Common menu bar

Breadcrumb

Education and Early Labour Market Outcomes in Canada - December 2007

4. Regression results

4.1 The effect of education on wages

The wage differences across educational levels discussed above do not control for any other differences that may exist between individuals with different levels of education. To illustrate the importance of controlling for such differences in order to better measure the wage effect of education, a set of linear regression models differing by the set of control variables were estimated. The linear regression models were estimated separately for each region and the results for males are presented in Tables 11a-11d. Female results are shown in Tables 12a-12d. In addition to these regression results, Tables 11e and 12e show results when data from all regions have been pooled. In each table, the entries in the first column show the log-wage difference across different levels of education (all relative to high school drop-outs) obtained from a regression which includes no other observable characteristics of the individuals. These figures naturally show the same pattern as in Tables 9a and 9b. In column two, a control for time since last in school was added. In order to control for differences in labour market ability, the regression specification in column three also includes information on average high school grades and self-reported scholastic abilities, all assumed to be highly correlated with labour market ability. The final specification also includes information on family background in addition to the regressors included in specification III.

Table 11a: Wage regressions for males in Atlantic Canada
	Specification
	I	II	III	IV
High School only	0.062	0.082	0.075	0.034
	(1.07)	(1.36)	(1.25)	(0.54)
Low PSE	0.246**	0.290**	0.278**	0.229*
	(4.14)	(4.07)	(3.89)	(3.12)
High PSE	0.187**	0.239**	0.176**	0.125**
	(2.61)	(2.67)	(1.90)	(1.30)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	526
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 11b: Wage regressions for males in Quebec
	Specification
	I	II	III	IV
High School only	0.005	0.009	0.012	0.007
	(0.12)	(0.21)	(0.25)	(0.15)
Low PSE	0.177**	0.148**	0.135**	0.120**
	(4.21)	(2.85)	(2.56)	(2.23)
High PSE	0.161**	0.119	0.069	0.065
	(2.34)	(1.43)	(0.79)	(0.74)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	601
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 11c: Wage regressions for males in Ontario
	Specification
	I	II	III	IV
High School only	0.090*	0.091*	0.095*	0.087
	(1.68)	(1.65)	(1.71)	(1.55)
Low PSE	0.272**	0.260**	0.248**	0.233**
	(4.74)	(3.95)	(3.71)	(3.42)
High PSE	0.241**	0.209**	0.181**	0.157*
	(3.74)	(2.65)	(2.17)	(1.86)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	853
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 11d: Wage regressions for males in Western Canada
	Specification
	I	II	III	IV
High School only	0.053	0.101**	0.122**	0.100**
	(1.48)	(2.73)	(3.21)	(2.64)
Low PSE	0.161**	0.278**	0.309**	0.292**
	(3.99)	(5.78)	(6.31)	(5.98)
High PSE	0.204**	0.359**	0.378**	0.352**
	(4.13)	(5.87)	(5.83)	(5.43)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	853
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 11e: Wage regressions for males in Canada (all regions)
	Specification
	I	II	III	IV
High School only	0.050**	0.070**	0.076**	0.062**
	(2.18)	(2.95)	(3.15)	(2.60)
Low PSE	0.190**	0.216**	0.217**	0.199**
	7.69)	(7.37)	(7.32)	(6.67)
High PSE	0.188**	0.218**	0.193**	0.166**
	(6.06)	(5.65)	(4.74)	(4.06)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	2,906
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

For males in Atlantic Canada, there are no significant wage difference between high school drop-outs and those with high school only. Graduating from a low PSE is associated with a significant wage gain, around 25%, and this wage gain is relatively stable across regression specifications. Similar to the finding in Table 9a, those with high PSE have lower average wages than those with low PSE but the difference is not statistically significant. In fact, in the most general regression specification, there is no significant wage difference between high school drop-outs and those with high PSE. As already discussed, this is likely due the nature of the sample which consists of respondents that are between 22 and 24 years old in December 2003. Moreover, a substantial fraction of the sample is still enrolled in some form of PSE at this date. In Quebec and Ontario, a pattern similar to that in Atlantic Canada is observed although the wage difference between high school drop-outs and those with low PSE is smaller in magnitude in Quebec than in the other two regions. Finally, for Western Canada, the estimates from the most general specification suggest significant wage difference between high school drop-outs and those with high school or more. Further, for this region, the effect of education on wages is the largest among all regions.

