segments, may suffer from truncation bias. This occurs when markets are not truly segmented and are divided between low and high wage earners. When this occurs, the rates of return to education and work experience are biased downwards compared to a sample that is undivided. However, Osterman (1977), Rumberger and Carnoy (1980) and Ryan (1981) have argued that Cain's argument is problematic. In fact, if segmentation does exist then estimates of rates of return to education and work experience in a pooled sample will be biased because they omit the segmentation variable. That is, some people with higher levels of education will have low earnings if they are stuck in a secondary labour market, and other people with the same level of education will have high earnings if they work in primary labour markets.

The critical issue in avoiding truncation bias is using selection criteria for defining labour markets that are independent of wages. Since the major objective of this study is to examine whether or not there are differences in union wage premiums in segmented labour markets we will use a number of different methods for defining labour market segments to ensure that the results are not an artifact of one particular method for defining labour market segments.

The labour market segmentation literature developed initially from the work of Doeringer and Piore (1971) on dual labour markets. Doeringer and Piore used the concept of internal labour markets to define primary and secondary labour markets. Primary markets were characterized by high wages and jobs with fringe benefits, stable employment with chances for promotion, due process in work rules and procedures for settling grievances. Secondary markets were characterized by low pay, absence of benefits, poor working conditions, high turnover and few chances for advancement.

One of the major points made by segmented or dual labour market theorists was that earnings did not depend solely on supply-side characteristics but were also a function of demand-side (institutional) characteristics. This was particularly important in its implications for anti-poverty programmes. Human capital theory holds that the best way to solve the problem of low wages and poverty is to increase training and education. In contrast, the view presented by segmented labour market theorists was, that while training and education were useful, they would not significantly reduce poverty and low wages because wages reflected the characteristics of particular types of jobs rather than the characteristics of individuals. For segmented labour market theorists, human capital theory may explain a significant amount of variation in earnings in the primary sector. However, in the secondary sector they assert that wages are largely independent of human capital. Thus, the segmented labour market model was formulated to challenge the "orthodox", i.e. human capital, model.

There are a number of different approaches to defining labour market segments and various studies have divided the market into two or more distinct segments. Segments can be defined using firm/industry characteristics, occupations, job characteristics, individual characteristics or some combination of all of these factors (Bibb and Form 1977; Hodson 1977; Beck, Horan and Tolbert 1978; Oster 1979; Rumberger and Carnoy 1980; Gordon 1986; Boston 1990).

This paper defines labour market segments using "job zones". Job zones are distinct groups defined by the level of specific vocational preparation necessary for a particular occupation, allowing for the comparison of skill levels and training for each occupation. The data on job zones comes from the Occupational Information Network database (O*Net[TM]98). The O*Net[TM]98 database was designed by the US Department of Labor's Employment and Training Administration to take the place of the Dictionary of Occupational Titles. Each job zone is defined by a range of Specific Vocational Preparation (SVP) scores.

For purposes of this paper, each labour market segment was defined by assigning each three-digit occupation, in the March Current Population Survey (CPS), to a job zone. The secondary labour market consists of all occupations with a job zone 1, which has an SVP score less than 4.0. These jobs require little or no preparation and most training is done on the job usually by a fellow worker in a relatively short period of time. These occupations involve helping people and following instructions. The lower tier of the primary labour market, also referred to as the subordinate primary labour market, consists of occupations in job zones 2 and 3, which have SVP ratings greater than or equal to 4.0 and less than 7.0. These jobs require some degree of preparation which might involve an associate's degree, an apprenticeship or some type of formal vocational education. Finally, the upper tier of the primary labour market, also referred to as the independent primary labour market, consists of occupations in job zones 4 and 5 with SVP ratings of 7 or greater. These jobs require a considerable amount of formal education as well as a significant amount of on the job training. In most cases at least a college education is required and in many cases an advanced degree is required. This method of defining labour market segments is similar to the methodology used by Rumberger and Carnory (1980) in defining labour market segments. (1)

Table 1 shows the breakdown of the various labour market segments using March CPS data from 1992-2002 by race and sex. Not surprisingly, both black men and women are more likely than their white counterparts to be in the secondary labour market and less likely to be in the independent primary labour market. Compared to others in the sample Hispanics are also more likely to have jobs in secondary labour markets and this is especially the case for Hispanic women. Compared to whites, there is only a slightly higher proportion of Asian men in secondary labour markets and, while the proportion of Asian women in secondary labour markets is larger, it is relatively smaller than the proportion of black or Hispanic women in secondary labour markets.

There are few differences in the proportion of white, black and Hispanic men working in the subordinate primary labour market. This same pattern generally holds for white, black and Hispanic women. Interestingly, there is a smaller percentage of Asians employed in the subordinate primary market, among both men and women.

Table 1 also shows that larger percentages of whites and Asian men are employed in the independent primary market compared to blacks and Hispanics.

Among whites and Asians about the same percentage of men and women are employed in secondary labour markets. However, the data indicate that black and Hispanic males are actually more likely to hold jobs in the secondary labour market than black and Hispanic women. Women of every race are more likely to be employed in the subordinate primary market than their male counterparts. Finally, white and Asian men are more likely than their female counterparts to be employed in the independent primary sector. By contrast, there are relatively small differences between the genders among blacks and Hispanics in the independent primary labour market.

This distribution of jobs across segments is consistent with the findings of Boston (1988) who found that race was a significant factor in determining labour market segmentation. In the case of gender, however, segmentation may not be as important a factor in explaining male-female differentials as occupational sex segregation.

Table 2 shows the average wage rate for each of the labour market segments. As expected, workers in the independent primary sector have the highest average wages followed by workers in the subordinate primary market. Not surprisingly, wages for workers in the secondary sector are …