Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.
- Data and concepts
- The evolution of the male-female wage gap
- The evolution of wage differences by age
- The wage gap between levels of education
- Changes in inter-industry wage differentials
- Changes in the occupational wage structure
Hourly wages are a key driver of employment income and—since employment income is the main income source for most people—a major determinant of financial well-being. Differential wage growth influences the likelihood of different groups of individuals entering into or moving out of low income and has an effect on their capacity to handle unexpected expenditures. Furthermore, changes in group-specific wages inform discussions of the extent to which different groups of workers have shared in productivity and gross domestic product (GDP) growth in Canada over the last three decades.
In Canada and many other industrialized countries, the hourly wages of various groups of workers grew at markedly different rates over the last thirty years. A massive literature dating back to the mid-1980s has attempted to uncover why the pay rates of some workers grew faster than those of others. Technological changes, growth in international trade, institutional factors (e.g., de-unionization, changes in minimum wages, and changes in the incidence of pay-for-performance), movements in group-specific labour supplies, and changes in social norms have been cited as potential drivers of differential wage growth. 1
While the relative importance of the aforementioned factors is still open to debate, two simple, but important, questions emerge:
- Which groups of Canadian workers and which sectors of the Canadian economy experienced relatively stronger real-wage growth over the last three decades?
- To what extent do individuals’ acquisition of education, general work experience, and job tenure within firms, as well as their movements into low- or high-paying occupations and industries, account for the differences in real-wage growth observed across groups of workers?
This study focuses on how real wages of Canadian workers evolved from 1981 to 2011 across five dimensions: gender, age, education, industry, and occupation.
Previous research has documented several stylized facts regarding the evolution of wages by gender, age, and education. While women still earn relatively less than men on average, their wages grew faster than those of their male counterparts between 1981 and 2008 (Baker and Drolet 2010). Young workers lost ground relative to older workers during the 1980s and 1990s (Picot 1998; Morissette and Johnson 2005), but wage differences between the two groups stabilized between 1995 and 2005 (Boudarbat et al. 2010). Between 1980 and 2005, university graduates—especially men—experienced faster wage growth than high-school graduates (Boudarbat et al. 2010). What happened to the wages of men and women, young and older workers, and highly-educated and less-educated workers through the latter half of the 2000s is one of the issues addressed in this study.
In addition to updating wage trends across gender, age groups, and education levels, the study also examines wage trends across industries and occupations. Many questions arise in this regard. For instance, it is unclear whether wages in highly skilled services-producing industries (such as finance) grew in tandem with those in lower-skilled services-producing industries (such as retail trade) over the last three decades. Did the introduction of computer-based technologies coincide with a greater increase in the educational attainment of the workforce—and thus stronger wage growth—in the former group than in the latter? What happened to wages in textile, leather, and clothing as this sector faced growing international competition? How did wages evolve in mining and oil and gas extraction over the last decade, as world prices of oil and minerals increased sharply? Did the relatively strong wage growth experienced by managers between the late 1990s and the mid-2000s, which was documented by Morissette (2008), persist in the subsequent years?
Apart from documenting the evolution of wages across the five aforementioned dimensions, the study investigates why some groups of workers, some industries, and some occupations experienced faster wage growth than others. It is well-known that education, work experience, and job tenure within a firm are positively correlated with wages. In addition, wages differ across industries and occupations for observationally equivalent workers. Hence, some groups of workers (e.g., women, older workers) may have experienced relatively strong wage growth because they increased, more than did other groups, their educational attainment, their work experience, and/or their job tenure with their employers. Part of their wage growth may also be accounted for by a relatively high propensity to be employed in high-paying industries and/or occupations. The study quantifies the importance of these factors—denoted below as compositional effects—in accounting for differences in wage growth across groups of workers, industries, and occupations.
In addition to compositional effects, other factors may have generated differential wage growth across groups of workers, industries, and occupations. These include shifts in labour demand or labour supply induced by changes in technology, consumer preferences, international trade, foreign direct investment, immigration, and group-specific preferences for leisure. 2 In addition, changes in labour market institutions (e.g., changes in unionization rates, in minimum wages, and in the incidence of pay-for-performance schemes), social norms, and labour market discrimination may have generated differences in wage growth between groups. Because identifying the causal impact of all these factors in a statistical analysis is a daunting task, this study makes no attempt at quantifying the contribution of these factors. 3
The study pays particular attention to the first decade of the new millennium, which witnessed significant wage growth following two decades of pay stagnation between 1980 and 2000 (Plummer and Shannon 2011). Since Canada’s unemployment rate was relatively low during much of the 2000s—reaching a low of 6.0% in 2007—and since wages of lower-skilled workers tend to increase when the unemployment rate reaches historically low levels (Mishel et al. 2009), the wage growth observed during the 2000s may have reduced wage differences between highly-educated workers and their counterparts with lower education or between workers employed in highly skilled occupations and other workers. Likewise, the recession of 2008/2009 may have affected pay rates of Canadian workers to a greater extent in some industries or occupations than in others.
The study is organized as follows. Section 2 describes the data and concepts used in the paper. Sections 3, 4, and 5 analyze the evolution of wages across gender, age groups, and education levels, respectively. Sections 6 and 7 document wage growth across industries and occupations, respectively. Concluding remarks follow, in Section 8.
2 Data and concepts
The pay concept used throughout this study is the hourly wage. It is the best measure of direct remuneration since it is the purest price measure of labour services available in Canadian surveys. Differences in the quantity of labour provided—the number of hours worked per day, the number of days worked per week, or the number of weeks worked per year—can influence other measures often used, such as weekly or annual earnings.
Following Morissette and Johnson (2005), the study constructs a time series of hourly wages using several household surveys:
- the Survey of Work History (SWH) of 1981;
- the Survey of Union Membership (SUM) of 1984;
- the Labour Market Activity Survey (LMAS) for the years 1986 to 1990; and
- the Labour Force Survey (LFS) for the years 1997 to 2011.
All these surveys cover the same population, are based on the LFS sample design, and contain information on hourly wages received in the main job held by paid workers. 4 Together, they allow an assessment of the evolution of hourly wages from 1981 to 2011. This assessment is possible largely because the LFS started collecting hourly-wage data in 1997. 5
Apart from hourly wages, all these surveys contain information on workers’ age, job tenure— how long a worker has been employed in a given firm—educational attainment, union status, marital status, province of residence, industry, and occupation. These covariates are used throughout the paper to investigate the sources of differential real-wage growth across groups of workers.
Because the industry and occupation codes available in the data sets used for this study changed in the late 1990s-moving from the Standard Industrial Classification of 1980 (SIC 1980) and the Standard Occupational Classification of 1980 (SOC 1980) to the North American Industry Classification System of 2007 (NAICS 2007) and the National Occupational Classification for Statistics of 2006 (NOC-S 2006)—the evolution of wages is documented for two sub-periods: 1981 to 1998, and 1998 to 2011. 6 Although industry and occupation codes are not comparable between sub-periods, they are comparable within each sub-period. Whenever possible, trends are presented for the entire 30-year period, from 1981 to 2011. 7
Unless otherwise noted, the sample consists of paid workers aged 17 to 64 who are employed full-time in the main job they held in May. 8 Attention is restricted to full-time jobs in order to analyze changes in the wage structure for a sample of relatively homogeneous jobs. Alternative samples focus on workers aged 25 to 54 and thus exclude younger workers who may be combining school and full-time employment and older workers who may be considering retirement. Throughout the study, hourly wages are expressed in constant 2010 dollars deflated using the Consumer Price Index (All items).
While remuneration per hour worked is an important job aspect, other aspects such as whether a job is covered by a registered pension plan (RPP) or whether a job is temporary also matter. These aspects are not covered in the current study, in order to keep the analysis manageable and also because microdata on RPP coverage and the incidence of temporary work is not available for the entire 1981-to-2011 period. 9
The study relies heavily on decomposition methods to make inferences about factors that explain why some groups of workers experienced faster wage growth than others over the last three decades. When doing so, the study allows explanatory variables to be correlated with unobservable factors that affect wages, but it makes the assumption that the dependence structure between explanatory variables and unobservable factors is the same in the groups for which the decompositions are performed. For instance, when accounting for the narrowing of the gender wage gap, the study assumes that, while occupations (or education) are correlated with workers’ ability, the selection biases associated with occupations (or education) are the same for men and women. Under this assumption, wage decompositions performed on men and women—or on young workers and their older counterparts—remain valid (Fortin et al. 2011, p. 6).
While the decomposition methods used in this study are useful for understanding differences in real-wage growth across groups of workers, they have certain limitations. One important limitation is that they ignore "general equilibrium" effects. For instance, job losses in one sector alter the alternative options available to all workers in a local labour market. For this reason, they may affect not only wages of workers in that industry but also wages of other workers employed in the same local labour market (Beaudry et al. 2011). In line with a vast literature in labour economics, this study makes no attempt at quantifying these effects.
Before analyzing how real wages grew across gender, age groups, education levels, industries, and occupations, it is important to understand how real wages evolved in the aggregate over the last three decades.
Chart 1 shows that the last three decades resulted in periods of markedly different real-wage growth. Average pay rates of Canadian workers grew slowly during the 1980s and 1990s, before rising at a faster pace during much of the 2000s. From 1981 to 1998, average real hourly wages of full-time workers aged 17 to 64 increased by roughly 4%, less than half the 10% increase observed during the shorter, 1998-to-2011, period. Overall, average hourly wages of full-time workers increased by 14% from 1981 to 2011.
3 The evolution of the male-female wage gap
Women have traditionally earned less than men on average, as documented by numerous studies (Doiron and Riddell 1994; Baker et al. 1995; Gunderson 1979; Drolet 2002; Fortin and Schirle 2006; Baker and Drolet 2010). In addition, Statistics Canada released for many years, until 2009, Income Trends in Canada, a publication that provided estimates of the male-female earnings gap.
There is a debate regarding what measure of earnings should be used when reporting this gap. The report mentioned above, Income Trends in Canada, has traditionally used the annual earnings of full-time, full-year workers. When applying the metric used in that report, it has been found that women’s average earnings have been about 70% of those of men since 1992. However, Baker and Drolet (2010) made a convincing case that, with the availability of wage data from the LFS, comparisons based on hourly wages—rather than on annual earnings—are more appropriate. The reason is that gender differences in work hours may bias comparisons of annual or even weekly earnings. Using this metric, Baker and Drolet (2010) reported that women’s average hourly wages were 85% of men’s in 2008.
Following Baker and Drolet (2010), hourly wages are used in this section to measure differences in pay rates between men and women. Since the study seeks to document changes in the wage structure, the focus is on the change in the gender wage gap, particularly that which has taken place since the late 1990s. 10 The analysis investigates the degree to which declines in the gender wage gap are attributable to differential changes in work experience, education, job tenure, industry, and occupations of men and women. Since union status, marital status, and province of residence are also potential determinants of wages, their influence is also taken into account in the multivariate analyses.