Table 12a: Wage regressions for females in Atlantic Canada
	Specification
	I	II	III	IV
High School only	-0.055	-0.065	-0.104	-0.062
	(0.63)	(0.74)	(1.22)	(0.70)
Low PSE	0.112	0.086	0.007	0.047
	(1.28)	(0.97)	(0.08)	(0.53)
High PSE	0.435**	0.368**	0.223**	0.222**
	(4.88)	(3.81)	(2.31)	(2.22)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	518
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 12b: Wage regressions for females in Quebec
	Specification
	I	II	III	IV
High School only	0.088	0.082	0.103*	0.102*
	(1.51)	(1.35)	(1.69)	(1.69)
Low PSE	0.355**	0.342**	0.310**	0.313**
	(6.28) **	(5.50) **	(4.92) **	(4.97) **
High PSE	0.459**	0.416**	0.321**	0.329**
	(7.26)	(5.51)	(4.14)	(4.16)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	611
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 12c: Wage regressions for females in Ontario
	Specification
	I	II	III	IV
High School only	0.155**	0.178**	0.172**	0.193**
	(2.28)	(2.58)	(2.50)	(2.76)
Low PSE	0.285**	0.317**	0.288**	0.312**
	(4.23)	(4.34)	(3.95)	(4.22)
High PSE	0.399**	0.423**	0.364**	0.381**
	(5.70)	(5.07)	(4.27)	(4.41)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	833
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 12d: Wage regressions for females in Western Canada
	Specification
	I	II	III	IV
High School only	0.124**	0.113**	0.119**	0.118**
	(2.30)	(2.06)	(2.15)	(2.11)
Low PSE	0.233**	0.215**	0.213**	0.216**
	(4.23)	(3.56)	(3.47)	(3.47)
High PSE	0.439**	0.405**	0.395**	0.396**
	(7.53)	(5.84)	(5.57)	(5.53)**
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	948
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

Table 12e: Wage regressions for females in Canada (all regions)
	Specification
	I	II	III	IV
High School only	0.110**	0.109**	0.116**	0.114**
	(3.49)	(3.41)	(3.67)	(3.58)
Low PSE	0.271**	0.267**	0.253**	0.253**
	(8.64)	(7.82)	(7.47)	(7.43)
High PSE	0.418**	0.392**	0.347**	0.343**
	(12.53)	(9.90)	(8.67)	(8.52)
Specification includes controls for:
Time since last in school and age	No	Yes	Yes	Yes
High School grade averages and Scholastic abilities	No	No	Yes	Yes
Family background	No	No	No	Yes
Sample size:	2,805
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

The entries in Tables 12a-12d show the effect of education on wages for females across the four regions (Table 12e shows results when data is aggregated across regions). Generally, the wage effect is larger than for males. In all regions, except Atlantic Canada, there are significant wage increases associated with increases in educational attainment. The wage difference between low PSE and high PSE is small in Quebec but sizeable in Ontario and Western Canada. Overall, consistent with the findings in Tables 9a and 9b, the effect of education on wages is substantially higher for women than for men. This finding is also consistent with findings in other studies, see for example Hansen (2006), Burbidge et al (2003), Ferrer and Riddell (2002). The fact that a substantial fraction of the sample is still enrolled in school in December 2003 may introduce a statistical problem to the analysis of the effect of education on wages. For instance, as mentioned above, this will be the case if those who are still enrolled in school are also endowed with higher unobserved labour market skills than those who are working. In an attempt to deal with this potential selection problem, estimates from selection corrected wage regressions are presented in Tables 13a and 13b. The estimates were obtained by specifying a likelihood function that models the probability of not being enrolled in school (and instead working) jointly with wages for those who are working. If, as hypothesized, unobserved labour market skills are correlated with the probability of being enrolled in school, we would expect a negative correlation between the idiosyncratic error terms in the schooling equation and in the wage equation. If the assumptions made in the selection corrected model specification are correct, the resulting estimates will consistently describe the effect of education on wages among those who are working.⁹

