Over-skilling and Over- education Peter J Sloane, Director, WELMERC, School of Business and Economics, Swansea University, IZA, Bonn and University of Melbourne Abstract There is now a substantial literature on the concept of over-education, but due to data availability a much smaller one of the concept of over-skilling. This paper compares and contrasts these two concepts. The policy relevance of over- education depends on the extent to which it represents a mismatch between workers’ levels or types of education and the requirements of the job. However, is there is substantial heterogeneity among individuals with particular levels of education, ‘over-educated’ workers may simply be those with lower ability levels given their level of education, so that there is no market failure. Using data on over-skilling from both Australia and Britain the paper argues first that over- skilling and over-education measure different things and second that the over-skilling measure is more likely to capture true mismatch than the over-education measure. CEDEFOP Research Arena Workshop on Skill Mismatch: Identifying Priorities for Future Research Thessaloniki, Greece, 30 May PRIFYSGOL ABERTAWE SWANSEA UNIVERSITY WELMERC
1. INTRODUCTION In Australia the university participation rate rose from 24% in 1988 to 38% in 1999, while in Britain, the participation rate rose from 13% in 1980 to 33% in This rapid increase has led to concerns about employer – employee mismatches (i.e. graduates in non-graduate jobs). Over-education rates are about 30% in both countries. Over-educated workers are paid more than matched co-workers, but less than matched individuals with the same qualifications as themselves. Does this represent individual heterogeneity or market failure? The paper examines: Whether overskilling is substantial in the two countries and has a similar pattern Whether overskilling is substantial in the two countries and has a similar pattern Whether there is a sizeable wage penalty in each country. Whether there is a sizeable wage penalty in each country.
2. OVERSKILLING AND OVEREDUCATION In both HILDA and WERS 2004 individuals report the extent to which they utilise their skills and abilities in the workplace This is less subject to bias as a consequence of individual heterogeneity than the over-education variable. The two variables measure different things. Green and McIntosh (2002) found that less than half over-educated were also overskilled. They found 20% of British workforce were overskilled and 4% under-skilled. Four possibilities Education and skill matching (- professional degrees)Education and skill matching (- professional degrees) Overeducation, but skill matching (- individual heterogeneity)Overeducation, but skill matching (- individual heterogeneity) Education matching, but overskilling (- grade inflation)Education matching, but overskilling (- grade inflation) Both over-education and overskilling (- constrained job search)Both over-education and overskilling (- constrained job search)
3. THE DATA HILDA is a panel of about 20,000 individuals, which has run from 2001 WERS is cross-section matched employer – employee data set, containing 2,295 establishments and up to 25 employees per establishment HILDA measures overskilling on a 7 point scale from 1 strongly disagree to 7 strongly agree on answers to the statement “I use many of my skills and abilities in my current job”
WERS measure is derived from the question “How will do the skills your personally have match the skills you need to do your current job? ” There is a five point scale defined as much higher, a bit higher, about the same, a bit lower, much lower. In HILDA 11.5% are severely overskilled (1, 2 or 3) and 30.6% moderately so (4 or 5). In WERS 21.1% are severely overskilled and 31.9% moderately so.
4. OVER-EDUCATION AND OVERSKILLING: A COMPARISON We assess the strength of the relationship between the two variables, using HILDA only and the empirical method to estimate overeducation Three measures: Definition 1- One education level above the modal level of education within the occupation Definition 2 - One standard deviation above the mean level of education Definition 3 - Half a standard deviation above the mean level. Whatever the definition 50% of those classified as over-educated were also overskilled, and of these 20% were severely overskilled and 30% moderately so. Table 3 The effect on wages: Model 1 -Overskilling alone Model 2 -Overeducation alone Model 3 –Combined Clearly the two measures are different. They both have a significant negative effect on earnings.
Table 3: The effects of overskilling and overeducation on wages - comparison of alternative overeducation definitions Note: Dependent variable is weekly wages. Standard errors in parentheses. ***/**/* denote significance at 1%, 5% and 10% respectively. Source: Hilda survey waves 4 and 5.
5. PATTERNS OF OVERSKILLING IN AUSTRALIA AND BRITAIN Incidence is measured for full-time workers only, using weekly earnings and correcting for hours worked, for different levels of education and using comparable explanatory variables in the two data-sets. Table 4 suggests i.Overskilling is more prevalent in Britain than in Australia ii.In Australia it declines with educational level, whilst in Britain it is invariant to educational level.
Table 4: Overskilling by education Note: Full-time employees only. Source: Hilda and WERS 2004.
6. ESTIMATION We estimate a standard wage regression in which log of weekly wages is regressed on a vector of characteristics for individual i in workplace j: Where includes a vector of individual characteristics such as gender, marital status, age,tenure and educational attainment Where includes a vector of individual characteristics such as gender, marital status, age,tenure and educational attainment includes a vector of employment characteristics such as size of establishment and industry includes a vector of employment characteristics such as size of establishment and industry is a dummy for severe overskilling is a dummy for severe overskilling is a dummy for moderate overskilling is a dummy for moderate overskilling denotes estimated returns to the characteristics vector denotes estimated returns to the characteristics vector is standard error term. is standard error term. Table 8 reveals wage costs for severely overskilled of 8.5% in Australia and 12% in Britain, with penalties for the moderately overskilled of 2.3% in Australia and 2.9% in Britain. Returns to obtaining a degree relative to no qualifications are, however, higher in Britain (56% compared to 42%). Table 9 reports the effects of over-skilling on earnings by educational level. These seem to increase with educational level in both countries. Table 10 shows that the effects tend to be stronger for men than for women.
Table 8: OLS and interval regression estimates for effects of overskilling on weekly wages - Australia vs. Britain Note: Standard errors in parentheses. Reference groups are as follows: age ; education attainment below yr 10; employed with current employer for less than a year; employed on continuing contract with a firm that employs at least 50 people. ***/**/* denote significance at 1%, 5% and 10% respectively. NA denotes that R square statistics are not available for interval regression equations.
Table 9: Effects of overskilling on weekly earnings by education level Note: Standard errors in parentheses. OLS regression results for Australia and interval regression results for Britain, with weekly wage as the dependent variable. A large number of covariates has been included and is reported in Appendix Tables A5a, A5b and A5c. ***/**/* denote significance at 1%, 5% and 10% respectively. NA denotes that R square statistics are not available for interval regression equations.
Table 10: Effects of overskilling on weekly earnings by gender Note: ***/**/* denote significance at 1%, 5% and 10% respectively. NA denotes that R square statistics are not available for interval regression equations.
7. CONCLUSIONS Both countries have a problem of overskilling, but the effects are greater in Britain than in Australia. In Australia incidence falls with educational level, whilst it is invariant to education level in Britain. In both countries the wage penalty increases with education. In the long run any benefits to employing overskilled workers are likely to be eroded by lower job satisfaction and a higher propensity to quit. There are likely to be costs to the economy in terms of lost output.