The generation of literacy skills

View the most recent version.

Schooling and background effects
Age and generational effects
Post-school, work related literacy acquisition

Schooling and background effects

We turn now to using both IALS 1994 and IALSS 2003 to examine the sources of literacy. We do this using a series of regressions, where we again focus on Document literacy as our representative measure. More specifically, we use the log of the Document literacy score as our dependent variable so our estimated coefficients can be interpreted as showing impacts in terms of percentage changes in literacy. We begin with the much larger IALSS 2003, examining the impacts of family background and schooling on individuals' literacy scores.

Before presenting the estimation results, we begin by setting out a brief, heuristic model of literacy generation. The model will help to put our estimates in context as well as providing guidance in thinking about identification issues. Consider a simple model in which individuals start out at birth endowed with two key characteristics: their ability and parental resources. By parental resources, we mean something quite broad, incorporating both parental income and parental willingness and ability to support their children's education and literacy acquisition. Pre-school children begin to acquire literacy based on these fundamental characteristics (ability and parental resources). Once they enter school, these characteristics interact with characteristics of the school such as teacher quality, class size and the attitudes and abilities of peers. New additions to literacy with each year of schooling are then functions of ability, parental resources, school characteristics and the literacy level at the beginning of the period. These influences may interact in complicated ways. These additions continue until the legal school leaving age. After that point until the end of high school, students make a decision each year on whether to continue in school. That decision will be a function of ability, parental resources and school characteristics, again, but it is also likely to be a function of literacy acquired to that point. The more literate a student is, the less onerous they are likely to find school and, thus, the more likely they are to choose to stay an extra year. Finally, after high school, whether an individual continues to go to school will be determined by a combination of their own decision to apply to continue and the decision of the college or university on whether to admit them. The latter decision will likely be a function of the student's literacy as reflected in her grades. Thus, schooling and literacy are co-determined with extra years of schooling leading to increased literacy but increased literacy also leading to more years of schooling, especially after the legal school leaving age. Indeed, once we account for expectations, the inter-relation between the two may be even tighter. Individuals who do not expect to continue with school past the legal minimum may rationally under-invest in acquiring literacy skills while they are in school.

Once individuals leave school, literacy acquisition is likely more difficult. Literacy skills may be acquired on the job if they are needed for carrying out tasks at work but otherwise further acquisition would require active investment in non-work hours. Indeed, it seems quite possible that individuals could lose literacy skills after they leave formal schooling if those skills depreciate when they are not used.

We are interested in characterizing as many of the components of literacy generation as possible. In particular, we are interested in the relationship of literacy to parental resources since that relationship is fundamentally linked to notions of equity: to the extent that one generation's literacy hinges on the resources of the previous generation, differences in literacy can be seen as arising from characteristics beyond the control of the people involved. This is a paradigmatic case for redistributional policy in the views of philosophers such as Dworkin and Rawls. We are also interested in the relationship between formal schooling and literacy since this is a main channel through which we could hope to affect the literacy distribution. Finally, we are interested in whether literacy declines or improves after leaving school and how this process is related to characteristics of an individual's job. If literacy has a "use it or lose it" form then there may be good reason to adopt policies such as subsidizing firms to provide access to literacy maintenance activities such as allowing employees to return to school. Many of these parameters of interest reflect causal relationships that are difficult to establish definitively. We will make efforts to estimate the causal parameters where the data permit but much of what we will discuss is necessarily in the form of correlations rather than clear causal impacts.

