Section 3: Literature review

Previous research has found that certain demographic characteristics are associated with differences in returns to investment in education. One of these is gender. Gittel et al. (2005) suggest that there are numerous reasons why a woman working full-time who is similarly educated as her male counterpart would earn lower wages. They suggest that women offer less labour because of gender roles in family responsibilities – women are significantly more likely to take time out of the labour market to care for children than men, either by leaving the labour market entirely or by choosing to work part-time when their children require more care. They also suggest that women tend to be concentrated in 'low-paying' fields of study such as the humanities and less concentrated in higher demand fields such as engineering. Similarly, McNeil and Lamas (1987) find that women are more likely to take time out of the labour market which leads to them having fewer years of employment with the same employer – and thus less likely to earn higher salaries and promotions. They also find that the gap in earnings for similarly-educated men and women can be explained in part by gender differences in occupational structure, with wages tending to be lower in female-dominated occupations.

An individual's age can also have an impact on returns to education. The relationship between age and earnings is two-fold: first, age reflects the number of potential years of experience in the labour market. Younger workers typically have fewer years of working experience than their older counterparts and therefore tend to command lower earnings (Card 1999). Furthermore, at different stages in life, people may be more or less willing to participate in the labour market. Examples of this may be child-bearing years for women or pre-retirement years for older workers. Finally, according to the job-matching or information-based model, younger workers tend to have more frequent short-term employment spells at the start of their careers as they look for a good match between their skills and employers' needs (Riddell 2007). Such short-term employment spells would contribute to lower earnings for younger workers early in their careers.

Immigration status and location of study also play a role in earnings levels. According to the 2006 Census, one in five Canadians1 was born outside Canada – the highest proportion since 1931. This proportion was even higher in Ontario, where over a quarter – 28.2% – of Ontarians was born outside of Canada. The source countries for new immigrants have also changed over time. Among the more than 1.1 million immigrants who arrived between 2001 and 2006, almost 6 in 10 were born in Asian countries, including the Middle East. In Ontario, 63.9% of new immigrants were from this region. In contrast, in 1971, 61.6% of newcomers to Canada were from Europe while 63.1% of newcomers in 1971 in Ontario were from this region. As a result of changes in immigrant source countries, the proportion of the foreign-born population who was born in Asia and the Middle East (40.8%) surpassed the proportion born in Europe (36.8%) for the first time in 2006 (Statistics Canada 2008a). This was also true in Ontario, where 40.5% of the foreign-born population was born in Asia and the Middle East, while 38.5% was born in Europe. One implication is that new immigrants are much less likely to have English or French as their mother tongue than previous generations of immigrants and large numbers have completed their schooling in their home countries, often in a language other than English or French.

There are many reasons why immigrants may experience lower returns to their credentials in the labour market. Bonikowska, Green and Riddell (2008), for example, find that the literacy-skills distribution is higher for the Canadian-born than it is for immigrants who completed all of their education abroad, noting that these differences in measured skills partly reflect proficiency in either English or French. They also find that lower literacy skill levels translate into lower earnings in the labour market. Finally, they note that part of the explanation for the earnings gap between immigrants and the Canadian-born is that immigrants' earnings reflect low, or even zero, returns to their foreign work experience. When only their Canadian work experience is taken into account, immigrants' earnings were more similar to those of the Canadian-born with the same years of experience.

A person's family situation may also play a role in labour market attachment. As previously noted, women in particular are more likely to choose to reduce their working hours while they have young children and this affects their employment earnings. Zhang (2009) finds that there is a sizable earnings difference between women who have children and women who do not. As well, this study reports that the impact on earnings of having a child was larger for postsecondary-educated women. Another way in which family situation can affect employment earnings is the impact of total family income, that is, if one member of the family is already making a fairly good salary, this might enable his or her partner to take a lower-paying (but possibly rewarding in some other way) job, to work part-time or to choose not to work at all. Hou and Myles (2007) find that, increasingly, individuals are tending to marry similarly-educated individuals (what they term 'homogamy'). This could lead to situations where highly-educated individuals voluntarily choose to have lower earnings, if their highly-educated mate already has high earnings.

Other characteristics that have been shown to affect employment earnings are program level and field of study (Finnie 2001; Finnie and Frenette 2003; Walters 2004; OECD 2008). Earnings trajectories of university graduates tend to be higher than those of college graduates, who, in turn, have higher earnings trajectories than high school graduates and those with less than high school (Walters 2004; OECD 2008). With respect to field of study, college and university graduates with credentials in fine arts, for example, would have a significantly different set of skills than someone with an engineering diploma or degree and that would affect occupational options. Most studies find that graduates in the more general liberal arts programs (such as the humanities and social sciences) tend to do more poorly in terms of employment and earnings outcomes than do graduates in more applied fields (Walters 2004).

A number of studies have addressed the issue of highly-educated workers with low earnings in a European context. For example, this phenomenon has been studied in Austria (Fersterer and Winter-Ebner 2002) and Sweden (Korpi and Talin 2008). Researchers there conclude that developed economies have been creating more skilled workers than skilled jobs in recent years and, as a result, the supply of educated labour has outstripped the demand for it. In other words, they suggest that weak employer demand for more highly educated individuals may provide part of the explanation. Other work in Ireland suggests that the phenomenon of highly-educated low-earnings workers can be partially explained by a drop in the level of ability associated with a postsecondary education (McGuiness and Benett 2007). Their explanation is that, as access to postsecondary education has increased, greater variance has arisen in the ability levels of postsecondary graduates with the result that variance in returns to education are due to the fact that higher-ability graduates are able to find higher-paying jobs, while lower-ability graduates do not. These kinds of hypotheses are out of scope of this report, however, since the first concerns the nature of demand for skills in the labour market while the second requires information about the abilities of individuals.

To summarize, the literature suggests that there are a number of factors that may help explain why some college- and university-educated individuals are in low-earnings situations. These include gender, age, immigration status, labour market attachment (full-time versus part-time), field of study and occupation. The contribution of each of these factors is investigated using logistic regression analysis. Before turning to that analysis, however, the next chapter first describes the data sources and methodology that form the basis for the statistical analysis.


  1. This includes Canadian citizens (by birth and naturalization), landed immigrants and non-permanent residents and their families living with them in Canada. Non-permanent residents are persons who hold a work or student permit, or who claim refugee status. The census also includes Canadian citizens and landed immigrants who are temporarily outside the country on Census day.
Date modified: