I’m too clever for this job: a bivariate probit analysis on overeducation and job satisfaction in Australia

Abstract

Using data from the first wave of the Household, Income and Labour Dynamics in Australia data set, this article establishes an empirical relationship between overeducation and workplace satisfaction for Australian adult males in the labour force. In a departure from much of the existing literature, both univariate and bivariate probit models are used to account for potential unobserved heterogeneity. We find that estimates in the univariate probit models are positively biased for three of the six measures of workplace satisfaction studied. This suggests that consideration should be given to the use of bivariate models when studying the determinants of workplace satisfaction and overeducation. Results show, although levels of satisfaction remain high, that across all measures of workplace satisfaction overeducated workers are less satisfied compared to their nonovereducated counterparts. This intimates that satisfaction levels should be viewed from a relative, rather than an absolute perspective. ‘Pleasure in the job puts perfection in the work’ Aristotle 384BC–322BC

Acknowledgements

We would like to thank Jim Taylor, Rodney Beard and an anonymous referee for their helpful feedback on an earlier draft and the Melbourne Institute and the Department of Family and Community Services for providing us with the HILDA data set. Views expressed in this article do not necessarily reflect those of the Melbourne Institute nor the Department of Family and Community Services. All errors and omissions remain our own.

Notes

¹ Overeducation is said to occur if a worker has excess education to that needed to perform his or her job. See Hartog (2000) for more information.

² It is often assumed that education is positively related to productivity, at least from the view point of human capital theory (see e.g. Nelson and Phelps, 1966; Welch, 1970; Schultz, 1975).

³ Hersch hypothesizes that firms hire overeducated workers because they require less training in their starting positions and thus form a pool from which promotions are made. This reduces hiring costs for higher-level positions in the firm.

⁴ We are unable to replicate this distinction as the HILDA data set does not contain enough information to separate educational qualifications and skill levels.

⁵ For more information on Wave 1 of the HILDA data set see Wooden et al. (2002).

⁶ The undereducated are job mismatched insofar as they possess article qualifications beneath their job requirements. However, undereducation is not a concern for policymakers as the undereducated tend to be more experienced workers with longer tenure at work. As the undereducated are not deemed a disadvantaged group in the labour market, a number of other studies in the education–job mismatch literature have also concentrated their research only on the overeducated. See e.g. Duncan and Hoffman (1981), Alba-Ramirez (1993) and Battu and Sloane (2004). In unreported results, we find the undereducated to be older, have more working experience, and have longer tenure than both the over and adequately educated. In essence, we surmise that these workers have been promoted into their jobs, in the absence of appropriate article qualifications, because they have developed the necessary skills to do that job.

⁷ The workplace satisfaction question given to respondents (E36 of the Person Questionnaire) is reproduced as Appendix B.

⁸ We also investigate more stringent restrictions on the identification of satisfaction dummies in order to test the robustness of our original measure. Specifically, we narrow the definition of satisfaction to include only those who answered by giving a value of eight or higher. Although there are some minor changes in parameter estimates and levels of significance, our results remain largely unaltered.

⁹ A corresponding table comparing the employed and nonemployed males in the HILDA data set is presented as Appendix C. As expected, the employed posit different characteristics compared to the nonemployed. Our definition of overeducated precludes the inclusion of nonemployed males.

¹⁰ However, Battu et al. (2000) do not find that the overeducated are more likely to be valued by their employers. Our hypothesis in favour of Hersch (1995) and Buchel (2002) is strengthened by Kler's (2005b) finding that overeducated graduates in Australia are more likely to be satisfied with job security.

¹¹ Correcting for potential endogeneity can also be done using the instrumental variable (IV) approach. However, there are shortcomings related to this approach that lead to a large trade-off between bias and variance (Davidson and MacKinnon, 1993; Norton et al., 1998; and Bollen et al., 1995). Should weak correlation exist between the instruments and endogenously determined explanatory variables, IV estimates can be seriously biased and the bivariate probit model is more appropriate and efficient under these conditions (Greene, 1997, 1998).

¹² See Appendix A for definitions of these groups.

¹³ See Alba-Ramirez (1993) and Battu et al. (1999) for evidence.

¹⁴ Should the levels of job satisfaction be lower for jobs that require less education, then one possible explanation of overeducation being endogenous with respect to job satisfaction is that they are in jobs that do not fully utilize their skills. There is evidence that the overeducated are more dissatisfied with jobs that do not maximize their skills (Burris, 1983). As well, in unreported findings, we find that levels of satisfaction are higher for those in occupations with higher entry requirements compared to those with lower entry requirements.