Table 13a: Selection corrected wage regressions for males
	Atlantic	Quebec	Ontario	Western	All regions
High School only	-0.031	0.036	0.094	-0.032	0.007
	(0.48)	(0.71)**	(1.52)	(0.76)	(0.25)
Low PSE	0.126	0.162**	0.195**	0.124**	0.122**
	(1.61)	(2.82)	(2.39)	(2.12)	(3.53)
High PSE	-0.016	0.212**	0.150	0.104	0.086*
	(0.87)	(2.17)	(1.52	(1.33)	(1.81)
Correlation	-0.092	-0.073	-0.176**	-0.239**	-0.155**
	(1.17)	(1.06)	(2.03)	(3.01)	(3.62)
Specification includes controls for:
Time since last in school and age	Yes	Yes	Yes	Yes	Yes
High School grade averages and Scholastic abilities	Yes	Yes	Yes	Yes	Yes
Family background	Yes	Yes	Yes	Yes	Yes
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

The estimates for males shown in Table 13a suggest smaller effects of education on wages than those found in Tables 11a-11e. Focusing on the results for all regions, the estimates suggest a wage difference between high school drop-outs and graduates from a low PSE program of around 12%, down from 20% in the uncorrected regression results. Further, the wage effect of a high PSE program is less than 10%, down from 17% in the uncorrected case, and only marginally significant. As previously discussed, these results would be expected if the effect of education is positively correlated with ability and those with higher ability are also more likely to be enrolled in school in December 2003. This means that the uncorrected estimates in Tables 11a-11e are exaggerating the effect of education on wages among those who work. Finally, for all regions the correlation between the idiosyncratic error terms in the schooling equation and in the wage equation is negative and precisely estimated for all regions except Atlantic Canada and Quebec.

Table 13b: Selection corrected wage regressions for females
	Atlantic	Quebec	Ontario	Western	All regions
High School only	-0.004	0.085	0.159**	0.097	0.084**
	(0.04)	(1.29)	(2.03)	(1.58)	(2.31)
Low PSE	0.094	0.251**	0.220**	0.121	0.153**
	(0.92)	(3.64)	(2.38)	(1.60)	(3.67)
High PSE	0.278**	0.307**	0.242**	0.208**	0.224**
	(2.48)	(3.44)	(2.19)	(2.36)	(4.48)
Correlation	-0.089	-0.217**	-0.128	-0.168*	-0.158**
	(0.98)	(3.24)	(1.47)	(1.83)	(3.49)
Specification includes controls for:
Time since last in school and age	Yes	Yes	Yes	Yes	Yes
High School grade averages and Scholastic abilities	Yes	Yes	Yes	Yes	Yes
Family background	Yes	Yes	Yes	Yes	Yes
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

The selection corrected estimates for females are presented in Table 13b. As for males, a comparison between the estimates in this table with the corresponding estimates in Tables 12a-12e suggests that the uncorrected estimates exaggerate the effect of education on wages. However, the wage differences across educational levels are still larger for females than for males. Further, as for males, the correlation between the idiosyncratic error terms in the schooling equation and in the wage equation is negative.

