The young and the old

Demographic transformations can occur fast, with profound effects on the functioning of economies.3 A century ago, life expectancy in the United States of America (USA) was around 40 years. Today, it is almost double that figure. In the Russian Empire on the eve of the 1917 revolution, life expectancy was around 31 years. By 1994, life expectancy in Russia was 66 years. And today, it is around 72. Russia’s total fertility rate (the number of births per woman) is estimated to have fallen from around 7.5 a century ago to about 1.4 in the mid-1990s, before rising to around 1.8 today (see Chart 1.1).

Demographic shifts both reflect and catalyse changes in education and technology. As life expectancy in Russia has risen from 31 years to 72 years, people’s ability to accumulate skills has changed dramatically, as have the prospects of achieving a financial return on those skills. Indeed, average years of education have risen rapidly as life expectancy has improved and the total fertility rate has declined. At the same time, technological change (such as the advent of electricity and computing; see Chart 1.1) has further increased demand for basic cognitive and computing skills, while also contributing to better standards of healthcare and further improvements in life expectancy.

The demographic changes that are being observed in emerging markets today are similarly fast-paced, as documented in the Macroeconomic Overview. Before the financial crisis of 2008-09, countries where labour force growth outpaced population growth accounted for 90 per cent of emerging market economies’ contributions to GDP. By 2040, labour force growth is expected to exceed population growth in only 20 per cent of the world’s emerging markets. This means that a given rate of productivity growth will translate into a lower rate of per capita income growth.

This chapter looks at how rapid demographic change influences countries’ development paths, examining the impact of trends in demographics, technology, migration and the accumulation of skills. On the basis of that analysis, this chapter highlights policies that can help to deal with demographic challenges, looking at how countries in emerging Europe can improve labour force participation rates and enhance the productivity of older workers, and looking at how Turkey, Azerbaijan and the economies of Central Asia and the southern and eastern Mediterranean (SEMED) can create jobs for their young people.

The chapter starts by looking at the demographic dividends that countries can secure as their birth rates fall and their life expectancy rises. These dividends manifest themselves in a higher ratio of workers to non-workers, higher levels of savings and greater investment in human capital. Over time, however, such dividends may be reversed and turn into demographic headwinds. The chapter then looks specifically at the economies in the EBRD regions that are in the early stages of that demographic transformation process – Turkey, Azerbaijan and the countries of Central Asia and the SEMED region – before turning its attention to the countries of emerging Europe, which are battling the headwinds of population ageing.

As we shall see, responses to population ageing may involve a combination of immigration, automation and longer working lives. This chapter looks at how countries can help people to stay productive for longer, exploring the ways in which pension and tax systems encourage people to remain in employment and assessing the factors that enable people to enjoy longer working lives (such as better health and lifelong learning). Chapter 2 then examines the subject of automation, and Chapters 3 and 4 look more closely at cross-border and internal migration.

In addition, labour force participation among women tends to increase as the birth rate declines, although trends vary in accordance with prevailing social norms and skill requirements in the workforce.5 Relatively strong labour force growth, in turn, raises the growth rate of per capita income (income convergence) for any given rate of output per worker, as discussed in the Macroeconomic Overview.

In addition, increases in life expectancy raise lifelong returns to education, while lower fertility rates enable both parents and the state to commit more resources to each student. Thus, a decline in the number of children leads to an increase in quality.7 As a result, the accumulation of human capital accelerates, providing a further boost to productivity growth. This is reflected in the strong relationship between higher average years of education and lower birth rates (see Chart 1.2). Increases in savings and spending on human capital in response to demographic change are referred to as the “second demographic dividend”. The increased spending on human capital may be of particular importance as the empirical relationship between demographic variables and aggregate savings in the economy tends to be weaker.