In the OLS 1 column of Table 3, we present our simplest OLS regression: one in which the dependent variable is the log of individual Document literacy and the independent variables are age, age squared, years of schooling, years of schooling squared, and a gender dummy. Thanks to the substantial sample size, all the variables are statistically significantly different from zero at either the 1% or 5% level but this does not mean their actual impacts are sizeable. Thus, the estimates show that women have lower literacy than men (conditional on school and age) but only by 1.4%. Similarly, the age and age squared coefficients are highly statistically significant but together they imply that the impact of an extra year of age on literacy is actually -0.1% at age 30.¹ This finding that there is essentially no relationship between literacy and either age or experience is a key part of the discussion in Green and Riddell (2003). The one relationship that is economically substantial is the one between literacy and schooling. One extra year of schooling, evaluated when the individual already has 12 years of education, increases literacy by 3.2%. This is very similar to what Green and Riddell (2003) calculated using the IALS 1994.²

In the OLS 2 column of Table 3 we add variables on parental education and immigrant status to our OLS regression. Introducing these variables has virtually no impact on either the gender or schooling variable effects. However, including them leads to a near doubling in the age coefficient. Given that the coefficient on the age squared variable also becomes more negative, the net effect of age is still quite small. The parental education variables are jointly highly significantly different from zero but, perhaps surprisingly, the effect is found almost entirely at low levels of parental education. Having a parent (either mother or father) who is a high school drop out, decreases average literacy by between 3% and 4%. However, parental education beyond high school graduation has no further impact on literacy. Interestingly, not knowing a parent's education level (which is the case for approximately 8% of the sample) has a strong effect, being associated with approximately 6% lower literacy. While we just included this variable in order to allow us to keep the observations for which parental education is missing, it seems possible it is actually proxying for something real. For example, children who do not know a parent's education likely did not have a close relationship with that parent. Thus, the estimated coefficient may reflect the extent to which literacy is generated through direct parental involvement. Finally, having a father who is an immigrant has a mild association with literacy (increasing literacy by 1.5%), while having a mother who is an immigrant has no impact. We also tested specifications in which we included a set of parental occupation dummy variables, but these were never jointly statistically significant. In particular, a test of the hypothesis that the set of father's occupation dummy variables jointly had zero effects has an associated P-value of 0.13. The same test for mother's occupation has a P-value of 0.79. We also find that a dummy variable representing whether the individual's mother was working when the individual was 16 does not have a statistically significant effect. Overall, the results point to a surprisingly weak association between literacy and parental background. Only schooling seems to have a substantial impact on literacy generation.

As we discussed earlier, literacy and years of schooling are likely to be jointly determined. In that case, our OLS coefficient on schooling provides a biased estimate of the impact of schooling on literacy. We attempt to address this in two ways. First, biases may arise because of a correlation between literacy and schooling arising from unobserved ability. If high ability people do not view it as particularly costly to either acquire literacy or go to school then we could observe a strong, positive coefficient on schooling in our regression because years of schooling is proxying for ability rather than as a reflection of a causal impact of schooling on literacy. This problem can be addressed if we have a measure of ability since once we control for ability, any relationship between schooling and literacy cannot be due to an omitted ability term. Note, though, that many studies that try to control for ability (in, for example, earnings regressions) actually use scores on tests much like our literacy tests. What we would require is a test score from a very young age – before the process we are trying to study really begins. Since we don't have that, we instead try to proxy for ability using two variables that are plausibly related to it. In particular, in the third regression, we include a dummy variable equalling one if the person agreed or strongly agreed with the statement that they got good grades in math when they were in school and another dummy variable equalling one if the respondent agreed or strongly agreed with the statement that teachers often went too fast and the person often got lost. Either of these could plausibly be seen as proxies for innate ability. Both of these variables enter significantly, with people who claimed to have gotten good grades in math having 2.7% higher literacy and those who thought teachers went too fast having 2.6% lower literacy. However, including these variables has almost no impact on the other estimated coefficients. Most notably, their inclusion does not change the impact of schooling on literacy at all.