To summarize, the results suggest that there are significant wage gains from graduating PSE in Canada for females. Female graduates from universities at levels of at least a Bachelor's degree have average wages that are around 22% higher than those of high-school drop-outs. The results also indicate that wages among PSE graduates are higher than wages among those with high school only. For males, the wage gain associated with PSE is smaller. It should be noted that the relatively low wage gains associated with PSE may in part be due to the fact that the sample consists of young individuals even though the regression specifications control for time since last in school.

4.2 Wage differences across major field of study

Previous work, e.g. Hansen (2006), Drewes (2006), Boothby and Rowe (2002) have documented substantial wage or earnings differences across different fields of study. In order to assess the importance of field of study on wages among PSE graduates, wage regressions were estimated separately for males and females, but aggregated across regions. The results are presented in Table 14, which shows estimates associated with eight different fields of study with business serving as the excluded category.¹⁰ Thus, the entries show average wages given a specific major field of study relative to students whose major field of study was business. Most of the estimates in Table 14 are not statistically significant suggesting that early wages (the samples were restricted to those who had graduated from a valid post-secondary program between January 2000 and November 2003) are relatively homogenous. However, wages for those who graduated from studies focusing on arts or communications technologies are significantly lower than wages among business students. For males, graduates from humanities/social sciences/education also have significantly lower wages than business students.

Table 14: The effect of major field of study on wages
	Males		Females
Major field of study	Estimate	T-stat	Estimate	T-stat
Arts/Communications Technologies	-0.134**	(2.14)	-0.164**	(3.42)
Humanities/Social sciences/Education	-0.197**	(3.98)	-0.030	(1.01)
Physical and Life sciences/Technologies	-0.039	(0.48)	-0.070	(1.29)
Math Sciences/Engineering/Agriculture	-0.027	(0.61)	-0.119**	(2.91)
Health	-0.039	(0.50)	-0.009**	(0.24)
Personal and transportation services	0.031	(0.52)	-0.064	(1.43)
Other	0.065	(0.79)	-0.011	(0.14)
Sample size:	926		1454
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Major field of study refers to respondents' first main field of study or specialization. Sample is limited to respondents who had graduated from a valid post-secondary program by December 2003. Respondents who completed/attended PSE before Jan 2000 were removed as the field of study was not determinable. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level. Regression model includes controls for time since last in school, age, high school grade average, self-reported scholastic abilities (reading and math), family background, regional dummy variables and occupation. Finally, the excluded major field of study category is business.

4.3 Wage differences across occupations

The wage differences across different occupations for PSE graduates are shown in Table 15. As for the estimates described above, the specifications control for a wide range of observable characteristics. While the entries suggest small and mostly insignificant wage differences across major field of study, the entries in Table 15 indicate existence of significant wage differences across different occupations. Because of sample size constraints, occupational categories had to be aggregated into five groups: Natural sciences and health; Social sciences/education/government/art; Sales/services; Trades/primary production/processing and business/management. The estimates in Table 15 show the average wage in a certain occupation relative to that in business/management. The entries suggest that average wages are highest in Natural sciences and health, and this is true for both men and women. Lowest average wages are found in Sales/service occupations. Perhaps somewhat surprisingly, average wages in Social sciences/ education/ government/art are higher than those in business/management. Again, this finding may be in part due to the young nature of the sample. A similar finding was reported by Hansen (2006), who used data from the National Graduates Survey (NGS). As both data sources focus on new entrants in the labour market, it is likely that the lower average wages in business relative to Social sciences is due to differences in entry wages across these occupations.

