Only a handful of studies have focused on returns to job-related formal learning for less-educated learners. Light (1995) uses the American National Longitudinal Survey of Youth to investigate whether adults who return to school later in life have lower returns than those who complete their schooling in one uninterrupted phase. Using a sample of young men between the ages of 16 and 32 from 1979 to 1989, Light finds that by the end of a six-year period there is no difference between the wages of those who delay their schooling and those who obtain the same amount of education continuously. Jacobson, LaLonde and Sullivan (2003) estimate the impact of community college on the earnings of displaced workers (35 years or older) living in Washington State. They found that one year of community college increased earnings by about 8 percent for males and 10 percent for females. Using the British National Child Development Study, Jenkins and colleagues (2002) analyzed the outcomes of individuals who returned to school between the ages of 33 and 42. They found that men who left school with only low-level qualifications earn substantially more if they undertake a degree via lifelong learning. They did not find a similar effect for women. However, significant positive employment effects were found for both men and women. In particular, the acquisition of vocational qualifications later on in life was associated with a higher probability of having moved into the labour market by 2000 for men and women who were out of the labour market in 1991 and with a higher likelihood of remaining in work for women employed in 1991. These findings are especially significant because they are able to control for the measurement problem that statisticians call ‘selection bias’.12
In Canada, there are numerous studies on returns to education but only two studies have attempted to distinguish the impacts of initial education from the impacts of education obtained later in life. A recent Statistics Canada study (Zhang and Palameta, 2006) uses a sample drawn from the Survey of Labour and Income Dynamics (SLID: 1993-1998 and 1996-2001) to analyze the earnings gains of individuals who obtained higher educational credentials later in life. The results show that most men and some women who obtained a post-secondary certificate later in life enjoyed sizable wage and earnings gains. Most significantly for our purposes, male learners with an initial education of high-school or less actually received higher returns than their more educated counterparts (10 percent versus 6 percent for wages, and 9 percent versus 6 percent for earnings). Female learners with high-school or less also enjoyed higher wage gains than their more-educated counterparts (10 percent versus 1 percent). However, for both less-educated and more-educated women learners, wage gains did not translate into gains in annual earnings. The authors speculate that one reason for this finding is that women may have used the increase in wages to cut back on the number of hours that they worked at several different jobs and focus on one better-paying or more satisfying job. The study asks whether these gains come from switching to a better job or staying in the same job but getting paid more money. Interestingly the data show that for less-educated men, only ‘job stayers’ and not ‘job switchers’ report significant wage gains. For less-educated women, both ‘job stayers’ and ‘job switchers’ report significant wage gains but these gains are only significant for the ‘job switchers’.
12 The term “selection bias” is used to refer to the problem that individuals who undertake lifelong learning are not a random subset of the population. Learners may differ from non learners on measured characteristics such as age and gender. Researchers address the problem of measured differences by including these differences in their statistical models. But learners may also differ from non-learners on other characteristics such as motivation and skill that are not usually measured in standard data sets. This is a problem because individuals who take training may be more motivated and/or more skilled than individuals who do not take training. Therefore, they may be likely to have successful labour market outcomes regardless of whether or not they took training. Moreover, encouraging less motivated or less-skilled individuals to take training may not have the desired effects if they do not have the necessary cognitive abilities to take advantage of the training opportunity. To address this problem, statisticians have devised a number of complex techniques that aim to separate out the effects of training from the effects of unobservable individual characteristics. A discussion of these techniques is outside of the scope of this paper. However, it is important to be aware of whether researchers control for selection bias or not. For a detailed discussion of these issues see Blundell, Dearden and Sianesi (2005) or Card (1999).