Analysis of Part-Time Pay in Canada

Abstract

This paper analyses the difference between part-time and full-time wages in Canada using the 1981 Survey of Work History. Since part-time and full-time workers are distinguished by hours worked and since hours worked and wages are interdependent, selectivity bias occurs if ordinary least squares regression is used. A modification of Heckman's procedure, to account for the fact that part-time hours are doubly truncated, is used to correct this problem. The results indicate that, adjusting for differences in characteristics, the wage differential between part-time and full-time workers is 10 per cent, or one-third the unadjusted differential. Results vary somewhat by sex and marital status. Analyse de la remuneration du travail d temps partiel au Canada. Ce memoire analyse les differences entre les salaires pour le travail a temps plein et a temps partiel au Canada a partir des resultats de l'Enquete sur l'activite en 1981. Puisque les travailleurs a temps plein et A temps partiel sont definis par le nombre d'heures de travail, et puisque le nombre d'heures de travail et les niveaux de salaire sont interdependants, il y a biais de selectivite quand on utilise la methode des moindres carres ordinaires dans l'analyse de regression. L'auteur utilise une modification de la technique de Heckman (pour tenir compte du fait que le nombre d'heures de travail a temps partiel est doublement tronque) pour corriger le probleme. Les resultats montrent que, apres avoir tenu compte des differences dans les caracteristiques des personnes, on observe un differentiel de niveau de salaire de 10 pourcent entre les travailleurs a temps plein et a temps partiel, soit le tiers du differentiel brut avant les ajustements pour tenir compte de caracteristiques differentes. Ces resultats varient quelque peu selon le sexe et selon la situation familiale. The trend towards part-time employment has generally escaped the attention and scrutiny of economists.' Part-time employment in Canada grew at an annual rate of 4.1 per cent between January 1972 and January 1984, compared 1 One exception is Nakamura and Nakamura (1983), who examine related labour supply issues for married females. Canadian Journal of Economics Revue canadienne d'Economique, xix, No. 4 November novembre 1986. Printed in Canada Imprim6 au Canada 0008-4085 / 86 / 798-807 $1.50 ?0 Canadian Economics Association This content downloaded from 207.46.13.159 on Sat, 22 Oct 2016 04:29:38 UTC All use subject to http://about.jstor.org/terms Analysis of part-time pay in Canada 799 with a growth rate of only 2.4 per cent for full-time employment. As a result 15.4 per cent of jobs were part-time as of January 1984 compared with 13 per cent in January 1972.2 The growth of part-time employment raises a number of issues. In particular, Labour Canada (1983) argues that the hourly wage differential between full-time and part-time workers reached 26.3 per cent in 1981. It is not clear, however, that this gross wage differential between full-time and part-time workers is a useful figure. A number of accepted, measurable characteristics that determine wage differentials among workers may confound this comparison unless full-time and part-time workers have identical characteristics. It would therefore be useful from a policy standpoint to eliminate these effects to identify part-time effects, and the gravity of the part-time pay problem, per se. This paper analyses the differences between part-time and full-time wages to try to distinguish part-time effects, whatever the source, from other established bases for differentials. In the first section a standard model for wage comparison among workers is considered. The source of selectivity bias is identified and a consistent estimation technique is described. The second section uses this technique to analyse the differences in part-time and full-time wages using microdata from the 1981 Survey of Work History used by Labour Canada (1983). The third section summarizes the results and provides suggestions for further research. A MODEL FOR WAGE COMPARISON WITH SELECTIVITY BIAS Consider a standard model used extensively for union and non-union wage comparisons (Lee, 1978; Duncan and Leigh, 1980; Robinson and Tomes, 1984; Simpson, 1985). Let Wf be a vector of observations on the natural logarithm of hourly wages paid to full-time workers and let X be a matrix of observations on measurable worker characteristics commonly used in such models, namely, educational attainment, age, sex, union status, public (vs. private) sector status, and region. Let wp and X be the corresponding observations for part-time workers. Let h* be the maximum hours of work permitted for classification as a part-time worker. Then we specify Wf Xff3f ? If if h > h* (1)