From 1981 to 2011, hourly wages of men increased modestly, with median wages and average wages growing by 5% and 11%, respectively (Table 1, Chart 2, and Chart 3). However, most of this growth took place after 2005, as men’s hourly wages remained largely unchanged from 1981 to 2005.
In contrast, women’s median and average wages grew substantially faster between 1981 and 2011, increasing by 23% and 26%, respectively. As a result, the male-female hourly wage gap narrowed. In 1981, women aged 17 to 64 employed full-time had average hourly wages that were 77% of those of men; in 2011, the corresponding figure was 87% (Chart 4). However, these results are based on raw data. When gender differences in industry, occupation, education, age, job tenure, province of residence, marital status, and union status are taken into account, women’ wages amounted to 92% of men’s in 2011 (See subsection 9.2). 11 Among full-time workers aged 25 to 54, the corresponding number was 91%.
Using raw data, the paper finds that the gender wage gap based on medians decreased by half over the last three decades, dropping from 26% to 13% between 1981 and 2011. When using averages, the paper finds that the gender wage gap decreased from 23% to 13%. 12 A similar decline is observed for men and women aged 25 to 54 (Table 24).
Considering the 1981-to-1998 and 1998-to-2011 periods (each of which has consistent industry and occupation codes), one finds that more than half of the narrowing of the gender wage gap took place from 1981 to 1998 (Table 1 and Table 24). Why did the hourly wages of men and women tend to converge?
Part of the answer lies in the changing characteristics of female workers (tables 2 and 3). In both periods, women increased their average level of job tenure to a greater extent than men. For example, between 1981 and 1998, the average job tenure of women aged 17 to 64 increased by 25.6 months whereas men’s grew by 3.5 months. The educational attainment of women also increased more than that of men, as did their presence in high-wage industries (such as public services). Furthermore, women moved away from low-paying occupations (such as clerical jobs) and into high-paying occupations (such as those in health, social sciences, education, and government service) to a greater extent than men. Finally, women experienced smaller declines in unionization than men.
Table 4 quantifies the contribution of these factors to the narrowing of the gender wage gap. 13 Between 1981 and 1998, changes in job tenure, industry, and occupation were the major contributors, accounting for 91.0% of the narrowing of the gap between male and female workers aged 17 to 64 and for 95.9% of the narrowing of the gap between male and female workers aged 25 to 54. 14 Differential changes in educational attainment and unionization account for an additional 21.6% of the narrowing of the gap between male and female workers aged 17 to 64 and for an additional 18.8% of the narrowing of the gap between male and female workers aged 25 to 54. Changes in unionization played a fairly minor role. Overall, differential changes in education, job tenure, industry, and occupation account fully for the narrowing of the gender wage gap observed between 1981 and 1998. 15
While differential changes in industry were a major source of wage convergence between 1981 and 1998, they were not an important factor from 1998 to 2011. During the latter period, differential changes in job tenure, educational attainment, occupation, and unionization further reduced the gender wage gap, although their overall impact was more limited than it was between 1981 and 1998. Indeed, these four factors reduced the gender wage gap by close to 0.03 log points—roughly 3 percentage points—from 1998 to 2011, compared with 0.06 log points—roughly 6 percentage points—from 1981 to 1998 (Table 4). As a result, these factors account for most (85%) of the narrowing of the gender wage gap between 1981 and 1998, but for about half between 1998 and 2011.
Male-female differences in hourly wages have narrowed considerably since the early 1980s, as the hourly wages of women grew at a faster rate than those of men. The growing propensity of women to obtain higher education, remain in their jobs longer, and work in high-paying industries and occupations accounted for all of the convergence observed between 1981 and 1998. Subsequently, differential changes in education, job tenure, occupation, and union status accounted for about half of the observed improvement between 1998 and 2011. This in turn implies that other factors, not observed in the data sets used for this study, also contributed to the convergence observed through the 2000s. 16
4 The evolution of wage differences by age
In addition to the narrowing of the male-female hourly wage gap, other changes in Canadian wages occurred in the 1980s and 1990s. One of these was the increasing dispersion of male wages, particularly across age groups, as the wages of young men fell relative to those of older men (Morissette 1998; Picot 1998).
Beaudry and Green (2000) examined this phenomenon from a cohort perspective and concluded that, from 1978 to the mid-1990s, earnings fell for each successive cohort of young men entering the labour market. This pattern was observed among both young men with lower education levels and their more highly-educated counterparts. Furthermore, as young men entering the labour market in the 1980s and early 1990s aged and acquired more experience, there was little evidence of earnings "catch-up." In other words, the evidence was consistent with a downward shift in the age-earnings profile of young men, with no steepening of the age–earnings profile. No satisfactory explanation for these facts has been presented so far. 17
Quite different patterns were observed among women. Beaudry and Green (2000) found little evidence of a downward shift in the age-earnings profile of young women over the period considered. Using a time series of cross-sectional data, Picot (1998) showed that age-related earnings differences increased among women during the 1980s and early 1990s, but did so to a much lesser extent than among men. Furthermore, the growing age-earnings differences found among men resulted from a decline in the earnings of young men combined with an increase in the earnings of older men. In contrast, age-earnings differences among women resulted from faster earnings growth among older than younger women, although both groups experienced increases.
