The incidence of educational mismatch in Canada

220 S.P. Vahey Economics of Education Review 19 2000 219–227 requirements as “otherwise identical”. Assuming earn- ings reflect marginal productivity, their results do not support Berg’s 1970 hypothesis: overeducated workers earn more—not less—than otherwise identical workers. A number of studies have focused on whether the returns to skill mismatch are gender dependent. Frank 1978 argued that the limited geographic mobility of women causes a male–female differential in the returns to education. If relocation is a family-based decision, then it will be based upon the needs of the primary earner—usually the male. As a result, the secondary earner—the female—is geographically constrained in her job search. Duncan and Hoffman 1981, Rumburger 1987, Hartog and Oosterbeek 1988, Groot 1996 and Kiker et al. 1997 have examined male–female differ- ences in the returns to skill mismatch. These researchers found that their results were similar for the two sexes: under overeducated workers earn less more than otherwise identical workers. Using US data, however, Hersch 1991 found that for females, the returns to edu- cational mismatch were insignificant. In this study, I use data from the National Survey of Class Structure and Labour Process in Canada NSCS to estimate the returns to educational mismatch. These are the only data available that contain self-report infor- mation on educational mismatch in Canada for either sex. For males, I find evidence of negative positive returns to under overeducation; but also find that the returns are sensitive to the level of required education. The results for the female sub-sample differ from those obtained using the full sample: the returns are insignifi- cant for all levels of required education. Hence, Berg’s 1970 hypothesis is rejected for both males and females: overeducated workers do not receive lower earnings than otherwise identical workers. The rest of the paper is organised as follows. In the following section, I discuss the incidence of educational mismatch in Canada. I then set out the empirical model and present the results. I draw some conclusions in the final section.

2. The incidence of educational mismatch in Canada

The data are taken from the NSCS, a cross-sectional survey that contains information on approximately 3000 respondents. This survey was carried out by Canada Facts, who conducted face to face interviews in 1982. Researchers at the Department of Sociology and Anthro- pology at Carleton University decoded the survey responses and transferred the information to tape. I have excluded workers over 64 and under 18, and anyone with non-positive 1981 earnings or hours worked per week. After removing the self-employed and those who did not work year round, the final sample is 993, of which 424 are female. 2 The unique feature of this Canadian survey is that respondents were asked about both their attained edu- cation and the education requirements for the job. The following question was asked about educational attain- ments: “What is the highest level of education you have completed?”. The answers were categorised into six classes: grade school diploma or less GRADE, some high school SOME, completed high school HIGH, collegevocational school COLL, bachelor’s degree BACH, and postgraduate or professional degree POST. The question asked about required education was: “What type of formal schooling is now normally required for people who do your type of work?” Individ- uals are defined as under overeducated if their attained schooling is less greater than their required education. Since the second question inquired about schooling “now normally required”, arguably the resulting variable understates overstates the extent of overeducation undereducation–education requirements have generally increased with time. 3 A drawback of this type of measure of skill mismatch is that it is by definition subjective. Workers who are dissatisfied with their jobs may misreport themselves as overqualified—introducing bias into the model. The main advantage of the self-report approach is that the measure is job specific. Requirements can differ greatly within occupations. For example, the schooling require- ments for a post as an economist can vary from an under- graduate degree for some private sector jobs, to a Ph.D. for research jobs. Other researchers have used occu- pation-based measures derived from either expert opi- nion of educational requirements e.g. Alba-Ramı´rez, 1993 andor deviations from average attainments Verdugo and Verdugo, 1989. Kiker et al. 1997 and McGoldrick and Robst 1996 review the advantages and disadvantages of each measure in detail. 4 2 I have excluded seasonal workers on the grounds that their earnings are influenced by very different factors—related to their labour market inflows and outflows. Mismatched workers could self-select into seasonal work, however, particularly if they are not rewarded by the market for their qualifications. Hence, the exclusion of workers who did not work year round could cause downward bias in the incidence of and the returns to educational mismatch. 3 In contrast, the question asked of the Panel Study of Income Dynamics respondents refers to schooling required “to get a job like yours?”. The question asked in Hersch’s 1991 survey in Eugene, Oregon enquired about schooling “needed to do a job like yours, not just be hired”. In both these cases the time frame is unclear. 4 In most studies utilising self-report evidence, the conven- tional measure of educational mismatch is denominated in years of schooling, rather than levels of achievement as in this study and Hartog 1986. In some cases, such as Sicherman 1991, 221 S.P. Vahey Economics of Education Review 19 2000 219–227 Table 1 Incidence of skill mismatch Attained Required GRADE SOME HIGH COLL BACH POST TOTAL Males n 5 569 GRADE 6.0 2.6 2.8 0.5 0.0 0.0 12.0 SOME 3.3 5.6 6.9 2.3 0.5 0.0 18.6 HIGH 1.4 4.0 8.3 1.8 1.9 0.7 18.1 COLL 2.1 2.6 9.8 12.7 2.3 0.7 30.2 BACH 0.0 0.4 0.7 1.8 9.0 0.9 12.7 POST 0.0 0.0 0.4 0.0 3.2 4.9 8.4 Total 12.8 15.3 28.8 19.0 16.9 7.2 100.0 Females n 5 424 GRADE 4.7 1.4 0.5 0.0 0.0 0.0 6.6 SOME 5.4 6.1 5.4 1.4 0.5 0.0 18.9 HIGH 1.9 1.7 12.7 2.1 0.7 0.0 19.1 COLL 0.5 2.6 13.7 16.7 3.8 0.5 37.7 BACH 0.0 0.7 1.7 1.9 9.4 0.2 13.9 POST 0.0 0.0 0.7 0.2 1.4 1.4 3.8 Total 12.5 12.5 34.7 22.4 15.8 2.1 100.0 The incidence of educational mismatch in the sample is described in Table 1. There are a number of striking features about these data. First, educational mismatch is a common phenomenon; but, the incidence of overeduc- ation males 30, females 32 is greater than the inci- dence of undereducation males 24, females 17. These figures are similar to those based on self-report measures for the U.S. McGoldrick and Robst, 1996 and Britain Sloane et al., 1996. Second, attained schooling is generally within one education level of required schooling; the incidence of skill mismatch outside this interval is small. Third, for both sexes, the peak in required schooling is at the HIGH level, but the peak in attained education is at COLL. Fourth, the distributions of attained and required education are flatter for males; the job market is particularly thin for females in the upper tail. As a result, the estimates for well-educated females should be interpreted with some caution. A number of researchers e.g. Sicherman, 1991; Groot, 1996 have noted that overeducation may be a short-run phenomenon. Entry level employees are often overqualified for their jobs, but go on to use their skills in later life. For the NSCS sample, overeducation is asso- the respondents were asked about the number of years in edu- cation. In others, for example Hersch 1991, the researchers converted the levels to years by making assumptions about the average equivalence between the two measures. The main advantage of using years is that the education variables are con- tinuous; the drawback is that, where a conversion factor is required, it is not job specific. ciated predominately with younger workers. Approxi- mately 57 of males and 33 of females under 26 are overeducated. 5

3. Empirical model