Table 15: The effect of occupation on wages
	Males		Females
	Estimate	T-stat	Estimate	T-stat
Natural sciences/Health	0.275**	(5.93)	0.295**	(7.96)
Social sciences/Education/Government/Art	0.198**	(3.75)	0.167**	(5.35)
Sales/Service	-0.064**	(1.47)	-0.169**	(6.02)
Trades/Primary production/Processing	0.135**	(2.88)	0.103**	(1.70)
Sample size:	926		1454
Source: Ordinary Least Squares (OLS) regressions based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Major field of study refers to respondents' first main field of study or specialization. Sample is limited to respondents who had graduated from a valid post-secondary program by December 2003. Respondents who completed/attended PSE before Jan 2000 were removed as the field of study was not determinable. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level. Regression model includes controls for time since last in school, age, high school grade average, self-reported scholastic abilities (reading and math), family background, regional dummy variables and occupation. Finally, the excluded major field of study category is business.

4.4 The effect of education on labour force status

In addition to yielding higher wages, PSE may benefit students by reducing the risk of experiencing unemployment and by increasing time spent in employment. The discussion in Section 2.7 above, which was based on the entries in Table 6, showed that the proportion of respondents that were unemployed decreased with educational attainment up to low PSE but that it was similar for graduates from a high PSE program and high school drop-outs. It was hypothesized that this finding may be due to the lack of controls for differences in time since last in school. The entries in Table 16 are based on model specifications that include controls for a number of observable characteristics, including time since last in school. Specifically, Table 16 shows differences in the probability of being unemployed and of being neither employed nor unemployed (Home) across different levels of education. The estimated marginal effects suggest that the probability of being unemployed is significantly lower for PSE graduates than for high school drop-outs. The effect is larger for males than for females. Regarding the effect of education on home time, there is only a weak negative effect of education for males while it is significant and sizeable for females. However, for females the results suggest that completing high school, regardless if further schooling is acquired or not, significantly reduce home time.

Table 16: Education and labour market status
	Males		Females
	Unemployed	Home	Unemployed	Home
High School only	-0.018	0.008	-0.002	-0.148
	(1.03)	(0.52)	(0.37)	(3.80)
Low PSE	-0.035**	-0.028*	-0.010*	-0.198**
	(2.42)	(1.70)	(1.73)	(5.49)*
High PSE	-0.045**	-0.033)	-0.011*	-0.120**
	(5.05)	(1.27)	(1.76)	(6.18)
Specification includes controls for:
Time since last in school and age	Yes		Yes
High School grade averages and scholastic abilities	Yes		Yes
Family background	Yes		Yes
Regional controls	Yes		Yes
Sample size:	2,895		2,484
Source: Estimation based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Samples restricted to those who, in December 2003, were either employed, unemployed (looking for work) or not employed or unemployed (Home). Full-time students and self-employed workers were excluded. Entries show marginal effects with absolute values of T-statistics in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

4.5 The link between major field of study and occupations

In addition to an examination of early labour market outcomes for post-secondary graduates, this paper will also investigate the correspondence between occupational structure among post-secondary school graduates and the qualifications they obtained during their training. As discussed above in Section 2.6, there is a fairly strong correspondence between major field of study and occupation among respondents in this sample. While the association between occupation and major field of study in that section is informative, it does not provide information on what characteristics are important to improve the likelihood of a correspondence between field of study and subsequent occupation. To investigate this further, Table 17 presents results from a regression of the correspondence between major field of study and current occupation on selected observable characteristics. The dependent variable equals one if one of the following combinations of major field of study and occupation was observed:

Major field of study	Occupation
Humanities/Social sciences/Education	Social sciences/Art
Business	Management/Business
Physical and life sciences/Technologies	Natural sciences/Health
Math sciences/Engineering/Agriculture	Natural sciences/Health
Health	Natural sciences/Health
Personal and transportation services	Sales
Math Sciences/Engineering/Agriculture	Trades/Primary production/Processing

For any other combinations of major field of study and occupation, the dependent variable is set to zero. While this measure of correspondence is imprecise it enables a preliminary assessment of who is more likely to succeed in obtaining a job that, at least weakly, corresponds to the major field of study. It also provides an indication of how well the system for higher education in Canada provides skills that are demanded in the labour market. Among males, around 60% were coded as working in an occupation that matched their main field of study. The figure was somewhat lower for females, about 55%.

