The Treasury

Global Navigation

Personal tools

Treasury
Publication

Adult literacy and economic growth - WP 04/24

5.3  Literacy and employment

5.3.1  Concepts and methods

A number of the studies of literacy skills and earnings outlined above also look at the effects of people’s literacy skills on their labour force status. In considering labour force status, people are typically classified as being either employed (part-time or full-time), unemployed (that is, actively seeking paid employment), or not in the labour force. People who are either employed or unemployed are considered to be in the labour force.

Empirical studies examine the effect of a change in literacy score on the probability of being employed or, alternatively, the probability of being employed full-time, unemployed, or in the labour force. Since employment is a binary variable (people are either employed or they are not) studies typically use logit or probit models, where the probability of being employed (or unemployed) is a function, bounded by 0 and 1, of a person’s literacy score and other control variables. These control variables are typically those which are also used in the earnings regressions described above. Raw coefficients from logit and probit models are difficult to interpret, however, so studies usually (although not always) translate these into the effect on employment of a unit increase in the corresponding variable. In particular, the following section is concerned with the difference in the probability of being employed (or unemployed) between people who differ by one ‘unit’ of literacy.

5.3.2  Results of studies

The studies presented in Appendix 1 show that literacy has a persistent, positive and statistically significant association with people’s labour force status. People with greater literacy skills are more likely than people with weaker literacy skills to be employed, even after taking account of other observed factors. People with greater literacy skills are also more likely to be employed full-time, more likely to be in the labour force and, not surprisingly, are less likely to be unemployed.

As with studies of literacy and earnings, the measure of literacy (prose, document or quantitative) included in the regressions makes little difference.[27] Those studies which both do and do not control for educational attainment find that including education as a control variable reduces the employment differential 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 employment.

Using the New Zealand IALS data, Maré and Chapple (2000) show that a 10% increase in the average of the three literacy scores raises the probability of a male being employed by 1.2 percentage points (p.p.) and raises the probability of a female being employed by 2.1 p.p.. Maré and Chapple also examine the effect of literacy on Maori and non-Maori employment prospects separately. They find that a 10 percent increase in literacy score raises Maori employment chances by 3 p.p., compared to only 1.5 p.p. for non-Maori.

Using IALS data for the United Kingdom, McIntosh and Vignoles (2001) find that men with Level 2 prose literacy skills are 9.0 p.p. more likely to be employed than men with Level 1 prose literacy skills; the corresponding figure for women is 13.5 p.p.. North American studies indicate that an increase in literacy of one standard deviation increases the probability of employment by around 2-4 p.p. for men and by up to 8 p.p. for women. In most of the studies in Appendix 1, the impact of literacy on employment is greater for women than for men. Pryor and Schaffer (1999) consider that this because female labour supply is more sensitive to hourly earnings than male labour supply, and hourly earnings are positively related to literacy (as shown in the previous section).

None of the studies in Appendix 1 considers more than one country, or looks at more than one point in time. However, given the results reported in section 5.2.2, it would be safe to assume that the effect of literacy on employment does differ across these two dimensions.

5.4  A note about birth cohort studies

IALS, and other similar surveys of literacy, are limited in the amount of information they collect about individuals. In particular there are no good measures in IALS of an individual’s innate abilities, childhood environment, family background and socioeconomic status (apart from parents’ education levels), personality, attitudes and ‘soft’ skills such as sociability and ability to meet deadlines. These factors might be correlated with both literacy skills and earnings (or employment), and their omission from almost all of the analyses reported in Appendix 1 might mean that the studies overstate the earnings (or employment) premium associated with increased literacy skills. Since longitudinal birth cohort studies generally do include this kind of information, the two birth cohort studies included in Appendix 1 warrant a separate mention.

The National Child Development Survey (NCDS) is a longitudinal study of people living in Great Britain who were born during one week in 1958. Information on the cohort has been collected from an early age and in 1995 a 10% sub-sample was tested on their basic literacy and numeracy skills. McIntosh and Vignoles (2001) use this data to regress hourly earnings against literacy and against numeracy, including as control variables the socioeconomic status of respondents’ parents and the results of reading and mathematics tests undertaken at ages 7 and 16. McIntosh and Vignoles refer to these as tests of ability, but it is not clear that they measure the individuals’ innate ability, especially the age 16 tests. As the authors admit, controlling for age 16 test scores almost certainly means that the model is measuring the effect of changes in literacy skill between ages 16 and 37, although this itself is useful as a way of estimating the potential impact of adult interventions to improve literacy.

