Chapter 5
Multivariate results

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.

Archived

This page has been archived on the Web.

[an error occurred while processing this directive]81-595-m[an error occurred while processing this directive] [an error occurred while processing this directive]

Three separate models are estimated for each dependent variable. The first model includes only the educational pathways variable and a dichotomous measure indicating if the respondent has been out of school for 2 years or for 6 years. This measure is a proxy for experience in the labour market. The second model adds background measures such as sex, place of birth and parental education, as well as two important factors during high school, frequency of working and marks. The third or full model adds variables from the year the labour market outcome is measured — interprovincial mobility between high school and the labour market year, age, presence of children in the household, marital status, presence of a long-term condition that limits work, province of residence, population size of the community where the respondent lived, and number of months worked (for the earnings regressions). For the most part, the discussion focuses on the results of the full models; however, any important and substantial changes across models also are noted.¹

5.1 Likelihood of being employed on a full-year basis for 12 months

5.1.1 Time 1 – 1 or 2 years after leaving full-time schooling

Table 5.1 presents results from a series of logistic regressions predicting the likelihood that respondents worked a full year (1 = worked 12 months, 0 = not) at two different time points following the exit from full-time school. These models produce odds ratios — an odds ratio less than 1 indicates a negative effect while an odds ratio above 1 indicates a positive effect.² A first glance at Model 3 for Time 1 does not reveal very many significant differences from the reference category (Non-gappers-University), with the following exceptions. At this initial time-point a couple of years out of school, high school leavers, 2nd Chancers with no post-secondary education, 2nd Chancers with some post-secondary education, and Gapper-post-secondary education leavers were all significantly different from the reference category; the latter two effects were weak, however. Specifically, for high school leavers, the odds of being employed were 55.8%³ less than those for Non-gappers with a university degree; for 2nd Chancers with no PSE the difference in the odds are even greater, at 64.9%⁴ less.

Table 5.1 Logistic regression results on full-year employment at two time points after respondents have left school on a full-time basis

In earlier models, youth with a high school diploma only (and who had not previously dropped out) had significantly lower odds of being employed than the university educated who went directly to post-secondary education after high school. However, this significant difference was removed once concurrent factors affecting employment were considered in Model 3. These contemporaneous factors also weakened the effects for 2^ndChancers with some post-secondary education and Gapper-post-secondary education Leavers.

The significance levels reported in Table 5.1 however do not tell the entire story since they only report the significant differences between each pathway and the control group, which in this case is Non-gappers with a university degree. A series of Wald tests were performed after each estimation in order to test for significant differences between the other pathways. These tests revealed that the greatest difference was between the Non-gapper-College-University pathway and almost all other pathways. With the exception of Non-gappers with college diplomas and Gappers with university degrees, Non-gappers who went to college and then university had significantly greater odds of full-year employment than every other educational pathway. This trend was also noted in the descriptive analysis. Moreover, at this first time point, there were no significant differences in the odds of full-year employment between the two 2nd Chance groups. In other words, a couple of years after school, youth who had ever dropped out of school and who then returned had similar odds of full-year employment, regardless of whether they went on to post-secondary education; these 2nd Chance youth also had similar employment levels to high school leavers who never returned.

In terms of whether taking time off between high school and post-secondary education matters for early employment, we do not see any significant differences. This suggests that going directly, or delaying the start of post-secondary education after high school, makes no difference for the odds of full-year employment a couple of years after leaving school; both groups have about the same chance of full-year employment. This finding is important as it enhances our understanding of the differences between pathways from what we observed with the descriptive analysis. With the multivariate findings we can now say with some confidence that the differences we observed in the descriptive analysis do not hold once other factors are taken into account; thus, the observed differences between Gappers and Non-gappers on employment a couple of years after leaving school appear to be explained by other factors.

In terms of the control variables, only two factors can be considered robust predictors of full-year employment. First, in terms of adolescent employment, similar to past research (Hango and de Broucker 2007; Marsh and Kleitman 2005), the results show that working 1 to 10 hours during high school increases the odds of employment by 50%; this decreases to 40% if they worked between 10 and 20 hours per week on average. However, these positive effects of working in adolescence on employment in early adulthood must be tempered somewhat by prior findings which showed that working a great deal in high school is associated with lower levels of educational attainment (Hango and de Broucker 2007). Second, in line with past research, having at least one child in the household decreases the odds of full-year employment by approximately 49% (also see Waldfogel 1998; Zhang 2009).⁵

5.1.2 Time 2 – 5 or 6 years after leaving full-time schooling

As before, Table 5.1 presents results from logistic regressions at both time points. Results from Model 3 at Time 2 reveal that the Non-gapper-University advantage in terms of full-year employment has all but disappeared. The only pathway with significantly lower odds of being employed on a full-year basis compared with the reference category is 2nd Chancers without any post-secondary education, who are almost 70% less likely than Non-gappers with a university degree to be employed. Wald tests reveal that this pathway of 2nd Chancers without any post-secondary education also has significantly lower odds of full-year employment than every other educational pathway, including high school leavers. Furthermore, these 2nd Chancers without any post-secondary education had significantly lower odds of employment than their 2nd Chance counterparts who had gone on to some type of post-secondary education. This difference did not exist at the first time point, suggesting that it takes several years for the positive effect of post-secondary education to emerge in the labour market for youth who had ever dropped out of high school but who then returned to complete their high school diploma.