Table 17: Effects of observable characteristics on the correspondence between major field of study and occupation
	Males		Females
	Estimate	T-stat	Estimate	T-stat
High level PSE	-0.158**	(3.68)	-0.061*	(1.79)
Time since last in school	0.040**	(2.82)	-0.012	(1.03)
HS grade 90+	0.093	(1.36)	0.214**	(3.90)
HS grade 80-90	-0.015	(0.41)	0.083**	(2.79)
Atlantic Canada	0.046	(0.75)	0.083	(1.60)
Quebec	0.029	(0.67)	0.096**	(2.85)
Western Canada	0.032	(0.79)	0.023	(0.71)
Sample size:	926		1454
Source: Estimation based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Samples restricted to those who, in December 2003, were either employed, unemployed (looking for work) or not employed or unemployed (Home). Full-time students and self-employed workers were excluded. Entries show marginal effects with absolute values of T-statistics in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level.

The entries in Table 17 suggest that graduates from high PSE are less likely to work in an occupation that matches the main field of study. This effect is larger for males than for females. A possible reason for this finding may be that many shorter post-secondary programs provide more specific skills than longer programs at universities that may focus more on general skills. The gender difference in this effect may be due to less patience in job search among male graduates. The regression results also indicate that the match between field of study and occupation improves with time since last in school. Thus, recent graduates from a PSE program may find themselves in a transition period between school and work during which they take jobs that do not necessarily match the qualifications obtained in school. However, this effect is only significant for males. The probability of a match also improves with high school grade average, but this effect is significant for females only. While there are no significant regional variations in the match for males, the probability of a match for females between field of study and occupation is higher in Quebec than in other regions.

The above analysis described how certain characteristics are correlated with the match of skills acquired in school and those used in the labour market. The entries in Table 18 show the effect of such a match on wages and the effect is positive and significant for both men and women. Indeed, the estimates are virtually the same across gender and indicate that graduates that are working in occupations that correspond to their field of study earn on average about 18% more than otherwise similar graduates. The fact that a match between field of study and occupation yields higher wages (and productivity), combined with observations that many respondents have not yet obtained a job in occupation that matches their main field of study, indicates that many PSE graduates may be overqualified for their jobs. Interestingly, these numbers are similar to those reported by Li et al (2006), who found that 48% of young university graduates (under the age of 30) worked in a position for which they were overqualified.¹¹ However, as the regression results on the determinants of a "match" indicate, the probability of a match increase with time. This suggests that the nature of over qualification may be a transitory phenomenon.

Table 18 : Effects of the correspondence between major field of study and occupation on wages
	Males		Females
	Estimate	T-stat	Estimate	T-stat
Correspondence between major field of study and current occupation	0.175**	(6.53)	0.177**	(8.26)
Sample size:	926		1454
Source: Estimation based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples are restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. Major field of study refers to respondents' first main field of study or specialization. Sample is limited to respondents who had graduated from a valid post-secondary program by December 2003. Respondents who completed/attended PSE before Jan 2000 were removed as the field of study was not determinable. Absolute values of T-statistics are shown in parentheses. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level. Regression model includes controls for type of post-secondary education, time since last in school, age, high school grade average, self-reported scholastic abilities (reading and math), family background, and regional dummy variables. See text for details on how the correspondence between major field of study and current occupation was defined.

