a more skilled population ahead age or cohort effects

A more skilled population ahead: age or cohort effects?

By Guillermo Montt.

Blog43.0bIn 2012, the OECD released the first results of the Survey of Adults Skills. Workers in Japan and Finland showed higher proficiency than workers in Spain or Italy. Those findings captured the media attention, as OECD country rankings typically do. But there is much more in the data.

The survey assessed the literacy and numeracy skills of adults aged 16 to 65 in more than 20 countries. One of the key findings is that proficiency has somewhat of an inverted U-shape. Literacy proficiency typically peaks among 25 to 34 year-olds and is lowest among those over 55. There are several ways to interpret this relationship, which could be a mixture of age and cohort effects.

If driven purely by age effects, this relationship would tell us that, as workers and individuals age, they acquire more skills (by continuing their studies in formal education, for example) but, after peaking at around age 28, skills depreciate (possibly because they are not put to use in the workplace, or because of innate biological processes). If this were the case, then twenty years into the future we would not expect a change in the skill level of the population (except if there were dramatic changes to the demographic age profile of the population).

If driven purely by cohort effects, age does not interfere: workers maintain their skill levels as they age. The higher performance of younger workers compared to older workers would tell us that, as a result of more and better skill development, new generations are better skilled. As older workers retire, the skill level of the working population is bound to increase, painting a bright future.

The policy implications for a declining age or declining cohort effect are completely different. While the latter calls for lifelong learning and better use of skills at work and everyday life to avoid skills depreciation, cohort effects require more attention to initial education.

The fact of the matter is that age and cohort effects cannot be distinguished in a cross-sectional survey like the Survey of Adult Skills. Fortunately, 15 of the countries that implemented the Survey of Adult Skills also participated in the International Adult Literacy Survey (IALS) in 1994-1998. Comparing the results from the two surveys allows us to distinguish those countries where the generalised age differences in PIAAC reflect age effects, from those where it reflects an upskilling among adults. Comparing the results from the two assessments is possible because the literacy scales share items and can be put on the same scale. (For more details, see the PIAAC Reader’s companion).

With the data from the two surveys we can ask, for example if all adults in 2012 are more proficient in literacy than adults in 1996. Adults aged 16-65 in Australia, Italy, Poland, England (UK) and Northern Ireland scored significantly higher in PIAAC than they did in IALS; Canada, the Czech Republic, Denmark, Germany, Norway, Sweden and the United States scored significantly lower.

Yet these changes can be due to demographic shifts, so it is important to look at specific cohorts to see where the changes lie.

In the first results from the PIAAC Survey the OECD compared age-specific results in the 1990s (IALS) with those of PIAAC for Australia, Canada, the Czech Republic, Finland, the Netherlands and the United States. The figure below summarises the results. In general, if the black line (PIAAC) is systematically above the blue line (IALS) this is evidence that more recent cohorts are more proficient.

This upskilling across cohorts seems to be the case, for some cohorts, in Australia and Finland. Adults aged 20 to 50 in Australia in 1996 when IALS was carried out and adults aged 28 or older in Finland in 1998 had lower proficiency than adults of that age in 2012. (In Finland, there is a potential down skilling among younger cohorts). In Canada, there seems to be a skills gain between the cohorts born in 1950s compared to those born in the 1930s, but skills loss when comparing the cohort born in the 1970s compared to those born in the 1990s.The age-differences in PIAAC do not point to any statistically significant upskilling in the Netherlands.


In England (UK) there are no statistically significant differences between adults aged 16-54 in IALS and PIAAC. There is, however, a significant increase in scores over time among the 55-65 year-olds suggesting that adults born in the 1950s have been better able to retain their skills when compared to adults born in the 1930s, possibly as a result of the increase in the school-leaving age.

Any differences in the cohort-specific results between PIAAC and IALS are not necessarily due to changes in educational. They could be due to changes within each cohort due to migration or changes in skills use patterns. Results must be interpreted with caution as well, though the two surveys share a common scale, they have differences in how they were carried out, potentially influencing the results.

In a few weeks the OECD will release a new round of results for countries that implemented the Survey of Adult Skills in 2014-15. Of these nine additional countries, Chile, New Zealand and Slovenia also participated in IALS, which could provide important insight as to whether any age effects observed in PIAAC reflect an upskilling of the population between the second half of the 1990s and 2014, or simply an age effect.