In contrast, a significant effect, albeit weak, emerges at this time point which did not exist previously. Several years after leaving school, Gappers with a university degree are significantly more likely than their Non-gapper counterparts to be employed on a full-year basis. These highly-educated Gappers are over three times more likely than their Non-gapper counterparts to work a full-year at Time 2. It is interesting to note that this difference was observed during the descriptive analysis as well (see Table 4.1). Furthermore, the Gapper-University pathway also has significantly higher odds of full-year employment than every other pathway, except for Non-gappers who obtained a college diploma prior to obtaining a university degree. No other differences were observed within, as well as between, education types for Gappers and Non-gappers at this time point. This implies that several years after leaving school, the university- and college-educated in general are remarkably similar, except for Gappers with a university degree who had significantly higher odds of full-year employment than everyone else, including their college-educated counterparts. Thus, at this second time point, the real difference between the university educated and the other pathways is between Gappers with a university degree and everyone else; this demarcation did not exist at Time 1.

Moreover, what these results suggest is that several years after leaving school, some changes with regard to employment appear to be occurring within certain educational pathways. As we saw in the descriptive analysis, almost every pathway had a full-year employment rate that was higher (and in some cases much higher) at Time 2 than at Time 1.⁶ The changes across time within pathway do not remain, however, once we control for important covariates. Moreover, Wald tests reveal that the odds of employment do not change significantly across time points, suggesting remarkable consistency over a 6-year span once factors such as age, sex, and other demographic variables are taken into consideration in Model 3.

With respect to the control variables, four new factors are now significant at Time 2 that were not present at Time 1: being female decreases the odds of full-year employment by 30%; being Canadian born decreases these odds by close to 75%;⁷ having a long-term physical or mental condition reduces the odds by 50%;⁸ and being married or living in a common-law relationship increases the odds of full-year employment by about 32%. At the same time, similar to Time 1, we also note a significant positive effect of working during high school and a significant negative effect of having at least one child in the household.

5.2 Ordinary least squares regression on log of yearly earnings

5.2.1 Time 1 – 1 or 2 years after leaving full-time schooling

The second labour market outcome examined in this report is annual earnings from salary or wages (indexed to 2006 using the Consumer Price Index), logged in order to reduce skewness. All effects are interpreted as percent change in earnings as a result of each independent variable. To obtain the exact percentage change in earnings the following formula is used: eâ-1, however, â is a good approximation of percentage change at relatively small values.

From Table 5.2, we observe in our final model of Time 1 that, in general, any pathway that does not lead to the attainment of a post-secondary education credential results in significantly lower earnings than university graduates who did not have a gap between high school and first post-secondary education program. This is similar to recent descriptive findings by Shaienks and Gluszynski (2009) using the same dataset. For instance, when controlling for all other factors, high school leavers earned about 36%⁹ less than the Non-gapper-University path. Moreover, all 2nd Chancers earned considerably less than Non-gappers with a university degree: 37% less for those without post-secondary education and 28% less for those with some post-secondary education.¹⁰ Also, there appears to be a penalty for not completing a post-secondary education program: post-secondary education Leavers earned about 22%¹¹ less than the Non-gapper-university path, regardless of whether the post-secondary education Leavers had delayed the start of their post-secondary education program. These effects are very similar to what was observed in the initial model for full-year employment.

Table 5.2 Regression results on logged earnings at two time points after respondents have left school on a full-time basis

When comparing educational pathways to each other (as opposed to only the reference category) via Wald tests, we find that 2nd Chancers have similar earnings regardless of whether they had continued on to post-secondary education. Moreover, 2nd Chancers did not earn significantly more than high school leavers. Additionally, no significant earnings differences were noted between Gappers and Non-gappers. However, at this early time point after leaving school, it is interesting to observe that the university advantage, as compared to non-post-secondary education pathways, only extends to youth who went directly to post-secondary education following high school graduation. In other words, youth with university degrees who delayed going to university after high school did not earn significantly more than youth with high school diplomas only or even youth who had dropped out of high school prior to obtaining their diploma.

With respect to control variables, we note that females have lower earnings than males, earning about 33%¹² less, on average (see Frenette and Coulombe 2007; Thomas and Zhang 2005). Other negative effects include having a long-term physical / mental condition that limits work; living in the Atlantic Provinces; and living in larger population centres. For instance, having a limiting condition lowers earnings by about 11%.¹³ Youth who live in the Atlantic Provinces earned about 10% less than youth from Ontario;¹⁴ in contrast, youth living in the West earned about 13% more than youth living in Ontario.¹⁵ Meanwhile, youth in larger population centres earned almost 8% less than youth from rural areas.¹⁶ These effects are weak, however, and should be interpreted with this in mind.