4.6 School-to-work transitions, school interruptions and wages

The last set of regression results, presented in Table 19, show the effects of having experienced unemployment during the first year after leaving school on subsequent wages. For both females and males, the effect of such an experience is negative and significant. The negative effect is larger for females than for males, and the magnitude of the coefficient for women suggests that the average wage for those who were unemployed for at least one month during the first year after leaving school is about 26% lower that the average wage of similarly endowed women who did not experience any unemployment during that transition period. For males, the effect is -15%. However, it is possible that the negative effect of being unemployed diminish with time since last in school. To test this, the unemployment indicator was interacted with a measure of time since last in school (measured in months). The interaction effect is positive and significant for females (0.004) and positive but insignificant (0.001) for males. Thus, for women, the negative effect of being unemployed is reduced by approximately 5% per year so that the negative effect has disappeared after 5 years. However, for males, the negative effect of early unemployment is persistent and does not diminish with time.

Table 19 also presents the effects of the duration of school interruption between secondary school and post-secondary school on subsequent wages. This variable is set to zero for all workers who did not attend any PSE. The effect is negative and significant for females and negative and insignificant for males, suggesting that moving straight from secondary school to PSE carries a significant, but small, wage premium. The measure for school interruption was also interacted with the type of PSE completed (low or high) and the interaction estimates suggest that the wage penalty associated with a school interruption is only significant for females who, as of December 2003, had not completed any PSE. Thus, it is possible that the negative estimate for school interruption on subsequent wages is not reflecting a causal effect of such interruption on wages but instead that interruption is correlated with unobserved characteristics that influence wages.

Table 19 : Effects of unemployment and school interruptions on wages
	Females		Males
	Estimate	Std err	Estimate	Std err
Unemployed	-0.263**	0.057	-0.147**	0.059
Interacted with time since last in school	0.004**	0.002	0.001	0.002
Duration (in months) of school interruption	-0.005**	0.002	-0.003	0.002
Interacted with Low PSE	0.002	0.002	-0.003	0.002
Interacted with High PSE	0.009**	0.004	0.004	0.006
Source: Estimation based on survey data from the oldest cohort in Youth in Transition Survey (YITS), cycle 3. Note: Dependent variable is the log of wage per hour. Samples are restricted to those with positive earnings per hour information from job held in December 2003. Full-time students and self-employed workers were excluded. * indicates statistical significance at the 10%-level while ** indicates statistical significance at the 5%-level. Regression model includes controls for type of post-secondary education, time since last in school, age, high school grade average, self-reported scholastic abilities (reading and math), family background, and regional dummy variables.

4.7 Regional mobility among PSE graduates

The analysis so far has studied different aspects of labour market outcomes in four regions: Atlantic Canada, Quebec, Ontario, and Western Canada. Each respondent region refers to the region of residence at the time of the Cycle 3 interview. However, some of the PSE graduates reside in a region that is different from the region in which they acquired their degree. For instance, among those PSE graduates that were residing in Quebec in 2003, 4% had obtained their credentials in another Canadian province. A somewhat higher figure, 8%, is recorded for respondents residing in Ontario. For respondents in other provinces, the figures are similar to those for Ontario. Thus, inter-provincial migration is relatively rare, even in this highly mobile sub-population.

In order to assess the wage effects of such migration, an indicator was created that took on the value one if the respondent acquired his/her PSE in the same province as he/she resided in at the time of the Cycle 3 interview, and it equaled zero otherwise. When inserted into wage regression equations, the coefficients associated with this indicator were positive but never statistically significant. This result was found for both men and women as well as when data was aggregated across regions and gender.

⁹ The estimates may be sensitive towards the normality assumption invoked in the selection corrected model. Another issue surrounding these types of models is identification, and in this paper identification is obtained through functional form assumptions. Alternatively, one could obtain identification by searching for variables that explain school enrollment decisions but not wages.
¹⁰The regression specifications include controls for similar observable characteristics as the specifications in Tables 13a and 13b but also include controls for region and occupation.
¹¹ Li et al (2006) used data from the Survey of Labour and Income Dynamics and defined respondents to be overqualified if they worked in an occupation that required at most a high school education. Similar to the findings in this paper, they also report that overqualification falls with time.

Primary navigation (left column)

Footer

Date Modified:: 2008-01-24

Top of Page

Important Notices