A rise in accumulated pension obligations will necessitate increases in taxation and public debt. For instance, the labour tax wedge (the difference between gross salary and take-home pay) tends to rise as the old-age dependency ratio increases.8 Pensioners may then start selling their assets (whether held privately or by pension funds on pensioners’ behalf). If this happens on a significant scale, that dissaving can lead to a rise in long-term interest rates.9

In addition, as the age of the median worker rises, the total stock of human capital may start to depreciate, at least when it comes to basic cognitive skills or physical abilities that are harder to retain with age.10 As a result, the second demographic dividend may also become a burden, typically after the first demographic dividend has been exhausted.

The large numbers of young people in the populations of such economies are sometimes referred to as the “youth bulge”, owing to the shape of the resulting demographic “tree” (see Chart 1.3). The demographic trees of those economies contrast sharply with those of the rapidly ageing economies elsewhere in the EBRD regions (referred to hereinafter as “emerging Europe”).

The demographic challenges faced by these economies are shared by many developing economies all over the world, from southern Asia to South Africa, that need to keep creating jobs for their many new entrants to the labour market. In Egypt, for instance, the number of jobs needs to grow by around 2 per cent every year (roughly 750,000 jobs), in order for the economy to be able to absorb all new labour market entrants.

At the same time, a strong macroeconomic policy framework and a favourable and stable investment climate are required in order to leverage global savings and manage the current account imbalances that are associated with investment persistently exceeding domestic savings. Turkey is a good example of a young economy with persistent current account deficits.

Boosting investment and the quality of human capital also involves strengthening economic and political institutions and improving the business environment. Lowering barriers to the entry and exit of firms and improving the quality of management can help young economies to develop those export industries in which young and growing populations can provide a strong comparative advantage.

Thus, many middle-income economies are at risk of growing old before they can grow rich. The challenges that this poses extend well beyond “mechanical” demographic headwinds to per capita income growth. The necessary adaptation of the economy – in terms of facilitating longer working lives, leveraging technological progress, fine-tuning migration policies and redesigning social safety nets – may well be more challenging in countries with less developed economic and political institutions and less mature market structures. This may be a particular concern for economies in the EBRD regions where governance tends to be weaker than the levels that could be expected based on countries’ per capita incomes.13

The demographic transformation process, which manifests itself in falling birth rates and ageing populations, is much more advanced in emerging Europe than it is in emerging Asia, Latin America, the Middle East or Africa. Many countries in emerging Europe reaped demographic dividends relatively early in the 20th century, similar to advanced economies. Following the start of the deregulation and liberalisation reforms in the early 1990s, the birth rates of post-communist countries fell further, from what were already relatively low levels, reflecting economic hardship and increased uncertainty. In some countries this effect was transitory, rather than permanent, with people delaying having children during the early years of the transition process.14 While the availability of subsidised high-quality childcare can have a positive impact on both fertility rates and labour force participation among women,15 it is unlikely to reverse the long-term decline in the number of births in higher-income economies, which reflects changes in the preferred number of children.16

In fact, the demographic profile in emerging Europe is similar to that seen in western Europe, with those countries’ populations ageing at approximately the same rate as the populations of advanced European economies (see Chart 1.6) on account of modest birth rates and continued improvements in life expectancy.

Moreover, on the basis of current trends, the working-age population of central and south-eastern Europe is expected to shrink by 17 per cent by 2040 (see Chart 1.7), with broadly similar developments forecast in eastern Europe. At the same time, however, those depopulation trends are expected to differ significantly from country to country, with implications for productivity and welfare, as discussed in Chapter 4.

Many advanced economies have experienced significant inflows of working-age migrants that have compensated for weak or negative growth in the native working-age population (see the upper-left quadrant of Chart 1.8). As a result, many high-income countries are continuing to enjoy the first demographic dividend (that is to say, their labour force is still growing faster than the total population). Indeed, the population of the USA remains younger than those of many economies in emerging Europe (see Chart 1.6).

Uniquely, emerging Europe has experienced the opposite, with migration generally exacerbating – rather than mitigating – demographic pressures (see the lower-left quadrant of Chart 1.8). Over the past two decades countries in emerging Europe have seen substantial net outward migration for both highly skilled and low-skilled labour, mostly to advanced EU economies.18 At the same time, Turkey and Jordan, countries with higher natural growth rates, have experienced a significant influx of refugees.