An alternative approach to the problem of identifying a causal effect is to find an instrumental variable for schooling and estimate our regression using two-stage least squares. An instrument is a variable which plausibly affects the right hand side endogenous variable directly but affects our dependent variable only through its relationship to the right hand side endogenous variable. Essentially, it allows us to get estimate effects based on the part of the variation in the right hand side endogenous variable that is correlated with the instrument and, hence, uncorrelated with the error term. In our case, this permits consistent estimates of the true impact of schooling on literacy providing two conditions are met: 1) the instrument truly does affect education; and 2) the instrument does not belong in the regression determining our dependent variable in its own right. We use the province where the individual resided when he or she was last in high school or primary school fully interacted with age as an instrument for schooling. The idea behind the instrument is that different levels of public resources applied to schooling in different provinces for different generations will lead to different schooling outcomes for otherwise identical individuals. On the other hand, we cannot see a reason for province of residence having an impact on literacy other than through changing schooling outcomes. In implementing this approach it is important that we also control for current province of residence in both the first stage (schooling regression) and the second stage (literacy regression). Province of current residence may be related to literacy if more literate individuals choose to migrate to provinces with a higher proportion of high literacy jobs and low literacy individuals chose to move to, for example, provinces with large numbers of resource jobs. In that case, to the extent that province of residence during high school and current province of residence are correlated, the province of residence would pick up this migration effect rather than the schooling effect we want it to capture. Controlling for province of current residence addresses this problem and means that we are really identifying the schooling effect from people who currently reside in the same province but were schooled in different provinces at different times.

The results from our two stage least squares estimation using province of residence during high school as an instrument is given in the IV column. We are unable to come up with a separate instrument for schooling squared. A standard approach to instrumenting for a squared term would be to use higher order terms in the instrument as instruments for the squared term. However, since our instrument is a dummy variable (or, more specifically, a set of dummy variables), the higher order terms equal the dummy variable itself and we are left without a separate instrument for the squared term. As a result, we estimate a specification that just includes the linear schooling variable. In the OLS 4 column, we present estimates from an OLS regression using the same specification for comparison.³

In the first stage (reduced form) regression, in which years of schooling is the dependent variable, the set of covariates include the full set of covariates from our main literacy regression: age, age squared, parental education, parental immigrant status and province of current residence. It also includes the set of dummies representing the province at time of high school. As expected, parental education is strongly positively related to years of schooling. Having parents who are immigrants also corresponds to having more years of schooling. The set of provinces at time of high school variables and their interactions with age are jointly highly statistically significant, indicating that the requirement that the instrument is a significant determinant of the endogenous variable is satisfied.⁴ The first stage estimates are presented in Appendix B.

In the OLS 4 column results, the estimated schooling effect is highly significant and of the order of magnitude we reported as the percentage impact of an added year of schooling on literacy when the individual has 12 years of schooling based on the quadratic specification. The estimate of the schooling effect presented in column IV, however, is approximately twice as large, implying even stronger schooling effects than we have estimated with OLS specifications. Interestingly, once we instrument for schooling, the parental background variables all become smaller in magnitude and lose statistical significance. This may indicate that the parental background effects we estimated earlier were actually disguised place-of-schooling effects or, at least, that the parental background and place-of-schooling variables are highly collinear. As in earlier specifications, gender and age continue to have small impacts on literacy. We do not present the coefficients corresponding to the provincial dummy variables for the sake of brevity but they show that the Atlantic provinces and Ontario have essentially similar literacy levels, Quebec has significantly lower literacy levels and the Prairies and BC all have significantly higher levels. Our main conclusion is that, if the assumptions underlying our instrument are correct, these results indicate that education has a strong causal effect on literacy and that schooling is the dominant determinant of literacy.⁵ To put the estimated effect in perspective, completing four extra years of schooling (e.g., moving from being a high school graduate to a university graduate) implies a 21% increase in literacy, based on the IV estimates. This would be enough to move the individual from the median to approximately the 90^th percentile of the literacy distribution in 2003.

Age and generational effects

One of the most striking results from our initial regressions is the lack of significance of the age variable. Taken at face value, this suggests that individuals neither lose nor acquire new literacy skills after they finish schooling. Essentially, literacy is acquired at school and then simply maintained thereafter. This finding is reinforced if we replace age and age squared by experience and experience squared in the first regression specification from Table 3.⁶ The coefficients on schooling, schooling squared and the female dummy are virtually identical to those presented in Table 3. The coefficient on experience is very small: 0.00051 with a standard error of 0.00045.

