The Treasury

Global Navigation

Personal tools


Adult literacy and economic growth - WP 04/24

5  Individual returns to literacy skills

5.1  Introduction

This chapter looks at whether studies of individual differences in literacy and labour market outcomes provide evidence that increasing literacy skills would be good for the economy. It is one of three chapters looking at the economic effects of increased literacy skills (these were introduced in Chapter 3).

The IALS results presented in section 2.4 above show that New Zealanders with higher literacy skills earn more, on average, than people with lower skills and are more likely to be employed. A whole range of job-relevant skills, however, and not just literacy, affect a person’s earnings and employment. It may be that some of these other skills are associated with both literacy and earnings (or employment) and that these associations explain some, most, or all, of the apparent link with literacy. In other words, employers value people with literacy skills because these tend to go hand-in-hand with other valuable skills. Suppose, for example, that people with good literacy skills tend, for whatever reason, to be better at working in teams than people with lower literacy skills. Then at least part of the reason why people with higher literacy skills are paid more may be because they are believed to be better team members. Simply increasing a person’s literacy skills through participation in a training course, for example, without also improving their other work habits, might have much less of an effect on their earnings than expected, or indeed have no effect at all.

A number of studies have looked at the relationship between literacy and earnings (or employment), controlling for various potentially confounding factors. The results of these studies are presented in Appendix 1 and are discussed below. Most of these studies use data from IALS or similar literacy surveys conducted prior to IALS such as the 1992 National Adult Literacy Survey (NALS) in the United States. Other studies take their data from longitudinal studies, which follow participants over a period of time.

5.2  Literacy and earnings

5.2.1  Concepts and methods

Empirical studies of literacy and earnings typically adopt the following model specification:

Equation 1. log w = rl + ax + bx2 + cy + dz… + u

where w is a measure of earnings (either annual, weekly or hourly earnings), I is a measure of literacy skills (e.g. the prose score from IALS), x is age or years of work experience, and the other control variables y, z, etc include factors such as ethnicity, education, marital status, region, occupation/industry, disability status, immigration status and language spoken at home. The variable u is an error term representing unobserved factors other than s, x, y, z, etc., that affect w. Since log earnings are the dependent variable, coefficients (r, a, b, etc.) can be interpreted as the proportionate effect on earnings of a unit increase in the corresponding variable. In particular, the coefficient r can be interpreted as the percentage difference in earnings, all else remaining equal, between workers who differ by one ‘unit’ of literacy, for example one point on an IALS literacy scale.

The most interesting question about the choice of control variables in Equation 1 is whether or not to include measures of educational attainment. One view is that literacy is largely the result of education, although education also teaches other job-relevant skills and specialised knowledge. The effects of literacy on earnings should in this case be assessed by controlling for education, because education is a proxy for the other skills and knowledge learned in school, which might be correlated with literacy. On the other hand, if a child’s early literacy skills shape their future schooling attainment, and these literacy skills also persist into adulthood, then the full effects of literacy should be estimated by excluding schooling from the earnings regression. The truth, as usual, is probably somewhere in between these two extremes: literacy is likely to be both a cause and an effect of education. It is useful, therefore, to consider earnings equations which control for education, and those which do not control for education, as placing some bounds on the impact of literacy skills on earnings. Also, for the purposes of this report, it is important to consider the differences between improving a person’s literacy as an adult and improving their literacy as a primary school student.[22]

5.2.2  Results of studies

The studies presented in Appendix 1 show that literacy has a persistent, positive and statistically significant association with people’s earnings per hour, or per week. People with greater literacy skills are paid more, on average, than people with weaker literacy skills, even after taking account of other observed factors. Studies which are based on IALS, or similar literacy surveys, find that the measure of literacy included in the regressions makes little difference: using either prose, document or quantitative literacy, or the average of the three, gives a similar result.[23]

Those studies which both do and do not control for educational attainment find that including education as a control variable reduces the earnings premium associated with literacy. This suggests that literacy has both an indirect effect (since people with better literacy skills stay in formal education for longer) and a direct effect on earnings. Some studies also find a positive association between literacy and the quantity of work people do, e.g. the number of weeks they work in a year.

Using the New Zealand IALS data, Maré and Chapple (2000) show that a 10% increase in the average of the three literacy scores raises male annual earnings by 4.0% and female annual earnings by 5.1%.[24] To look at the effect on earnings per unit time, Maré and Chapple add controls for the normal hours each person works per week, weeks worked during the previous year, and whether people work full-time or part-time. They find, using these controls, that a 10% increase in literacy score increases male and female earnings by 5.0% and 3.2% respectively.