Emigration does bring economic opportunities, however, if the skills and experiences of returning migrants and diasporas abroad can be harnessed and used to improve domestic productivity, boost innovation and broaden export markets. A section of Chapter 2 looks more closely at the economic impact of emigration in emerging Europe, and Chapters 3 and 4 examine migration patterns in the region in greater detail.

On average, however, countries in the EBRD regions have been automating work processes more slowly than advanced economies with similar demographic profiles. This suggests that there is significant potential for rapid automation in the future, leveraging the regions’ relatively high levels of human capital (as confirmed by the analysis in later sections of this chapter). At present, robots are primarily deployed in the automotive, electronics, appliances, chemicals, and machinery and equipment sectors, but they may be used more widely in other sectors in the future.

Leveraging technological advances may help to reduce upward pressure on wages and raise labour productivity, thereby strengthening economies’ competitiveness in spite of demographic headwinds.20 This may, however, entail greater polarisation of jobs – that is to say, rising numbers of low-skilled and high-skilled jobs, and declining numbers of medium-skilled jobs (see Chapter 2 for a discussion of this issue). Consequently, economic policy needs to respond to the combined impact that demographic shifts and technological change have on income distribution and the nature of jobs.

Thus, the economic impact of demographic pressures cannot be analysed in isolation. What matters for economic outcomes is the interplay between demographics and trends in technology and migration, as well as the way in which education and social safety nets respond to those trends. This is true both in rapidly ageing societies and in young economies (see the upper and lower panels of Chart 1.10 respectively). For instance, while an increase in immigration may support growth in an ageing economy, it will pose challenges in terms of inclusion and social cohesion. Increased use of automation can ease labour shortages that arise due to ageing and emigration but may further exacerbate inequality.

In young economies, meanwhile, it may be difficult to create jobs and boost labour productivity in the absence of increases in capital-to-worker ratios (automation) and significant outward migration. There, joblessness among young entrants to the labour market also has the potential to threaten social cohesion. These interactions are explored in greater detail in other parts of this report, with Chapter 2 looking at labour markets, Chapter 3 focusing on migration across borders and Chapter 4 looking at migration within countries. The remainder of this chapter focuses on the lengthening of working lives in ageing economies.

Retirement ages in the EBRD regions are set to increase by an average of two years over the next two decades, on the basis of legislation that is already in force. However, old-age support ratios (defined as the retirement-age population as a percentage of the working-age population) will continue to rise. By 2040, the average old-age support ratio in emerging Europe is projected to be well above the G7 equivalent, with one retirement-age adult for every two working-age adults (see Chart 1.12).

In these circumstances, pension systems cannot deliver both broad coverage of the population and generous entitlements (high pension-to-wage ratios) while remaining fiscally sustainable. Different countries approach this trilemma in different ways. In the EBRD regions, the fiscal burden of public pensions tends to be manageable, with social security systems focusing on providing basic benefits aimed at ensuring that pensioners remain above the poverty line.26

As a result, average retirement incomes in the EBRD regions tend to be less generous than those observed in advanced economies (relative to wages). A number of public pension systems include funded (defined contribution) pillars, but the coverage of those second-pillar components remains modest, with relatively low contribution rates. Private pension savings have also remained low as a percentage of countries’ GDP, while real (inflation-adjusted) returns on pension assets held by public and private pension funds have been negative in many countries.27

Consequently, younger cohorts of workers need to be fully aware of the need to work longer and/or accumulate substantial savings in order to maintain higher levels of income in later life. At present, reforms that seek to improve the sustainability of public pension systems by reducing the generosity of pension entitlements are often strongly opposed by voters.

Reforms that raise the retirement age, if successfully implemented, are likely to provide stronger incentives to work longer. Evidence suggests that workers are much more likely to remain in the labour force if the benefits they are eligible to receive on exiting it are lower.28 In Iceland, for instance, workers who choose to work on beyond the statutory retirement age of 67 (which will rise to 70 in the future) are eligible to receive a larger pension, with their pension entitlement rising by 0.5 percentage point per month for a maximum of five years. Large numbers of eligible workers have taken advantage of this opportunity.