Interpreting the coefficient on age in a cross-sectional regression requires some care, however. As has been extensively studied in the literature on immigrant earnings, differentials between two age groups in a cross section could reflect a variety of possible combinations of true age and generational effects. Thus, while we are tempted to view the literacy level of 35 year olds in the IALS 2003 as a reflection of the literacy level the 25 year olds are likely to be at in 10 years time, we need to bear in mind that the 35 year olds come from an older generation and their observed literacy reflects a combination of any generational differential as well as any ageing effect. Only if there are no systematic differences across generations does the cross-sectional literacy–age profile reflect the true impact of ageing on literacy.

A more complete investigation of generational effects (often called cohort effects) and ageing effects requires the use either of true panel data or of at least two cross-sectional datasets constructed in such a way that we can follow "synthetic" cohorts through time. We make use of the IALS 1994 and the IALSS 2003 for this purpose. More specifically, in the public use version of the IALS 1994, we can observe a set of 10 year wide age groups for the respondents (i.e., ages 26 to 35, 36 to 45, 46 to 55, 56 to 65, and 65and over). Since we have a continuous age variable in the IALS 2003, we can construct age groups that correspond to the age people in these initial groups would be 9 years after IALS 1994 (i.e., 35 to 44, 45 to 54, 55 to 64, 65 to 74, and 75 and over). We call the set of people who were age 26 to 35 in 1994 and 35 to 44 in 2003, cohort 1 and number the remaining cohorts in ascending order from there. If both surveys provide random samples then each provides an unbiased estimate of the literacy distribution for the cohort at two different points in time and we can follow the progress of a given cohort over time. In part of our examination, we will further break the cohorts down by education. Linking across surveys would not provide unbiased estimates of the progress of cohort–education groups in this case if the educational composition of the groups being studied changes over time. Thus, we cannot say that the average literacy of 35 to 44 year old high school graduates in IALSS 2003 provides a consistent estimate of what happened to the 26 to 35 year old high school graduates we observe in IALS 1994 over the intervening 9 years if some of the initial high school graduates would be expected to have increased their education level. In that case, the average literacy in 2003 would reflect a combination of ageing effects and the systematic selection of people out of the high school graduates group. To avoid this problem, we focus on individuals over age 25, after which point changes in formal education across our broad categories are rare. It is also worth noting that the oldest cohort (made up of those over age 65 in 1994) is special in that its composition is likely to change over time because of deaths. As a result, while we include it in our analysis, we do not place much weight on results related to that cohort in our conclusions.

We begin with plots of the Document literacy densities from the IALS 1994 and IALSS 2003 broken down by age, education and cohort groups. Thus, in Chart 2, we present the density plots for individuals who are age 26 to 35 in 1994 and the individuals who are the same age in 2003. These densities correspond to the literacy for the youngest cohort we examine in 2003 and for the cohort just before them at the same age (observed in 1994). The density for the younger cohort (observed in 2003) has noticeably less spread, with both the left and right tails being thinner. In other words, the younger cohort experiences an improvement in literacy at the low end of the distribution but a worsening at the top end. This pattern is reflected in a 10^th percentile of 223 and a 90th percentile of 363 in 1994 compared to values of 248 and 354, respectively, in 2003. The reductions in inequality in both tails nearly cancel out, as reflected in a median value of 299 in 1994 as compared to 306 in 2003. Thus, one might conclude from a measure of central tendency that little has changed, whereas the tails reflect more substantial differences.

