Literacy and labour market outcomes: self-assessment versus test score measures

Abstract

This paper looks at the determinants of literacy and the relation between literacy and labour market outcomes while focusing on comparisons of self-assessment versus test score measures of literacy. The test score measure performs considerably better than the self-assessments when literacy is treated as an outcome variable in terms of the overall fit of the model and the specific coefficient estimates, with the self-assessments sometimes actually generating wrongly signed parameters. The test score measure also performs much better as an explanatory variable in the employment models, with the self-assessment variable generating significant underestimates of the effects of literacy on the probability of being employed. Finally, the test score is also superior in the income models, although the self-assessment measure is at least a reasonably good performer in this regard, suggesting that the main results reported in much of the existing literature (based on such measures) should perhaps be taken as good representations of the true underlying relationships.

Acknowledgements

The authors are grateful to the Centre for Education Statistics at Statistics Canada for support for this research, and to Mike Charette, Scott Murray and William Gillen for their helpful comments.

Notes

Each individual in the survey was asked: ‘On a scale of 1–5, 1 being poor and 5 being excellent, how would you rate your reading and writing skills in English (French)?’

While ERROR could theoretically range from −4 to 4 it actually ranges from −3 to 4. The distribution of ERROR is:

Table

The total number of observations used for the English and French calculations is less than the other two comparison groups because individuals with a first language either than one of these were deleted from these groupings.

The μ values are generated by the estimation package (LIMDEP v.7) used to estimate the models. The dependent variable y _i is observed and takes some value from 0 through 4. The value taken by y _i depends on an unobserved stochastic variable not bounded above – hence there is no μ₄ involved. μ₀, μ₁, μ₂, μ₃ are unknown parameters.

Putting these results together with those for education reported just above might at first seem contradictory: females tend to underestimate the effects of education on their literacy levels more than men, whereas males tend to underestimate their skill levels in an overall, ceteris paribus manner. These findings would, therefore seem to suggest that men are generally ‘humble’ than women (at least with respect to their self-evaluated literacy skills), but that these ‘overall’ effects are counteracted by the effects of various other factors (such as education), thus implying a variation in the gender differentials along these various dimensions in a way which is not captured in the gender pooled models.

When discussing regional differences in literacy, less emphasis should be placed on the French (first language) sample, since 91% of those individuals reside in Quebec. To a lesser extent the same is true for the Foreign Born equation, of whom 52% live in Ontario.

Some authors employ the IV technique to deal with the potential endogeneity of literacy in an earnings equation (Charette and Meng, 1994), a few present both OLS and IV estimates (Chiswick and Miller, 1995, Table 5), however most research simply uses OLS estimates while controlling for selectivity into work. Following most research in this area estimate incomes using the latter approach are used here.

The LSUDA data base uses a class variable for age, with the upper limit being 69 years. Elimination of individuals who did not have positive weeks worked in the previous 12 months will eliminate retired people from the sample.

One reason why this result may occur is because many immigrants reside in ‘linguistic enclaves’ where they do not have to learn the dominant language(s) of their host country very well. Within this context, weak tested literacy may not be a significant deterrent to economic success.