Following a period of decline, wages of young workers recovered in the 2000s. Median real wages of men aged 25 to 34 declined by about 10% between 1981 and the late 1990s, remained fairly constant until 2005, and increased afterwards, returning to their 1981 level by 2010 (Chart 5). Among men aged 17 to 24, the decline in median wages between 1981 and the late 1990s was larger, at over 20%, and the recovery by the end of the period was only partial. Meanwhile, wages of men aged 45 to 54 rose by about 10% in the early 1980s and have been fairly stable since then. 18 As a result, the difference in median real wages between men aged 25 to 34 and men aged 45 to 54 widened from 2% in 1981 to 23% in 1998, and then narrowed to 17% in 2011 (Table 5). 19 The corresponding numbers based on averages are 5%, 24%, and 18%.
As with men, wage growth among women varied according to age group between 1981 and 1998. While median and average wages of women aged 25 to 34 changed little over this period, those of women aged 45 to 54 grew by at least 20% (Table 6 and Chart 6). From 1998 to 2011, wages of women aged 25 to 34 grew slightly faster (i.e., by roughly 3 percentage points) than those of women aged 45 to 54. In sum, wages of men and women aged 45 to 54 grew faster than wages of their younger counterparts from 1981 to 1998 and slower from 1998 to 2011 (Chart 7). 20
Tables 7 and 8 illustrate why this happened. From 1981 to 1998, the unionization rates of young men and women decreased substantially—by at least 12 percentage points—while those of men and women aged 45 to 54 either dropped slightly or increased. 21 Young workers’ average job tenure fell whereas older workers’ (especially older women’s) job tenure increased. Young workers moved away from higher-wage sectors such as public services while older workers increased their presence in this sector. Finally, young workers’ presence in higher-wage occupations (such as those in natural and social sciences) grew less than that of older workers. These compositional effects accounted for about 40% of the widening of the wage gap between younger and older men during the period and for about three-quarters of the widening of the wage gap between younger and older women. 22
Compositional effects had a different impact from 1998 to 2011. While young men’s unionization rates changed little, the unionization rate of men aged 45 to 54 fell by 10 percentage points (Tables 7 and 8). Young men moved into high-wage industries, such as construction and mining and oil and gas extraction, to a greater extent than did men aged 45 to 54. As well, while men aged 25 to 34 maintained a stable share of employment in health, education, social services, and public administration, men aged 45 to 54 saw their employment share in these sectors drop by almost 6 percentage points. These differential changes in unionization, industry, and occupation explain roughly 60% of the narrowing of the wage gap between younger and older men through this period (Table 9). Differential changes in job tenure also played a minor role. About one-third of the narrowing of the age-wage differential among men remains unexplained.
As with men, differential movements in unionization, industry, and occupation favoured younger women between 1998 and 2011. In addition, younger women increased their educational attainment more than did older women over this period (Table 8). Together, these four factors account entirely for the faster wage growth of younger women since the late 1990s, and thus explain fully the modest narrowing of age-wage differences between younger and older women over the period (Table 9).
From the early 1980s to the late 1990s, compositional effects exerted upward pressure on the relative wages of workers aged 45 to 54 and downward pressure on the relative wages of workers aged 25 to 34. Conversely, through the 2000s, compositional effects exerted upward pressure on the relative wages of workers aged 25 to 34. From 1981 to 1998, compositional effects accounted for about 40% of the widening of the wage gap between younger and older men, and for about three-quarters of the widening gap between younger and older women. From 1998 to 2011, compositional effects accounted for virtually all of the modest narrowing of the gap between younger and older women and for about two-thirds of the more substantial narrowing of the gap between younger and older men. The end result was that, from 1981 to 2011, average hourly wages increased by 17% among men aged 45 to 54, but increased by only 1% among men aged 25 to 34. Hourly wages of women aged 45 to 54 grew by 33%, more than twice the 14% rate observed among younger women.
5 The wage gap between levels of education
Changes in the wage differentials between men and women and younger and older workers occurred while Canadian firms, along with those of many other Western industrialized countries, experienced substantial changes in technology and faced growing competition from abroad. 23 The technological changes that firms implemented and the growing presence of low-wage countries in international trade potentially affected the demand for highly skilled and lower-skilled workers and, thus, the wage growth experienced by these groups. In this context, did wage differences across education levels change? 24
The first wave of studies examining this question (Freeman and Needels 1993; Murphy et al. 1998; Burbidge et al. 2002) concluded that, for the Canadian labour force as a whole, the wage gap between workers with a university degree and other workers was relatively constant for men and declined for women over the 1980-to-2000 period. According to these studies, the "wage premium" associated with a university education had changed little over this 20-year period.
This view was subsequently challenged by studies that used different data sets and more control variables. Morissette et al. (2006) found that, among private-sector workers, wages of university-educated workers increased relative to those of high school graduates from 1980 to 2000. The increase was observed among both young men and their older counterparts as well as among younger women. No such increase was observed among public-sector workers. Looking at the 1980-to-2005 period, Boudarbat et al. (2010) concluded that, for the labour force as a whole, wages of university-educated workers rose relative to those of their less-educated counterparts, more so among men than among women. Most of the increase in the bachelor’s degree/high school wage ratio observed for men occurred between 1980 and 2000.
The LFS allows an investigation of what happened after 2000. 25
From 2000 to 2011, wages of male university degree holders grew more slowly than those of less-educated males. For instance, average hourly wages of men with trades certificates grew by 7.2%. 26 In contrast, the average hourly wages of males with bachelor’s degrees increased by only 2.7% (Table 10). 27 As a result, the wage gap between males with bachelor’s degrees and males with trades certificates narrowed during the 2000s. The wage gap between male bachelor’s degree holders and male high school graduates also narrowed, but to a lesser extent.