In Chart 3, we present the same kind of density comparison but for individuals aged 36 to 45. The outcome is much the same as in Chart 2 except that the relative improvement in the lower end of the distribution is not as great and the relative decrease at the top is decidedly larger than what we observed in Chart 2. In contrast, Chart 4 shows that for the 46 to 55 year olds, the results in 2003 are better across the distribution. This is important, in part, because observing different relative changes between 1994 and 2003 for different groups suggests that we are witnessing something real rather than just a difference in the tests in the two years. If all plots for all groups showed improvements at the bottom and declines at the top between 1994 and 2003 then the simplest explanation for the patterns would be that the test changed in such a way that it generated better results on the easy questions that will constitute most of the scores at the bottom but worse results on the harder tests that will define the shape of the top of the distribution. Finally, Chart 5 also shows improvements in 2003 for the 56 to 65 age group, though the 1994 and 2003 densities are very similar at the very top end.

We next examine the changes between the two years broken down by education group. Thus, in Chart 6, we plot the Document literacy densities in 1994 and 2003 for people with less than a high school diploma. The improvement at the bottom end of the overall distribution is also evident at the bottom of this distribution and is quite large. The 10^th percentile for this drop-out group rises from 145 in 1994 to 175 in 2003. At the other end, the differences are smaller. While the 95^th percentile declines between 1994 and 2003, the 90^th percentiles are actually identical in the two years. Of course, the 90^th percentile of their distribution is relatively low – being around the 70^th percentile of the overall distribution.

Even by the time we turn, in Chart 7, to high school graduates, however, there is clearer evidence of a worsening in the right tail of the distribution with a much smaller improvement in the left tail than was evident with the high school drop-outs.

Both the Non-university Post-Secondary and the University groups show, in Charts 8 and 9 respectively, declines in the upper part of their distributions, with the former group also showing some declines lower in the distribution as well. Both the Some Post-Secondary and the University groups register approximately 10% declines in their 90^th percentiles between the two years (from 373 to 338 for the Post Secondary and from 392 to 359 for the University group). Thus, in terms of within education group movements, the improvements at the bottom of the overall distribution are really only clearly evident in the high school drop-outs distribution while the declines at the top are mainly observed in the distributions for the top two groups.

The changes observed for the overall distributions may, in part, also stem from changes in the educational composition of the population between the two years. The replacement of older, less educated cohorts with younger, more educated ones shifts the educational distribution. Thus, the proportion of the population with a Bachelor of Arts rises from 17% in 1994 to 21% in 2003. This improvement undoubtedly plays a role in the improvement in the lower tail of the overall distribution. It is difficult to see, however, how this change could cause the declines at the top of the distribution. If a person of the same underlying ability was a high school graduate in the oldest cohort but a university graduate in the newest cohort then if we were to replace the former with the latter, one would expect the overall distribution to either improve (if university education increases literacy) or stay the same (if it does nothing). Even the change at the top of the distribution for university graduates is unlikely to be related to an expansion of the university educated pool. While the additional university graduates (those who would not have attended university in an earlier generation) are likely to be of lower ability than those who would have been university graduates in both periods, that would affect the lower not the upper part of the university educated distribution.

In Charts 10 through 13, we present density plots for each year for specific birth cohorts, as defined earlier. Note that this contrasts with Charts 2 through 5, where we examine the same people of the same age in the two surveys. The Chart 2 through 5 comparisons thus present comparisons across cohorts (at the same age) while the Chart 10 through 13 comparisons follow individual cohorts through time.

Chart 10 shows the plots for the cohort which was age 26 to 35 in 1994 and 35 to 44 in 2003. The plot indicates little difference at the low end of the distribution for the two years but a clear worsening at the top end. In fact, the 10^th percentile of the literacy distribution increases from 223 to 232 for this cohort while the 90^th percentile falls from 363 to 344. The cohort that was age 36 to 45 in 1994 shows a very similar pattern, plotted in Chart 11. In contrast, the densities for the cohort that was 46 to 55 show, if anything, an improvement from 1994 to 2003 (Chart 12). The 10th percentile for that cohort increased from 187 to 200 and the 90th percentile increased from 320 to 328 between the survey years. Again, this is important for considerations of the comparability of data since it shows that this cohort answers the full range of questions essentially as well (or slightly better) in 2003 as in 1994. The differences in patterns across cohorts would be very hard to rationalize simply as being due to more difficult top level questions being asked in 2003 than 1994. A more likely explanation of the patterns is that literacy skills deteriorate after leaving school but that the rate of deterioration declines with age.

