4 Accounting for change

Warning View the most recent version.

Archived Content

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.

In this section, we use the Oaxaca-Blinder decomposition to determine the extent to which compositional changes can 'account' for changes in mothers' employment rates and earnings. Our analysis proceeds in two parts. First, we ask to what extent changes in demographic composition can account for changes in the probability of being employed among all single mothers. Second, we consider the effects of compositional changes on the log earnings (annual and weekly) of those who were in paid work. In light of our descriptive results, we conduct our analysis separately for mothers under and over age 40.

4.1 Mothers' probability of being employed

The ordinary least-squares (OLS) regression estimates for the probability of being employed are shown in Appendix Tables 1 and 2 and the results of the multivariate decomposition are summarized in Table 5.

Table 5
Decomposition of change in employment rates among lone mothers with children under 18 years of age

Over the two decades, single mothers' employment grew by just over 12 percentage points in both countries. The U.S. gains occurred mainly in the 1990s and while compositional shifts accounted for all of the increase in the 1980s they accounted for less than a third of the change over the entire period. Canadian gains were more evenly spread over the two decades and changes in composition accounted for almost two-thirds of the total change. These results for all single mothers, moreover, reflect large cross-national differences in the experience of younger and older mothers.

In both countries, compositional shifts accounted for most of the change among older (40 and over) single mothers—84% (15.1/17.9) in Canada and 72% (9.2/12.8) in the United States. Since Canadian single mothers are older than their U.S. counterparts, these compositional effects had a larger impact on trends for all Canadian single mothers.

Among younger single mothers, changes in behaviour and other unmeasured variables were the main source of rising labour force participation and occurred mainly in the 1990s. Changes in composition played a minor role. Behavioural effects (changes in the coefficients), however, were much larger in the United States than in Canada—9.2 percentage points and 4.7 percentage points, respectively. Since U.S. single mothers are younger than their Canadian counterparts, these behavioural effects had a much larger impact on the trend for all single mothers.

4.2 The earnings of employed mothers

The OLS regression estimates for log annual earnings of employed mothers are shown in Appendix Tables 3 and 4 and the results of the multivariate decomposition are summarized in Table 6.

Table 6
Decomposition of change in log annual earnings among employed lone mothers with children under 18 years of age

In both countries, virtually all of the earnings gains over the two decades went to women over 40. Over the two decades, earnings for older single mothers rose 29% in the United States and 22% in Canada, with most of the gains coming during the 1980s. Moreover, compositional shifts were the main source of change, accounting for 65% of the gains among older U.S. single mothers and all of the gains among older Canadian single mothers.

Among younger single mothers, there is little change to account for. In the 1980s, their earnings actually declined by about 4% in both countries. During the 1990s, however, the earnings of younger single mothers in the United States rebounded somewhat, rising by about 9%. Moreover, all of the increase is attributable to changes in the coefficients, while compositional changes actually depressed earnings slightly. In Canada, by contrast, the earnings of younger single mothers were essentially stagnant.

In summary, the largest earnings gains in Canada and the United States went to older single mothers over the two decades and were mainly the result of changes in socio-demographic composition in the 1980s. In the United States, but not in Canada, earnings of younger single mothers rose in the 1990s and were mainly the result of behavioural change, a result consistent with the conclusion that U.S. welfare reform was an important causal agent.

The results for annual earnings, however, leave several important questions unanswered. Changes in annual earnings reflect changes in both wage rates and labour supply (hours and weeks worked). One must disentangle the effects of these two factors to answer a number of important questions. For example, the large gains in annual earnings experienced by older single mothers as a result of changes in their educational levels and other characteristics could occur for two reasons: first, one would expect that higher levels of education would result in higher earnings per hour and week worked; second, it could be that the better educated were simply working more weeks and hours per year.

To address this sort of question, we compare changes in log annual earnings with changes in log weekly earnings, controlling for whether the respondent usually worked on a full-time or part-time basis.7 We take advantage of the fact that the change in the mean of log annual earnings is simply the sum of the change in the mean of log weekly earnings and the mean of log weeks worked so that changes in log weeks worked can be simply calculated as a residual.

Among older (40 and over) single mothers (Table 7) the results indicate that most of the gains reflected their increased earnings power as a result of higher education levels. Over the two decades, annual earnings rose by 22% and 29% in Canada and the United States, respectively. Comparison with the results for weekly earnings indicate that most of these gains—59% (0.13/0.22) in Canada and 73% (0.21/0.29) in the United States—were the result of higher earnings per week rather than more weeks worked per year. Moreover, in both countries, the increase in weekly earnings was mainly the result of compositional changes that occurred in the 1980s.

In contrast, the pattern among younger single mothers, especially in the United States, indicates that most of the gains in annual earnings were the result of increased work effort. Consistent with the welfare reform thesis, annual earnings in the United States rose by 10% in the 1990s but rose only by 2% on a weekly basis, indicating that most of the gains were the result of increased labour supply (weeks worked). In Canada, a small decline in weekly earnings (4%) was offset by a modest increase in weeks worked, and annual earnings rose by 2%.

Comparison of the decomposition analyses for weekly earnings among younger single mothers is also instructive. Both countries have experienced declines in the wage rates of younger workers relative to older workers since the 1970s (Boudarbat, Lemieux and Riddell 2006; Card and DiNardo 2002) and this trend also affected the earnings of younger single mothers. In both countries, small compositional gains reflecting higher levels of education and other characteristics were offset by a large decline in the 'returns' to these characteristics. Over the two decades, returns to labour market relevant characteristics (change in the coefficients) declined by 9% in the United States and by 14% in Canada on a weekly basis, declines that were only partially offset by compositional changes. As a result, total weekly earnings of younger single mothers declined by 2% in the United States and by 9% in Canada over the two decades.

In short, gains in annual earnings among older employed single mothers mainly reflected their increased earnings power as a result of higher education levels rather than increased labour supply. In contrast, the modest gains of younger lone mothers mainly reflected an increase in weeks worked.

 

7. Ideally we would take account of changes in both weeks worked and hours worked per week (it equals the total hours worked per year) to calculate changes in hourly wages. The Canadian census provides no way to estimate annual hours worked.