The narrowing of the wage gap was particularly evident among the young, as wage growth was very strong among less-educated younger workers. From 2000 to 2011, hourly wages grew by 16% among full-time male workers aged 17 to 34 with trades certificates and by about 8% among their counterparts with high school diplomas. Among bachelor’s degree holders, growth was roughly 1% (Table 11). Differences in wage growth across education levels were less pronounced among workers aged 35 to 64 (Table 12).
This recent narrowing of male wage differences across education levels is also observed in multivariate analyses that control for workers’ potential experience (charts 8 and 9) and in analyses that use weekly wages, instead of hourly wages, as a measure of pay rates. 28 A narrowing of pay differences is also found in such analyses among women (charts 10 and 11). Nevertheless, highly-educated workers today still earn much more than their less-educated counterparts. Controlling for potential work experience, average hourly wages of male bachelor’s degree holders were 41% higher than those of male high school graduates in 2011, down from 47% in 2000. 29 The corresponding result for women in 2011 was 55%, down from 61% in 2000. 30
The increase in the university wage premium over the 1980-to-2000 period is often viewed as resulting from a more rapid increase in the demand, relative to supply, for the university-educated than for the high-school-educated. Technological changes that are biased towards skilled workers are assumed to have increased the relative demand for highly-educated workers (Katz and Murphy 1992). This interpretation has often been applied to the United States, which experienced faster growth in the university wage premium than Canada. It also fits with other changes taking place in the Canadian labour market.
For instance, employment in high-knowledge industries grew by 84% between 1981 and 2001, compared with 52% and 32%, in the medium- and low-knowledge industries, respectively (Morissette et al. 2006). Furthermore, high-knowledge industries not only experienced rapid employment growth, but the educational attainment of their workforces also rose more quickly than that of workforces in other industries. One would expect these factors to increase the labour demand for university graduates much more than that for high school graduates. However, according to Murphy et al. (1998), the supply of university graduates grew rapidly, more rapidly than the supply of high school graduates. The fact that the supply of university graduates—relative to high school graduates—grew faster in Canada than in the United States was viewed by these authors as the main reason why the university wage premium grew, if at all, less in Canada than in the United States.
The results above strongly suggest that the 2000s were different, as wage differences across education levels narrowed. This new pattern emerged in a period of three important economic shocks: a severe employment contraction in the computer and telecommunications industries between 2001 and 2004 with slow recovery thereafter; strong growth in the construction sector; and an increase in the world prices of oil and other commodities. 31 While the first shock may have reduced upward pressures on labour demand for certain university graduates, the next two supported employment growth in construction and in the mining and oil and gas extraction industries, thereby potentially increasing the demand for other workers despite the continued trend decline in manufacturing employment over the 2000-to-2011 period. Additional research is needed to confirm these hypotheses.
Overall, LFS data indicate that, between 2000 and 2011, education-related wage differences narrowed for men and women. This narrowing was most pronounced among workers under age 35. Combining these results with those of Morissette et al. (2006) and Boudarbat et al. (2010) suggests that, after increasing from 1980 to 2000, the wage gap between the less-educated and the highly-educated decreased during the 2000s, particularly among individuals under age 35, largely as a result of faster wage growth among less-educated workers. Nevertheless, highly-educated workers today still earn much more than their less-educated counterparts.
6 Changes in inter-industry wage differentials
Some Canadian workers may have experienced relatively stronger wage growth over the last three decades not only because they increased their education, labour market experience, and job tenure to a greater extent than others, but also because they moved to high-paying sectors, i.e., industries that pay observationally equivalent workers higher wages.
The existence of inter-industry wage differentials has been well documented (Gera and Grenier, 1994; Krueger and Summers 1988; Dickens and Katz 1987; Murphy and Topel 1987). Yet the factors underlying these wage differences are less well-known. The most obvious explanation emphasizes that different industries pay different wages simply because they employ individuals who hold different occupations and who differ in terms of experience, observed skill levels, and degree of unionization. However, sizable inter-industry wage differences remain even after accounting for these factors. Unobserved differences in worker ability could also play a role, but various studies (e.g., Murphy and Topel 1987) suggest that accounting for such differences does not eliminate inter-industry wage differentials.
The fact that sizable inter-industry wage differentials remain even after taking account of workers’ observed and unobserved characteristics suggests that inter-industry differences among employers matter. One explanation assumes that industries whose firms face high training costs or high costs of monitoring worker effort will pay higher-than-average wages, denoted "efficiency wages," in order to reduce worker turnover or increase productivity. 32 Alternatively, industries that have relatively high profits per worker may share some of these profits with their workers by offering them higher-than-average wages. 33 , 34
While the relative contribution of the aforementioned factors is still an open question, most studies find that inter-industry wage differentials are relatively stable over time. However, inter-industry wage differentials do change somewhat over time, as different industries are exposed to, and respond to, different changes in their economic environment. 35
How did wage growth vary across industries over the last three decades? Did pay rates in the goods sector grow in tandem with those in the services sector? Was wage growth uniform across all services industries? What happened to wages in textile, leather, and clothing during the 1980s and 1990s, as this sector faced growing international competition? How did wages evolve in mining and oil and gas extraction over the last decade, as world prices of oil and minerals increased? This section answers these questions.
The definitions of industrial sectors available in the data sets used in this study changed significantly during the three decades considered, moving from being based on SIC 1980 during the 1981-to-1998 period to being based on NAICS 2007 during the 1998-to-2011 period. Except for a few sectors, these two industrial classifications are generally not comparable. 36 Rather than attempt to reconcile the differences between these classifications and establish one single coding system—which requires aggregating several industries and the potential loss of inter-industry wage variation—the analysis is conducted separately for the two sub-periods.