Finally, in Charts 14 through 16 we plot densities following a particular education/cohort group. Here, we define wider cohorts in order to ensure enough observations. Chart 14 shows the plots for those with less than a high school education. There is clear evidence of deterioration over time in literacy skills at all levels above about the 10^th percentile. The same is also true for the high school graduates in Chart 15. The university graduates (shown in Chart 16) show mixed changes up to about the median but, again, clear deterioration above that point. Thus, there is evidence of deterioration of higher level literacy skills for all education groups.

To pursue these issues of ageing and cohort effects further, we estimate a series of regressions in which we pool the IALS 1994 and the IALS 2003 data and include cohort dummy variables in our standard specification. The public use version of IALS 1994 does not contain a continuous age variable so we adjust our standard specification by using age dummy variables instead of age and age squared variables. We report the results from this specification using simple OLS estimation in the OLS column of Table 4. Note that all of the following specifications include the parental education and immigrant status variables. In all cases, these variables exhibit the same patterns as in earlier estimation (with low parental education having a negative impact but the remaining variables not being significant) and we do not report them for the sake of brevity. As in the earlier specification, the schooling variables enter strongly and significantly. The schooling effect is larger than we witnessed in earlier tables. This is evident once we include the cohort dummy variables. Essentially, the schooling variable in the earlier specifications was partly picking up the cohort effects shown here: older cohorts have both less schooling and higher literacy, leading to an under-estimate of the true impact of schooling.

The age and cohort group effects in the pooled OLS results show interesting patterns mirroring those seen in the charts. In particular, when we estimate our standard specification using IALS 2003 and replacing age and age squared with the age interval dummy variables, we again find a relatively flat age profile for much of the age range. In particular, the first two age dummy variables (corresponding to the 36 to 45 and 46 to 55 age groups) do not have large coefficients and do not show statistically significantly different effects relative to the base age group (the 26 to 35 age group). The older two age groups show negative effects, with the 65 and over age group dummy having a coefficient of -0.16. This fits closely with the estimated patterns using age and age squared variables. Once we include the cohort dummy variables, however, the age effects assume a continuous, downward sloping profile, with the 36 to 45 year olds having 2.4% lower average literacy and the 46 to 55 year olds having 5% lower average literacy than the base group. At the same time, all the cohort dummies apart from the last one have significant, positive effects that increase with the cohort. The cohort that was 56 to 65 in 1994, for example, has average literacy levels that are 5% higher than those for the base cohort, which was 26 to 35 in 1994. The very oldest cohort has literacy levels that are approximately the same as the youngest cohort. Overall, the implication from these results is that the flat literacy age profile (at least up to age 55) arises from a combination of declining literacy with age and lower average literacy for more recent cohorts.