Another challenge is to limit the impact that rising old-age support ratios have on labour taxes and incentives to substitute capital for labour. Traditionally, both funded pension schemes and unfunded pay-as-you-go schemes have been financed using social security contributions (a form of labour tax). As the labour force dwindles and numbers of retirees rise, ever higher levels of social security contributions may be required in order to fund pension obligations, making labour more expensive from an entrepreneur’s perspective and strengthening incentives for automation, offshoring of jobs and tax avoidance.

Funding pensions out of consumption and wealth taxes may represent a more robust option, with a broader tax base and a smaller impact on the economy’s competitiveness and incentives to substitute capital for labour. However, taxes on consumption tend to be more regressive than income taxes. Equity considerations relating to such reforms need to be addressed using appropriate expenditure policy measures.

The role that good health plays in explaining the likelihood of a person being employed or seeking employment can be investigated using data from the Gallup World Poll – an annual household survey covering more than 160 territories around the world (including all economies in the EBRD regions). Each round of the survey involves at least 1,000 respondents in each country, with participants being asked several questions about their physical and mental health. Participants’ responses can be aggregated to establish a rescaled composite index ranging from 0 (very poor health) to 1 (good health).

The self-assessed health of male respondents in emerging Europe declines sharply after the age of 50 (see Chart 1.13). This mirrors the decline in labour force participation rates among the same cohorts (see Chart 1.11). In contrast, self-assessed health tends to decline later in life in other emerging markets and the G7. However, causality may run in both directions, with a number of studies finding that early retirement may lead to a less active lifestyle and a further deterioration in health.31

More broadly, deteriorating health is strongly linked to people’s decisions not to seek employment.32 Looking at Gallup World Poll data for all countries with GDP per capita in excess of US$ 1,100 at market exchange rates, around 22 per cent of respondents between the ages of 30 and 65 stay out of the labour force. The regression analysis reported in Table 1.1 looks at determinants of the likelihood of labour force participation across various countries, taking account of several different factors (such as gender, age, being in full-time education or training, and living in a particular country in a given year). Self-reported health is a major determinant of the decision not to seek employment, even when all of those other factors are taken into account. Health has a stronger impact on labour force participation among people between the ages of 50 and 65.

In emerging Europe, the impact of poor health is particularly strong among people over the age of 50 (see columns 1 and 3 of Table 1.1). If self-reported health improves from the 25th to the 75th percentile of the distribution of answers (corresponding to an increase of 0.4 on the 0-1 scale), that is associated with an increase of approximately 7 percentage points in the likelihood of seeking employment for respondents in emerging Europe between the ages of 50 and 65.33

A similar relationship can be observed in data collected by the Organisation for Economic Co-operation and Development (OECD) as part of its Programme for the International Assessment of Adult Competencies (PIAAC). Those PIAAC data, which are discussed in greater detail below and focus on skills and education, identify individuals who are in employment, education or training. Survey respondents are also asked to assess their health using a five-point scale. In the interests of comparability with Gallup data, those responses have been rescaled such that they range from 0 to 1. As with the Gallup data, good health is strongly associated with an increased likelihood of being in employment, education or training (see column 4 of Table 1.1).

The strong relationship between good health and an increased likelihood of labour force participation among people aged 50 or over calls for improvements in healthcare in the EBRD regions. It also highlights the need to make workplaces more suitable for ageing workers. Pilot projects at a BMW plant in Dingolfing (Germany) and a Renault plant in Novo Mesto (Slovenia) confirm that relatively inexpensive changes to equipment, uniforms and factory floors at manufacturing facilities (such as the introduction of softer floors, orthopaedic footwear, ergonomic chairs, vertically adjustable tables and magnifying lenses for work involving small parts)34 can significantly improve workers’ health and reduce sick leave among older employees. The potential to reduce rates of absenteeism and improve productivity among older workers mean that such investment is worthwhile for the companies in question.