1981 to 1998
From 1981 to 1998, average wages grew by 3.4% at the aggregate level (Table 13). Most of the broad industrial groupings considered in Table 13 for which wages increased registered an average wage growth that varied between 1% and 6%. The one exception was business services, which displayed an average wage growth of 13.5%; within this industrial grouping, the financial sector experienced wage growth of 20.1%. 37 In contrast, construction and retail trade saw average wages decline by roughly 6% and 2%, respectively. As a result, substantial cross-industry differences in wage growth were observed.
These substantial movements in industry-level wages occurred while the Canadian workforce became more educated and while the mix of occupations within industries was transformed as a result of technological changes, among other factors. However, these changes in occupations and in workers’ education were not uniform across industries. For instance, the percentage of clerical workers dropped by 14 percentage points (from 54.2% to 40.1%) between 1981 and 1998 in finance but fell by only about 4 percentage points (from 21.8% to 18.0%) in retail trade during this period. Conversely, the percentage of workers with a university degree increased by 13 percentage points (from 13.7% to 26.7%) in finance but by only 4.5 points (from 3.8% to 8.3%) in retail trade. As a result, the profile of workers changed much faster in finance than it did in retail trade. This likely affected inter-industry wage differentials.
Table 14 confirms this hypothesis. A comparison of numbers reported in columns 1 and 3 for retail trade and finance indicates that almost half of the difference in wage growth initially observed between these two sectors—i.e., the "extra" wage growth experienced by workers in finance relative to their counterparts in retail trade—can be accounted for by differential changes in workers’ age, education, job tenure, and occupation. 38 The same conclusion applies when comparing finance and construction. In many cases, differential changes in workforce characteristics and occupations accounted for a significant portion of the inter-industry differences in wage growth observed over the 1981-to-1998 period. 39
Compositional effects did not matter everywhere, however, as they fail to explain why wages in some manufacturing industries grew faster than wages in other industries. For instance, average (log) wages in paper and allied industries and in electrical products grew much faster than those in the leather, textile, and clothing industries, a sector that experienced a decline in real wages (Table 15, Column 1). Most of the differences in wage growth initially observed between these industries remain after controlling for changes in worker characteristics and occupations (Column 3), thereby indicating that changes in worker characteristics and occupations were not important factors. Other factors, which affected these industries differently, appear to have played a major role. Of these, changing international competition is a potential candidate. However, assessing its impact is beyond the scope of this study.
1998 to 2011
While most industrial sectors registered fairly modest wage growth during the 1980s and 1990s, the late 1990s and 2000s were different, as average wages increased by 10.3% from 1998 to 2011 (Table 16). Coincident with the commodity boom observed over much of this period, average wages in the resources sector (mining and oil and gas extraction) grew the fastest, at 21.0%. Wholesale trade, finance, and professional and technical services also posted above-average wage growth. Likely fueled by strong demand for housing, hourly wages in construction rose by 9.8%, more than offsetting the wage declines experienced from 1981 to 1998. In contrast, manufacturing and retail trade experienced growth in average wages of 5.0% and 3.0%, respectively.
As was the case from 1981 to 1998, increases in worker skills and employment movements towards highly paid occupations accounted for a significant portion of the wage growth observed in finance between 1998 and 2011. This can be seen by comparing columns 1 and 3 of Table 17. A similar qualitative conclusion applies to wholesale trade as well as to professional, scientific, and technical services. In contrast, compositional effects explained very little of the wage growth observed in the mining and oil and gas extraction industries. Wage growth in this sector was likely related primarily to rising labour demand associated with the commodity boom. Likewise, most of the wage growth in construction remained after accounting for compositional effects. This suggests that wage growth in construction was driven mainly by the shifts in housing demand observed during the 2000s.
From 1998 to 2011, average log wages in mining, quarrying, and oil and gas extraction grew 0.15 points faster than those in retail trade. Two-thirds of this difference remains after controlling for compositional effects, implying that differential changes in worker characteristics and occupations accounted for only one-third of this difference.
Combining data from both periods, one notable development is that average wages in finance grew by 20% from 1981 to 1998 and by 12% from 1998 to 2011, whereas average wages in retail trade fell by 2% from 1981 to 1998 and rose by 3% from 1998 to 2011 (Chart 12). As Chart 13 indicates, however, much of this difference in wage growth resulted from the relatively rapid upskilling of the workforce in the financial sector, a pattern that has also been observed in the United States (Philippon and Reshef 2007).
Overall, pay rates grew little in most industrial sectors from 1981 to 1998. The obvious exception was the financial services sector, which experienced strong wage growth, due largely to a fast-growing share of university graduates and changes in the types of occupations used. Among workers with comparable characteristics and occupations, wages fell substantially in the leather, textile, and clothing industries. From 1998 to 2011, several industrial sectors experienced close to two-digit wage growth. Yet there was considerable inter-industry variation in wage growth. The mining and petroleum sector experienced the fastest wage growth. Compositional effects had little to do with this fast wage growth, which coincided with increases in world oil prices and world commodity prices. 40 In contrast, retail trade posted fairly modest wage growth. Average real hourly wages in construction grew by roughly 10% from 1998 to 2011, after falling by about 6% from 1981 to 1998.
7 Changes in the occupational wage structure
Since the early 1980s, wages of Canadian workers may have grown at a different pace, not only across industries but also across occupations. The reason is that the human capital requirements of some jobs, the relative scarcity of individuals able to fill specific positions, the ability of workers to extract "rents" in some occupations, and social norms, all of which are potentially important determinants of pay rates, may have changed over the last three decades.