We investigate these phenomena further using quantile regressions in the remaining columns of Table 4. Quantile regressions are run using the whole sample of individuals and show how, for example, the 10^th percentile of the literacy distribution conditioning on the included covariates changes as each of the covariates change marginally. Thus, the coefficient on schooling can be used to answer the question, "how does the 10^th percentile of the literacy score distribution for people with 11 years of schooling differ from the 10^th percentile of the literacy score distribution for people with 10 years of schooling?" Several points stand out from these regressions. First, the impact of schooling declines across quantiles: years of school have a greater impact in shifting lower quantiles than upper quantiles. In examining this evidence, it is important to take account of the second order terms since the schooling effect declines faster with additional years of school at the lowest quantile while it is closer to linear at the top. This means that schooling has a much larger effect in shifting the bottom than the top of the distribution at low years of schooling but about equal effects at the top and bottom of the distribution around 16 years of schooling. Second, the declining literacy with age effect is much larger at the top than the bottom of the distribution. Thus, as in the charts, the bottom of the distribution for various age groups is relatively similar but older age groups have much lower 90th percentiles. According to the table, for example, the 56 to 65 year old group has a 90th percentile that is 17% lower than the 90th percentile for the base (26 to 35 year old) group, controlling for education and parental education. Third, while the cohort effects are evident across the distribution, they are much stronger at the top end. Thus, the cohort who was 56 to 65 in 1994 (i.e., who were born between 1929 and 1938 have a 90th percentile that is 13% higher than that for the base group (who were born between 1959 and 1968. The intriguing question is whether this reflects a decline in effectiveness of the school system. The fact that the charts show some improvements in literacy across cohorts at the bottom end of the distribution while the regression results show declines is likely due to the fact that the latter control for education and parental education: newer cohorts have more education, as do their parents. Taken together these results may suggest that schools are doing a poorer job of imparting literacy at any given level but that there have been real benefits to the fact that successive generations have attained greater schooling.

In Table 5, we present quantile regressions for those with a high school or less education and those with a university education in order to take a closer look at education related effects. For the high school or less educated group, the ageing effects are virtually non-existent at the lowest quantiles, which perhaps reflects that there is little loss with age of very basic literacy. Also, there is evidence that the lowest quantiles for this education group have actually increased across cohorts. At the top end of the distribution, though, the results are much like the overall results in Table 4: strong declines with age combined with declines across recent cohorts. Thus, it seems possible that changes to public and high school education have been equalizing across cohorts, with improved literacy for those at the bottom being offset by not providing as good literacy training for those at the top. For the university educated, the declines with age are evident across the whole distribution. There is also evidence of declines across cohorts but these are not statistically well defined. This could be simply a sample size problem or it could suggest that the real difficulties in literacy across generations are associated with other post-secondary education rather than with university education. We need to study this point further.

Before leaving the cross-generational/age discussions, it is important to provide a caveat. This discussion hinges on the test scores being comparable across the IALS 1994 and IALSS 2003. There is reason to believe that this is the case (since the two share questions and the remaining questions have been scaled to allow for such comparisons, and also because of the differences in patterns across groups discussed above), but it is still a potential source of concern. The results here would be strengthened if we had access to the scores just for the 40% of questions that are common across the two surveys.

Post-school, work related literacy acquisition

The results in the previous section indicate that adults lose literacy skills with years since leaving school. In this section, we investigate a "use it or lose it" model of literacy. In particular, we examine whether individuals who use literacy skills on their jobs maintain higher literacy, controlling for age. To do so, we re-run our basic specification using the IALSS 2003 including variables on literacy use at work and occupation. The literacy use at work questions ask about frequency of performing reading, writing and mathematical tasks. Thus, for reading, questions are asked about five tasks. There are also questions on five writing tasks and five math tasks. We construct dummy variables equalling 1 if the individual responded that he or she performed four or five of the reading related tasks at least once a week and similar variables for the writing and math tasks. We also constructed dummy variables corresponding to performing one to three of the tasks at least once a week for each of reading, writing and math. Finally, we constructed a dummy variable corresponding to individuals who answered that they performed all of the tasks in a given area (e.g., reading) "rarely". In the regressions, the omitted category is then people who responded that they used some of the tasks but at a frequency of less than once a week.