Studies suggest that learning new skills requires more effort the older you become, reflecting the declining adaptability of the brain. Consequently, it may be that acquiring new technical skills (such as the ability to speak a foreign language or develop software) ceases to be worthwhile once workers reach their early thirties36– well below the age of the median worker. What is more, by 2030 the median worker is projected to be in their early forties in the EBRD regions (and as old as 46 in Bulgaria).

The cost-benefit profile of acquiring new skills after the age of 40 can be improved if the cost of mid-career learning can be lowered. For example, changes could be made to basic educational qualifications (including university degrees) in order to help individuals to learn how to keep learning, rather than trying to provide students with specific skills that will remain relevant for decades. Indeed, research suggests that around half of all acquired skills now lose their relevance within five years in the absence of retraining.37 Workplace practices and professional training may also have an important role to play when it comes to retaining and updating skills.

The average skill levels observed across all age groups in the PIAAC data are strongly correlated with the country-level measures of human capital (based on average years of education) that were used in growth accounting in earlier Transition Reports. At an individual level, a better performance in PIAAC tests is associated with better socio-economic outcomes: an increased likelihood of being employed, higher wages and greater job satisfaction. The analysis below looks at these links in more detail.

However, as in advanced economies, average performance in the EBRD regions tends to be weaker among older survey participants (see Chart 1.15). Across economies, scores initially improve as survey participants gain extra years of education and training, before declining from around the age of 30 onwards. This decline in performance among older workers is broadbased, with the dispersion of scores remaining broadly stable across the various age groups.

The downward trend seen after the age of 30 reflects a combination of two factors: first, the impact that the ageing process has on basic skills; and second, cohort effects (that is to say, the fact that older respondents were educated at an earlier point in time, when education enrolment rates may have been lower and different curricula were in place). In other words, older workers may need to update and broaden their skills – rather than simply retaining them – in line with changes in technology and working practices.

Studies suggest that such cohort effects may have a large impact.39 Indeed, with younger people tending to be more familiar with computers and information technology (IT), it is noticeable that, across countries, average problem-solving scores (which involve a computer-administered test) decline faster with age than literacy and numeracy scores (which are based on a pencil-and-paper assessment).

As the PIAAC surveys have only been conducted once, it is not possible to distinguish between the impact of ageing and cohort effects. Nonetheless, this analysis remains informative. Measures that enhance mid-career learning can help people not only to retain skills in the face of ageing, but also to update them (thereby reducing the differences between the various cohorts).

As one would expect, having a tertiary (university) degree is associated with higher average skills across all age groups when compared with respondents who have only completed secondary education (see Chart 1.16), which represent the baseline group in this analysis. Indeed, for 57 per cent of respondents in the EBRD regions, the completion of secondary education represents their highest academic qualification (with a similar percentage being observed in comparator countries), while one in five survey participants in the EBRD regions hold a tertiary degree. Having a research degree such as a PhD (held by 0.5 per cent of respondents in the EBRD regions) is associated with an additional increase in test scores.

A large positive effect in terms of the retention of skills appears to come from the frequent use of skills at work. Survey participants were asked how frequently they had to use problem-solving, numeracy and literacy skills in the workplace, with answers ranging from “never” to “every day”. Those answers have been aggregated to form an index measuring the use of skills at work, which ranges from 0 (skills never used) to 19 (frequent use of all skills referred to in the survey). If the use of skills increases from the 25th to the 90th percentile of the distribution of this index, that is associated with an increase in test scores of up to half a standard deviation. The impact of this is particularly large for people aged 35 and over (see Chart 1.16).

Employee and employer training is also associated with higher levels of skill throughout a person’s working life. A training index constructed for the purposes of this analysis takes account of recent training organised by employers. Only 19 per cent of respondents in the EBRD regions reported having undertaken such training, compared with 30 per cent in other economies covered by the PIAAC tests. The index also covers private tuition (such as language classes) that has been undertaken by the respondent in the last 12 months – and again, the participation rate in the EBRD regions (14 per cent) is lower than that observed in comparator countries (18 per cent). The last component of the training index is the extent of the respondent’s willingness to learn, as reported by respondents themselves using a four-point scale. The overall index ranges from 0 to 10, and an increase in a person’s training score from the 25th to the 90th percentile of the distribution is associated with an increase in skill assessment scores of up to 17 per cent of a standard deviation.