Because their wages have grown rapidly in recent years in North America, senior executives belong to an occupation that has received considerable attention over this period. While the growth in their wages has been well documented (e.g., Piketty and Saez 2006), the factors that underlie it are not well understood. Potential explanations cited in the literature include changes in pay-setting methods and social norms, the increasing ability of executives to set their own wages (and, thus, to extract rents from shareholders), and the rise in wages associated with an increase in the size and market capitalization of firms. Whatever the cause, the growing gap between senior executives and other workers has contributed to the growth of wage inequality in North America (Lemieux 2008).
The occupational categories available in the data sets used in this study do not allow an explicit identification of senior executives. However, "managers" can be identified. As was the case in the industry analysis, the special surveys used prior to 1998 followed a different occupational coding system than that followed by the LFS post 1998. Hence, a time series with comparable definitions of occupations cannot be produced for the entire 1981-to-2011 period. However, comparable occupational data can be produced for the two sub-periods, 1981-to-1998 and 1998-to-2011.
1981 to 1998
At the high level of aggregation used in Table 18, there was relatively little change in the occupational wage structure from 1981 to 1998. Some occupations displayed negative growth in median or average wages (e.g., professionals and managers, service occupations, and blue-collar occupations), while others displayed positive growth (e.g., occupations in natural and social sciences, clerical and sales occupations). However, in both cases, the change in average wages was relatively small, ranging from -5.3% to +7.3%. In particular, professionals and managers did not experience rapid wage growth. As columns 1 and 3 of Table 19 indicate, controlling for differences among occupations in gender, age, education, job tenure, and industry leads to a reduction in the inter-occupational variation in wage growth, but the basic result remains: relatively little difference in wage growth across occupations took place over the 17-year period.
1998 to 2011
The story for the subsequent years is quite different. As Table 20 shows, managerial occupations registered much higher wage growth than any other occupation from 1998 to 2011: the average wages for this type of occupation increased by 26.9% during that period. None of the broad occupational groups considered in Table 20 experienced a decline in wages during this period. Most displayed average wage growth ranging from 2% to 15%. Along with managers, workers employed in the natural and applied sciences and workers employed in primary industries also displayed relatively strong wage growth (about 14%). In contrast, workers employed in occupations unique to processing, manufacturing, and utilities registered more modest wage growth.
Was the strong increase in managers’ wages due to some form of upskilling—with job tenure or education levels rising over the period—and to changes in the industry of employment? Table 21 shows that about 40% of the wage growth experienced by those in managerial occupations can be attributed to such compositional effects (columns 1 and 3). Likewise, about 40% of the wage growth observed among workers employed in the natural and applied sciences and in the primary industries—two occupational groups that experienced relatively large increases in wages—can be attributed to similar compositional changes. 41
Columns 1 and 3 of Table 21 also show that most of the substantial difference in wage growth observed between individuals in managerial occupations, on the one hand, and workers employed in occupations unique to processing, manufacturing, and utilities, on the other, cannot be accounted for by compositional effects. Even after taking account of these compositional effects, average log wages of managers still grew much faster than those of workers in occupations unique to processing, manufacturing, and utilities. This suggests that other factors—such as those mentioned above as well as specific factors putting downward pressures on the wages of workers employed in occupations unique to processing, manufacturing, and utilities—underlie most of the difference in wage growth observed between these two occupational groups from 1998 to 2011.
This study examined how real wages of Canadian workers evolved from 1981 to 2011.
Although women today still earn less than men on average, the gender hourly wage gap decreased significantly over the last three decades. During that period, women increased their productivity-enhancing characteristics at a faster pace than men did. Virtually all of the decline in the gender wage gap over the 1981-to-1998 period can be accounted for by this process. During the more recent period, 1998 to 2011, differential changes in education, job tenure, occupation, and union status accounted for about half of the narrowing of the gap.
From 1981 to 1998, average wages of men and women aged 45 to 54 grew roughly 20 percentage points faster than those of their counterparts aged 25 to 34. Differential changes in unionization, job tenure, industry of employment, and occupation explain about 40% of this difference for men and about 80% of it for women. In contrast, men and women aged 25 to 34 experienced faster wage growth than their counterparts aged 45 to 54 from 1998 to 2011. Changes in unionization, industry of employment, and occupation tended to favour young workers (relative to older ones) during that period and account for most of the difference in wage growth observed across age groups during that period. Over the entire 1981-to-2011 period, average wages of workers aged 25 to 34 grew by at least 15 percentage points less than those of workers aged 45 to 54.
After growing from 1980 to 2000, the wage gap between university- and high-school-educated individuals decreased from 2000 to 2011, especially among workers under age 35. This decline was driven by more rapidly rising wages among the less-educated, a reversal of the trend in earlier decades. The narrowing of the gap was associated with a shift in labour demand towards industries, such as mining, oil and gas extraction, and construction, which are less intensive users of highly-educated workers. Nevertheless, highly-educated workers today still earn much more, on average, than their less-educated counterparts.
Wage growth varied significantly across industries over the last 30 years. Most notably, the finance sector experienced much faster wage growth than other sectors during the 1981-to-1998 period. Most of this sector’s wage growth was related to changes in the worker characteristics and occupations used in the sector, as upskilling appeared to be taking place. About one-half of the differences in wage growth between finance and construction—the sector that experienced the largest decline in wages during that period—was attributable to the more rapid acquisition in the finance sector of attributes associated with higher wages.