We present simple OLS results from the basic regression including these literacy-at-work variables plus a job tenure variable in the OLS 1 column of Table 6. In order to reduce clutter in the table, we control for but do not report the parental education and immigration. Their estimated effects are very similar to what is reported in the earlier tables. The sample includes all individuals, whether working or not. The non-workers are captured in the dummy variable corresponding to job tenure being unknown. The estimated effect of job tenure is positive and statistically significant but very small: one extra year of job tenure increases literacy by 0.06%. Thus, literacy effectively does not change with job tenure. The literacy at work variables indicate that those who use literacy intensively at work (who report performing 4 or more literacy tasks in a group at least once a week) have higher literacy. Performance of reading tasks and performance of writing tasks at work are highly correlated, with 83% of those who say they read rarely also reporting writing rarely at work. Thus, in interpreting the coefficients, it is best to add together the coefficients from the reading and writing scores. Based on this, a person who performs four or more reading and four or more writing tasks per week has approximately 3.7% higher literacy on average. At the other end, a person who rarely writes or reads at work has approximately 2% lower average literacy. These effects are large enough to be noteworthy, but are not huge. That is, there is clearly a correlation between literacy of the individual and literacy use at work, but those who use literacy skills at work are not substantially more literate. Note that we do not know which way any causality runs behind these results. Literate individuals may be more likely to be employed in high literacy jobs, or high literacy jobs may help individuals maintain their literacy, or some combination of the two. Math use is not as correlated with the other two types of literacy at work groups as they are with each other. Thus, of the people whom we classify as reading rarely, only 43% say they also use math rarely. The estimation results do not report a positive effect on document literacy from using math often at work but those who use it rarely have over 3% lower literacy than the base group.

In the OLS 2 column of Table 6, we add in occupational dummy variables. Their inclusion has some effect on the literacy at work variables but mainly in increasing their standard errors due to collinearity rather than substantially reducing their estimated effects. Thus, reading and writing often at work still implies 2.5% higher literacy and doing so rarely implies 2.1% lower literacy. The base occupational group is composed of managers and relative to them, those in the professions, clerical and technical jobs all have similar literacy levels. Interestingly, so do skilled agricultural workers and craftsmen and trades workers (recall that these results hold after controlling for education). On the other hand, service workers, machine operators and elementary labourers all have significantly lower literacy. Again, though, these effects are not huge. For example, the difference in average literacy between a manager and an elementary labourer is 5%. Compare this to a difference of over 30% between a person with 12 years of education and one with 16 years of education.

As a final check on the importance of literacy use at work, we re-ran the pooled specifications but only for those in the professional, technician and clerical occupations on the assumption that these are high literacy use occupations. If being in such occupations allows one to maintain literacy then we would expect the estimated negative age profile to be less steep for this group. However, when we carry out this estimation, the estimated age coefficients are very similar to those presented for all individuals in the OLS column of Table 4. This may be because of slippage in the samples (i.e. the set of people who are in these occupations in 1994 need not be the same as those in 2003) but this is likely small relative to the strength of the result. Overall, the evidence on the impact of literacy use at work is mixed. There are clear correlations between literacy use at work and average literacy levels but these are not large relative to the impact of schooling. Also, being in a high literacy occupation does not seem to forestall the decline in literacy with age experienced by the individuals in this sample.

---

Notes

1. This number is obtained by taking the first derivative of the age profile and evaluating it at age 30. For this quadratic profile, this first derivative equals the coefficient on the linear age variables plus (two times the coefficient on the quadratic age variable times 30).
2. This calculation is made in the same way as the one described in footnote 1.
3. Note that the number of observations is now fewer because we have dropped observations for which province where the individual went to high school is not stated.
4. The test statistic corresponding to the joint hypothesis that the province of residence dummies and their interactions with age are significantly different from zero is distributed as F(22,13821) and has an associated p-value of 0.009.
5. This conclusion would not be true if province of residence when in high school has a direct impact on literacy outcomes. This might be the case if individuals living in resource rich provinces while in school purposefully under-invest in literacy because they know they are going to drop-out and take a job that does not require literacy when they are old enough. Essentially, we must assume that such a relationship does not exist, at least after controlling for current province.
6. Experience is defined according to the standard Mincer expression (i.e., experience = age - years of schooling - 6). The coefficient on the experience variable in our simplest OLS regression (which includes only a female dummy, schooling, schooling squared, experience and experience squared) is 0.00051 with a standard error of 0.00045 and the coefficient on the experience squared term is -0.000073 with a standard error of 0.0000083.