An Econometric Analysis of the 'Backward-Bending' Labor Supply of Canadian Women
IN THIS ARTICLE
This study investigates the behavior of the labor supply of Canadian women at different composite hourly wages of all paid jobs in 2009. Since it is widely demonstrated in the literature that variables such as age and levels of education, as well as the demographic, social, and financial characteristics of the household influence a woman’s decision to join the labor market, such variables are included in the study.
Although there have been a few studies on the women’s labor supply in both developed and developing countries, there are no recent studies as of 2009 that focus on the “backward-bending” labor supply of Canadian women. While some studies have focused on labor supply and its elasticities by considering the aspect of poverty in the labor supply analysis (El-Hamidi 2003, Sharif 1991, Dasgupta & Goldar 2005), and others have focused on testing the hypotheses advanced by the Nakamuras (1981) of finding a backward-bending supply curve for females similar to that of males (Robinson and Tomes 1985), this study tests the hypothesis that a backward-bending labor supply exists for Canadian women by using cross-sectional data from 2009. The results of this study offer strong support to the conclusions reached by the Nakamuras (1981), and Robinson and Tomes (1985).
The second motivating issue is of gender discrimination affecting Canadian women. Because of gender discrimination, women who perform the same tasks as men are often paid less and receive fewer benefits from their work. Even in developed countries like Canada, women earn only 70.4% of what men earn – a percentage lower today than in the 1990s [Gender Discrimination in Canada]. By understanding female participation behavior, policy makers will be in a position to assess the likelihood of tackling this issue, and provide effective policy prescriptions.
The analysis in this paper focuses on Canadian women in ten provinces aged 24 to 60. The objective of this paper is to test the assumption that the canonical model of labor supply is backward-bending for Canadian women. The second objective is to test if factors that determine labor supply decisions differ according to the economic well-being of the household where the female worker lives. However, this paper does not provide any policy prescriptions regarding the nature of the labor supply for Canadian women.
According to Robinson and Tomes (1985), given that men typically work more hours and receive higher wages than women, the larger income effect was expected to dominate the substitution effect for men, resulting in a backward-bending labor supply curve. On the other hand, for women the dominant substitution effect generates a positively sloped labor supply curve consistent with the upward trend in female labor force participation. In the United States, studies by Hall (1973) and Boskin (1973) provide empirical support for these arguments. In addition, the results for Canada reported by Carliner et al. (1980)2 for Canadian women are also congruent with this view.
However, this consensus has been challenged in a series of influential papers by Alice and Masao Nakamura3 (Nakamura, Nakamura, and Cullen, 1979; Nakamura and Nakamura, 1981, 1983) and also later challenged by Robinson and Tomes (1985) by affirming the conclusions reached by the Nakamuras with more recent and different data. Their results show that the estimated labor supply elasticities are predominantly negative, implying a backward-bending supply curve, and are broadly consistent with values typically reported for men. Hence the results of the Nakamuras and Robinson and Tomes suggest that there is no significant disparity between the labor supply elasticities of working men and women.
The results of the Nakamuras were of controversial nature at the time, and Nakamuras’ hypothesis was subject to independent tests especially by Robinson and Tomes (1985) and others. According to Robinson and Tomes (1985), past studies used census data that suffer from several drawbacks. They described that the first problem was that the actual hours of work were not recorded and instead intervals were used. The use of intervals is particularly problematic for females, because more females tend to supply hours outside the “normal” range and hence fall in much wider intervals than the males. The second problem was that an actual wage was not recorded. The wage has to be computed by dividing observed annual earnings by computed annual hours which results in a problem of division bias [Borjas (1980)]. Similar to the study of Robinson and Tomes (1985), this study deals with these problems but in a different manner4. By using cross-sectional data of the Survey of labor and Income Dynamics (SLID) from Statistics Canada, a direct measure of the hourly wage rate and the direct hours of work for a subset of women in Canada were obtained.
At the beginning of the 20th century, the average factory worker worked almost 60 hours per week, the equivalent of six 10-hour days (Borjas, 1996). It was not until the 1960s that the average hours in manufacturing conformed to the more familiar 40-hour workweek (eight hours per day, for five days). Figure 1.1 below illustrates the apparent degree of flexibility that Canadian workers have in the number of hours they work per week. The work patterns of women are quite varied. Men are more likely to work the typical 40-hour week. However, even for working men, less than half work 40 to 49 hours per week. Almost a quarter work more than 50 hours, and an equal proportion work fewer than 40 hours. Women, on the other hand, are more likely to work part-time than 40 hours (Ibid).
Figure 1.1: Distribution of Hours Worked per Week by Gender, 1996 5
Table 1.1 below traces the standard work week in Canadian manufacturing and shows a pronounced and continuous decline over time in hours of work. Between 1901 and 1981 the standard work week declined from almost 60 hours to less than 40 hours. The decline slowed down in the depression years of the 1930s, and the war years of the 1940s, and it appears to be slower in the postwar period. However, as the last column illustrates, when vacations and holidays are considered the decline in average working hours is more noticeable. In essence, in recent years the work force has reduced its working hours more in the form of increased vacations and holidays rather than a reduction in hours worked per week. The decline in net weekly hours in the postwar period and in standard hours prior to World War II give a long-run trend reduction of about two hours per decade (Borjas, 1996).
Table 1.1: Standard Weekly Hours in Manufacturing, Canada, 1901-19816
Since real wages have risen over the century, this long-run decline in hours worked appears inconsistent with an upward-sloping labor supply function (Borjas, 1996). Instead, it suggests an independent effect of increased wages: as people become wealthier, they need not toil as hard, and can afford to take more time off. Consequently, the classical theory will clarify the ways in which wages can affect labor supply.
A plausible explanation of the declining hours of work is the classical theory of labor supply. The classical theory of labor supply states that at low levels of income the substitution effect dominates which results in a positive elasticity of labor supply (raising wages raises hours of work). On the other hand, at high wage levels, the income effect dominates resulting in a negative elasticity (raising wages reduces hours of work). In other words, the income effect of higher wages means workers will reduce the amount of hours they work, because they can maintain a target level of income through less work. On the other hand, the substitution effect of higher wages means workers will give up leisure to do more hours of work because more work leads to higher rewards. As a result, the labor supply schedule forms a backward-bending shape for an individual (Robins, 1930). Since Canada is a developed country, this type of a backward-bending labor supply schedule for the majority of Canadians is expected in this study.Continued on Next Page »