The Economics and Statistics Division maintains archives of previous publications for accountability purposes, but makes no updates to keep these documents current with the latest data revisions from Statistics Canada. As a result, information in older documents may not be accurate. Please exercise caution when referring to older documents. For the latest information and historical data, please contact the individual listed to the right.
<--- Return to Archive
For additional information relating to this article, please contact:
February 26, 2020STUDY: UNDERSTANDING DEVELOPMENTS IN INDIVIDUALS' EARNINGS DISPERSOIN IN CANADA USING MATCHED EMPLOYER-EMPLOYEE DATA Statistics Canada recently released a study on the dispersion of individuals’ earning over 2001 to 2013, finding that overall earning dispersion declined slightly over this period as rising dispersion in top half of the distribution was offset by convergence in the bottom half. Earnings dispersion is a measure of the inequality of workers earnings comparing how much more is earned at the top end compared to the bottom end of the earnings distriubtion. It compares the amount earned by the workers who earn the most with those that are earning the least. For the study, the overall earnings dispersion is measured by the p90/p10 ratio that compares the earnings of workers at the 90th percentile with the earnings of a workers at the 10th percentile. A rising p90/p10 ratio indicates earnings are becoming more unequal with a further seperation or dispersion between those at the bottom and at the top of the earnings distribution.
The study notes that the literature shows that inequality in Canada and other countries has risen since the late 1970s and recent studies have attempted to assess the role of firms in explaining inequality. Studies of US data (Song et.al 2019) find that rising differential earnings between firm account for a significant portion of overall rising inequality. Widening dispersion between firms may reflect changing worker composition or changes in productivity between firms that arise from new technology, globalization, or market power. This paper uses the first available Canadian data to look at the firm-inequality issue with the Canadian Employer-Employee Dynamics Database (CEEDD). The findings are consistent with recent studies that show that inequality while higher than the late 1970s has been steady of slightly declining since 2000.
The CEEDD data base is matched employer-employee database based on business and individual tax files. The dataset allows studying between-firm and within-firm earnings dispersion with information on individual employee’s firm, age, gender, marital status, immigrant status and job earnings. The firm-level data allows for firm specific productivity measures and information on industry, firms size, payroll, tangible assets, revenues, expenses and profits. For the study, overall earnings dispersion is measured by the ratio of the 90th percentile earner with the 10th percentile earner (p90/p10 ratio). Change in the upper half of the earnings distribution are measured by the ratio between the 90th and the 50th percentile (p90/p50 ratio) and the lower half of the earnings distribution by comparing the 50th and 10th percentiles (p50/p10 ratio).
For Canada, overall earnings inequality increased 5 per cent from 2001 level to peak in 2005 afterwards it shrank to be 1 per cent lower in 2013 than in 2001. The top half of the earnings distribution increased over this period with the p90/p50 ratio rising steadily with a 4 per cent increase from 2001 to 2013. The dispersion of the lower half of the distribution declined after 2004 with a decline of about 5 per cent over 2001 to 2013. A compression in the lower half of the earnings distribution is consistent with a previous study (Fortin and Lemieux, 2015) that suggests minimum wage increases since 2005 could have made it possible for bottom of distribution to grow faster than middle or top end of the distribution.
Earnings dispersion changes were not the same for all provinces. Most provinces, except those most reliant on natural resources (Newfoundland and Labrador, Saskatchewan, Alberta, and British Columbia) had decreases. Nova Scotia experienced a 15.2 per cent decline in the p90/p10 ratio between 2001 and 2013. In contrast to Canada, the Maritimes and Quebec experienced contraction in the p90/p50 ratio over the 2001 to 2013. However, the decrease in the p50/p10 ratio was larger than the decrease for the upper end of the distribution for these provinces. All provinces except Newfoundland and Labrador experienced a decrease in dispersion in the lower end of the earnings distribution.
Earnings dispersion by sector were similar to the national trend with respect to direction of change. All sectors (agriculture, resources, utilities, construction, manufacturing and services) experienced declines in overall earnings dispersion (p90/p10) ratio with dispersion falling at the lower end (p50/p10). All sectors, except utilities, had dispersion increasing at the upper end of the distribution.
Earnings inequality was higher for larger firm size. Over 2001 to 2013, dispersion increased only among firms with more than 500 employee driven by rise in the upper end of the distribution. Workers in firms of all sizes experienced a decrease in dispersion (p50/p10 ratio decline) at the lower end of the earnings distribution.
Both male (+1.2%) and female (+1.3%) workers had a slight increase in earnings dispersion between 2001 and 2013. Male workers experienced a large increase in upper end of distribution and large decrease in lower end of distribution while changes in the female earnings distribution were smaller. Male workers experienced more pronounced polarization in earnings with larger increase at the top and bottom of distribution compared to the median. Although both males and females had increases in their dispersion, the overall earnings distribution converged slightly due to a composition effect. The top earnings are more effect by male workers, which grew slower than bottom earnings that are more effect by female workers.
Earnings dispersion can be separate into effects between-firms and within-firms using the variance of the distribution of earnings. Over 2001 to 2013, the contribution of the within-firm earnings variance to overall inequality account for more than 60 per cent of total earnings variance. The within-firm variance increased until 2007 but decrease afterwards. The between-firm variance increased slightly over the period. The average within-firm variance decreased 5.7 per cent since 2001 while the between-firm variance increased 2.8 per cent. The results for Canada are directional consistent with US data (Song et al. 2019) but the increase variance for between-firm earnings was larger in the United States.
Overall, falling inequality was driven by the within-firm variance in earnings declining. The increase in between-firms did place upward pressure on overall inequality and was driven by a widening gap between workers in firms in different industries.
The between-firm variance increase in Canada between 2001 and 2013 could be due to difference in firm performance. Productivity difference between firms has been linked to between-firm inequality for the post-2000 period in OECD member countries. The underlying theory is that firms experience technology driven productivity increases are likely to pass some of increase to workers wage as form of rent-sharing. Overall dispersion in labour productivity increased after 2001 in both manufacturing and service sector with notably separation for the 90th percentile compared to the median.
The study estimates correlation between firm productivity and earnings for various specification. Labour productivity dispersion and earnings dispersion are positively correlated. The correlation was stronger for the bottom half of the distribution than the top half suggesting that top-level jobs at firms with the highest productivity are have excessively high earnings above relative productivity. The link between labour productivity and firm earning dispersions was weaker in Canada than other OECD countries. This maybe due to difference in market competitiveness between Canada and other countries with less competitive markets tending to have a weaker link between wages and productivity.
At a firm level, earnings were positively correlated with firm-level productivity even after controlling for firm size, industry, region and firm fixed effects. The rent sharing passthrough was a 1% increase in firm-level labour productivity resulted in a 0.129% increase in firm-level average earnings. These estimates suggest that for 2001 and 2013, the rising dispersion of labour productivity contributed to about 22 per cent of the increase in the earnings dispersion between firms. The rent-sharing estimates appear to decline over time, suggesting that even as firm productivity diversion increase, the passthrough to wages was declining and limiting pressure on earnings inequality between firms.
Source: Statistics Canada, Understanding Developments in Individuals’ Earnings Dispersion in Canada Using Matched Employer–Employee Data
<--- Return to Archive