Table 5 shows the increase in earnings and employment associated with more advanced literacy and numeracy skills, compared to having low skills, under a variety of different specifications. Adding more controls to the model progressively decreases the returns to medium and high level literacy and numeracy skills. In a number of the models the returns to medium skills are relatively small (considering they involve increasing skills by the equivalent of at least one IALS level) and not statistically significant, even at the 10% level. Models (c) and (f), in particular, are the only ones which control for educational attainment, but few of the coefficients in these models are statistically significant.

Table 5 – Increase in earnings and employment over having low literacy or numeracy skills, NCDS
  Model (a) Model (b) Model (c) Model (d) Model (e) Model (f)
Increase in hourly earnings (%)
medium literacy skills 14.8** 8.5** 2.6 7.1* 4.7 1.3
high literacy skills 28.2** - - 16.3** 13.4** 8.0*
medium numeracy skills 14.7** 10.8** 6.9* 8.9** 7.7** 5.7
high numeracy skills 33.2** - - 18.0** 14.8** 7.6*
Increase in the probability of employment (p.p.)
medium literacy skills 5.1** 3.4 -0.03 3.9 3.0 0.2
high literacy skills 6.9** - - 5.6* 4.7 1.0
medium numeracy skills 4.5** 4.5** 2.7 4.8** 4.2* 2.9
high numeracy skills 9.0** - - 7.6** 6.3** 4.0
Controls
Background   X X X X X
Age 7 ability       X X X
Age 16 ability         X X
Education level     X     X

** statistically significant at the 5% level

* statistically significant at the 10% level

- results not shown in the report

Background controls are for gender, ethnicity, parents’ education levels and social class and a measure of family financial difficulties.

Low literacy is the equivalent of Level 1 in IALS, medium literacy ≈ Level 2 and high literacy ≈ Levels 3-5.

Low numeracy is the equivalent of Levels 1-2 in IALS, medium literacy ≈ Level 3 and high literacy ≈ Levels 4-5.

Source: McIntosh and Vignoles (2001).

Machin, McIntosh, Vignoles and Viitanen (2001) extend this analysis of NCDS data to take account of individuals’ attitudes and soft skills, as measured at age 16 and at age 37.[28] The main focus of the paper, however, is to regress earnings and employment against age 16 test scores. The age 16 test scores do not measure literacy and numeracy skills as they are usually conceived,[29] and using them in this way is a departure from the previous work by McIntosh and Vignoles. Presumably the authors were compelled to use this specification because it gave them a much greater sample size to work with (they weren’t restricted to the 10% sample of 37-year-olds). In a secondary analysis, Machin et al do regress hourly earnings against ‘real’ literacy as measured at age 37, but in all cases control for age 16 test scores. Therefore what is being measured is the impact of improvements in literacy and numeracy between ages 16 and 37. Few of the coefficients on literacy and numeracy are significant, however, under this specification.

The Dunedin Multidisciplinary Health and Development Study (DMHDS) also contains information on literacy and employment outcomes, at least for young adults. The DMHDS is a longitudinal study of a birth cohort of around a thousand children born in Dunedin in 1972 and 1973. Using DMHDS data, Caspi, Entner Wright, Moffitt and Silva (1998) find that poor reading achievement at age 15, as measured by the Burt word-recognition reading test, predicts later unemployment. After controlling for a range of individual, family and school variables, measured when participants were aged 15,[30] young people with low reading skills were 12.1 p.p. more likely than young people with high reading skills to be unemployed between the ages of 15 and 21, and averaged 1.7 more months of unemployment when unemployed. Some of this effect was due to more people with better reading skills staying longer in school and gaining more qualifications. In other words, young people with poor reading skills were at risk for unemployment, in part, because they left school at an earlier age. Even after accounting for this, however, there remained a direct impact of low reading skills on unemployment in later adolescence.

Notes

  • [27]Again, when literacy and numeracy are both included in the same regression, the coefficient on one measure is usually driven down to an insignificant level.
  • [28]The age-16 survey asked respondents about their attitudes to school, collected their official attendance record in school, and asked their teachers and parents about the respondents’ personality, ability to get on with others, and propensity for anti-social behaviour. At age 37, respondents were asked about their people skills, ability to trust others, tendency to argue, attitudes towards achievement, need for control, and caring skills.
  • [29]The mathematics test at age 16 consisted of “31 multiple choice questions examining a range of topics from the school mathematics syllabus, covering areas such as geometry and algebra”. The reading test consisted of “35 sentences, each with one word missing, and requires respondents to select a word from a choice of five that is most suitable to complete the sentences”. These are not literacy and numeracy skills as they are usually conceived and, at least for the mathematics component, are not tests of basic skills as measured, for example, in Murnane et al (1995).
  • [30]These were gender, parent’s occupational status, achievement of School Certificate, family structure, family conflict, parental attachment, school involvement, delinquency, mental illness and poor physical health.
Page top