Movements in industry-level wages were also quite different during the 1998-to-2011 period. Unlike in the earlier period, during which few industries posted robust wage growth, several industrial sectors experienced close to two-digit wage growth after 1998. Wages in the resource sector (mining and petroleum) grew the fastest, and virtually none of this growth was related to changes in worker or job characteristics. In contrast, the retail trade sector experienced weaker wage growth. Between 1981 to 2011, pay rates in this sector changed little.
Finally, wage growth varied little across broad occupational groups from 1981 to 1998. However, during the first decade of this century, average wage growth for managerial occupations exceeded that observed in other aggregate occupational groups by a substantial margin. About 40% of the wage growth in managerial occupations between 1998 and 2011 was due to changes in attributes associated with higher wages. Even after such changes were accounted for, managers continued to register the highest wage growth. Furthermore, only about one-quarter of the difference in wage growth between managers and workers in processing/manufacturing/utilities was attributable to changes in worker or job characteristics over the period. This suggests that other factors appear to be driving the relatively rapid wage growth among managers. Quantifying the contribution of these factors is an area for future research.
9.1 Wage and hours concepts used in household surveys, 1981 to 2011
1981 Survey of Work History (SWH): usual wage or salary before taxes and other deductions; no reference is made to tips, commissions, bonuses, or overtime
1984 Survey of Union Membership (SUM): usual wage or salary before taxes and other deductions; no reference is made to tips, commissions, bonuses, or overtime
1986 Labour Market Activity Survey (LMAS): usual wage or salary before taxes and other deductions; no reference is made to tips, commissions, bonuses, or overtime
1987 to 1990 Labour Market Activity Survey (LMAS): usual wage or salary before taxes and other deductions; includes tips, commissions, bonuses, and paid overtime-all together
1997 to 2011 Labour Force Survey (LFS): wage or salary before taxes and other deductions, including tips and commissions; whether respondents include overtime pay is unclear
1981 Survey of Work History (SWH): usual days per week + usual hours per day; no reference is made to overtime
1984 Survey of Union Membership (SUM): weeks worked in 1984 + usual hours per day; no reference is made to overtime
1986 Labour Market Activity Survey (LMAS): usualpaid days per week + usual paid hours per day; no reference is made to overtime
1987 to 1990 Labour Market Activity Survey (LMAS): usualpaid days per week + usual paid hours per day; no reference is made to overtime
1997 to 2011 Labour Force Survey (LFS): usualpaid hours per week; explicitly excludes overtime
9.2 The wage gap between men and women in recent years
This appendix highlights two key findings. First, it shows that, among full-time workers aged 17 to 64 who were employed in broadly comparable industries and occupations, women’s wages amounted to 92% of men’s wages in 2011. Among full-time workers aged 25 to 54, the corresponding number is 91%.
Second, it shows that using a detailed set of industry and occupation controls allows one to explain a larger fraction of the gender wage gap in 2008 than is explained by Baker and Drolet (2010) using less detailed controls. While Baker and Drolet (2010) were able to explain 16% of the log wage gap observed between men and women aged 25 to 54 who were employed full-time in 2008, using a detailed set of industry and occupation controls allows one to explain 37% of this gap. The implication is that-while Baker and Drolet (2010) suggest that, among full-time workers aged 25 to 54 who were employed in broadly comparable industries and occupations, women’s wages amounted to 85% of men’s wages in 2008-using a detailed set of industry and occupation controls raises this number to 89%. The difference in estimates is due largely to the fact that the broad occupational codes used by Baker and Drolet (2010) suggest—surprisingly—that women are over-represented in high-paying occupations while detailed occupational codes suggest the opposite, i.e., that women are still over-represented in low-paying occupations.
9.2.1 The gender wage gap in broadly comparable industries and occupations in 2011
Table 22 is based on equation (2) of Baker and Drolet (2010, p. 447). It shows that, in 2011, the gender log wage gap amounted to 0.129 (roughly 13%) among full-time workers aged 17 to 64. Almost half of this gap can be explained by the fact that women are over-represented in low-paying industries and occupations. The unexplained component amounts to 0.076, i.e., roughly 8% [exp(0.076)-1], thereby implying that, in broadly comparable industries and occupations, women’s wages amounted to 92% of men’s wages in 2011. A similar exercise performed for full-time workers aged 25 to 54 yields an unexplained component of 0.087, i.e., roughly 9% [exp(0.087)-1]. This implies that, among full-time workers aged 25 to 54 employed in broadly comparable industries and occupations, women’s wages amounted to 91% of men’s wages in 2011.
9.2.2 The effect of using detailed occupation codes
Because they wished to use consistent industry and occupation codes over the 1981-to-2008 period and because industry and occupation classifications underwent significant changes during this period, Baker and Drolet (2010) were forced to use fairly broad occupation controls when they conducted their regression analyses (See the appendix in Baker and Drolet, p. 455-464). In contrast, this study uses 47 occupation controls (along with 23 industry controls) when accounting for the gender wage gap. As Table 23 shows, using detailed occupation controls has two major impacts. First, it allows one to explain a larger fraction (37%, or 0.062/0.166) of the wage gap than that explained by Baker and Drolet (2010) (16%, or 0.026/0.166). Second, it suggests that, contrary to the findings of Baker and Drolet (2010, Table 5, p. 449), women are still over-represented in low-paying occupations. In fact, 17% (0.029/0.166) of the gender log wage gap can be accounted for by this fact. In contrast, the negative estimate of -0.023 obtained by Baker and Drolet (2010) suggests that women are over-represented in high-paying occupations.