Last, as with employment, we also observe a positive effect of working during high school on later earnings: the positive impact peaks at 10 to 20 hours a week. These youth have about 28% higher earnings than youth who did not work during high school.¹⁷ As expected, positive effects are also noted for age, respondents who were married or in common law relationships and for number of months worked during the year. Interestingly, net of education, a weak significant effect is observed for marks in the last year of high school. Specifically, youth with average marks between 60% and 69% earned about 14% less than youth whose average marks were between 70% and 79%.¹⁸ This appears to be in line with past research using more objective measures of cognitive ability (see Green and Riddell 2001; McIntosh and Vignoles 2001).

5.2.2 Time 2 – 5 or 6 years after leaving full-time schooling

Table 5.2 shows the results from regressions predicting the log of annual earnings at Time 2. Model 3 illustrates very similar findings to those from Time 1, namely, that high school leavers, respondents with high school only, all 2nd Chancers and all PSE leavers earned significantly less than Non-gappers with university degrees. However, at this time point, we also now observe significantly lower earnings for all college graduates (regardless of gap) and Gappers with a trade or other type of certificate compared to Non-gappers with university degrees. For instance, college graduates who had not delayed college attendance after high school earned 16% less than Non-gappers with a university degree; their Gapper counterparts earned about 22% less.¹⁹

Moreover, with respect to other education levels, Wald tests do not reveal any significant earnings differences between Gappers and Non-gappers, or between the two 2nd Chance pathways. These findings are consistent across both time points, suggesting a different pattern than that observed for employment where 2nd Chancers with post-secondary education and university-educated Gappers had greater probabilities of full-year employment than their counterparts with no post-secondary education and Non-gapper university graduates.

Also, in terms of change across time, the earnings results differ from the employment results. Between Time 1 and Time 2, the earnings coefficients were significantly lower for individuals in the Non-gappers with college diplomas pathway and Gappers with trade certificates or other types of diploma pathway. This suggests that these two pathways may not be keeping pace with Non-gappers with university degrees several years after leaving school.

A key finding of the analysis at Time 2 is that the university educated, regardless of gap, appear to be pulling away from almost all other educational pathways in terms of yearly earnings.²⁰ The exceptions are Non-gappers with trade / other certificates versus all university educated and Non-gappers with college diplomas versus Gappers with university degrees. This trend was also observed in the descriptive analysis. However, these multivariate findings indicate that the pattern is real and substantial after controlling for important factors known to affect labour market outcomes. Thus, as youth are in the labour market longer, having a university degree appears to pay off more compared to having a college diploma or some other type of post-secondary education credential.

With respect to the control variables, several factors remain consistent over time. For instance, the negative effects of being female, having a physical or mental health condition, living in the Atlantic Provinces and having low marks remain; while the consistent positive effects of working during high school, being in a marriage or common-law relationship, living in the West and number of months worked also remain. This latter effect, while still significant at Time 2, decreases somewhat, which is in line with findings by Thomas and Zhang (2005), who suggest that number of hours worked has a larger impact in determining earnings at the beginning of one's career.

---

Note

1. All results presented in Tables 5.1 and 5.2 make use of standard errors derived from re-sampling each model 1000 times using the provided bootstrap weights (Statistics Canada 2003). In all analyses, the statistical program Stata Version 10.1is used (StataCorp 2008).
2. Interpreting odds ratios can often be confusing. Refer to Long (1997) for an explanation and correct interpretation.
3. 1-0.4421
4. 1-0.3509
5. The research on effects of children has almost exclusively been confined to women however. (1-0.5064)
6. The exception was the Non-gapper-College-University pathway which stayed about the same over time.
7. This counterintuitive finding is likely due to the composition of the analytical sample. As shown in Table 2.1, the sample selected for analysis has a lower education level than the average Youth in Transition Survey (YITS) sample, and additional analysis demonstrated that the foreign born with lower levels of education have higher employment rates than the Canadian born, at Time 2. As a result, we may be more likely to observe higher employment rates among the foreign born at Time 2 simply because the Time 2 sample has, on average, a lower level of education. These alternative models are available upon request.
8. Calculated as follows: 1-0.6914; 1-0.2635; 1-0.4981, respectively.
9. exp(-0.4455) – 1 = -0.3595
10. exp(-0.4582) – 1 = -0.3676; exp(-0.3285) – 1 = -0.2800
11. exp(-0.2567) – 1 = -0.2264 or exp(-0.2371) – 1 = -0.2111
12. exp(-0.3969) – 1 = -0.3276.
13. exp(-0.1191) – 1 = -0.1123
14. exp(-0.1114) – 1 = -0.1054
15. exp(0.1222) – 1 = 0.1300
16. exp(-0.0807) – 1 = -0.0775
17. exp(0.2503) – 1 = 0.2844
18. exp(-0.1506) – 1 = -0.1398
19. exp(-0.1785) – 1 = -0.1635; exp(-0.2483) - 1 = -0.2199
20. This includes both Gapper and Non-gapper university paths, and also the Non-gapper-College-University path.