Also using IALS data, Denny, Harmon and O'Sullivan (2004) calculate the earnings benefits of literacy in 17 countries, including New Zealand. In New Zealand, a 10 point increase in the average literacy score in IALS is associated with a 2.4% increase in hourly earnings. Results differ considerably between countries, from a return of 1.3% in Germany to 3.3% in the Netherlands. Results for New Zealand were in the middle of this range. Blau and Kahn (2001) also find that the earnings benefits of literacy vary by country, although their sample does not include New Zealand. In most of the studies in Appendix 1, results differ between men and women although no obvious or consistent pattern emerges. The benefits of literacy also appear to change over time. Murnane, Willet and Levy (1995), for example, follow two cohorts of young people in the United States, separated by eight years. They find that the earnings premium associated with basic reading and mathematics skills, measured in the last year of high school, was much greater for the most recent cohort than for the earlier one.

While they vary across studies, countries and times, the results of the different studies are still fairly consistent. Across the studies, a 10-point increase in literacy, on the 500 point scale used in cross-sectional literacy surveys, results in an increase in earnings of around 1 to 5%. A 3% earnings return to a 10 point increase is a reasonable, middle-of-the-road assumption to make. Expressed in a different way, a one standard deviation increase on a literacy test results in an increase in earnings of around 4% to 20%.[25] By way of comparison, a year of schooling is typically associated with an earnings increase of around 7 to 10% a year.

As equation 1 makes clear, studies typically model a linear relationship between log earnings and literacy, so that 10 point increase in literacy score will necessarily have the same percentage effect on earnings at high levels of literacy as at low levels. Those studies which do test for nonlinearity, however, report a variety of results. Maré and Chapple (2000) look at whether literacy has a significantly larger earnings elasticity for people with low literacy skills but could find no support for this hypothesis. However, their log-log specification does imply that a 10-point increase in literacy score at low levels of literacy will be more highly rewarded than a 10-point increase at high levels of literacy.[26] Rivera-Batiz (1990) finds that a quadratic term involving literacy has a negative sign (indicating a stronger effect at low levels of literacy) but that this is only marginally significant. Denny et al (2004) allow for non-linearity in their results by using dummy variables for each quintile of the IALS score distribution instead of the literacy score itself. They find that in New Zealand the biggest increase in earnings comes from moving from the first to the second quintile of IALS score (i.e. at low levels of literacy). Other countries have different patterns of returns, however. In Great Britain, for example, the biggest jump in returns comes with moving from the fourth to the fifth quintile of literacy score.

Lee and Miller (2000) and McIntosh and Vignoles (2001) look specifically at the difference in earnings between people at various levels of literacy in IALS, although most of the coefficients reported in both these studies are not statistically significant, and should therefore be treated with caution. Lee and Miller report, using Australian data, that the biggest increase in earnings for men comes with moving from Level 1 to Level 2, but that the biggest increase for women comes with moving from Level 4 to Level 5. McIntosh and Vignoles report, for the United Kingdom, that the earnings premium for being at Level 2 of the prose literacy scale compared to Level 1 is 11.5% for men and 14% for women. The premium for being at Levels 3-5 drops for men to 9.5% but grows to 19.2% for women.

In some ways, though, it makes little sense to worry about whether the effects of literacy gains are higher at the bottom of the distribution than at the top. Osberg (2000) makes the point that a 10-point increase, or a 10% increase, refers to the literacy score rather than to literacy itself (the underlying concept). Literacy scores in IALS are essentially ordinal, not cardinal, numbers. A person with a higher score can be considered more literate than one with a lower score, but it is not possible, or meaningful, to say how much more literate they are.


  • [22]If the aim of policy is remedial, that is to improve literacy amongst adults, then it is too late to resurrect people’s school careers and the earnings benefits of increased literacy are best estimated by controlling for past schooling. However, if the aim of policy is to improve literacy in children’s formative years then the effects of this increase in literacy might well include the flow-on effects at school. It is therefore more appropriate to estimate earnings benefits without controlling for schooling.
  • [23]Some studies do treat literacy and numeracy as distinct skills. However, when they are included together in the same regression, as in Charette and Meng (1998), the coefficient on one of literacy or numeracy is usually driven down to an insignificant level. This is not surprising since these measures of literacy are, at least in IALS, highly correlated.
  • [24]In contrast to the other studies based on IALS or similar surveys, Maré and Chapple use the log literacy score as their dependent variable. Their coefficients therefore represent the percentage increase in wages associated with a percentage increase in literacy score.
  • [25]Standard deviations on IALS-type tests typically range from 40 to 60 points, but the range given here also includes the results of longitudinal studies which use different types of literacy tests.
  • [26]A 10-point increase from 100 points (a low score) is a 10% increase; a 10-point increase from 400 points (a high score) is only a 2.5% increase.
Page top