Overall, these results reinforce the view that providing mid-career training can help employees to retain and upgrade their skills. Technological advances are opening up new avenues in terms of the provision of such training (as in the case of distance-learning programmes, for example). Government-backed loan initiatives with income-contingent repayment over the longer term can help to make degree courses and mid-career training programmes more affordable and attractive. And packages given to employees in the event of redundancy can include ring-fenced retraining allowances (as in the case of the restructuring of Slovak Telekom, for instance). Moreover, transparent evaluation and rating systems can help to improve the quality and relevance of training.

In addition, greater use of teams with a variety of different ages can facilitate on-the-job learning, as the skills and experiences of younger and older employees tend to be complementary.40 Moreover, increased rotation across tasks can enable employees to use and upgrade their skills at work, while in some cases also reducing the physical and mental strain associated with repetitive actions.41

If degrees do not help to develop skills demanded by the market, a policy aimed at increasing the percentage of the population who have tertiary qualifications may lead to individuals becoming overqualified for the jobs they are doing. This, in turn, may have a strong negative impact on job satisfaction – and, ultimately, people’s well-being.

The PIAAC survey provides an opportunity to relate a measure of overqualification to job satisfaction. Each respondent was asked to name the qualification that a successful applicant for his/her job would be expected to hold at present. This qualification can be compared with the respondent’s level of education, with survey participants who are educated to a higher standard than their job would normally require being deemed to be overqualified. Inevitably, this measure of overqualification involves a subjective element – the respondent’s opinion about the qualification that is needed to do his/her own job.

Around a third of respondents in the EBRD regions were deemed to be overqualified (with the same percentage being observed in comparator countries). Younger people (those aged between 16 and 24) were more likely to be overqualified (with 42 per cent of them falling into that category). Incidence of overqualification was also above average among university graduates (37 per cent), rising to 51 per cent for holders of research degrees.

Participants also reported their level of job satisfaction on a five-point scale ranging from 0 (extremely dissatisfied) to 4 (extremely satisfied). Overall, 80 per cent of respondents in advanced economies were satisfied or extremely satisfied with their jobs, compared with 75 per cent in the EBRD regions.

Overqualification is one of the key causes of low job satisfaction, even after the level of education and performance in PIAAC tests have been taken into account alongside other factors such as the respondent’s country, industry, gender, age and type of employment (see Table 1.2). The negative relationship between overqualification and job satisfaction is even stronger in the EBRD regions, where being overqualified lowers job satisfaction by about 25 per cent of a standard deviation (see column 3). In contrast, having a higher skill score and having a research degree are both associated with a higher level of job satisfaction.

The strongly negative relationship between overqualification and job satisfaction reinforces the need to focus on the quality of degrees. In this regard, greater private-sector participation in the design of educational and retraining programmes and the drawing-up of their curricula can help to reduce skills mismatches in the labour market.

Educational reforms in these countries – which included changes to curricula, teacher training and student assessment – were phased in gradually over many years and at differing speeds. Certain changes were made quickly, including increases in foreign language tuition, the introduction of computers and the removal of ideological subjects. In addition, in many countries people graduating before those early transition reforms were assigned to their future employers by central authorities, whereas people graduating after those reforms had to look for a job themselves and were free to choose their employers.

Do the current skills of people educated before those early transition reforms systematically differ from the skills of people educated after the start of the transition process? PIAAC data can shed light on this question. For every individual, survey records include the total years of education and the year when the highest qualification was obtained. And for every country, the timing of early transition reforms can be determined on the basis of the year in which the prices of most goods and services were liberalised42 – information that can be derived from the EBRD’s transition indicators. So, for everyone in the seven post-communist countries covered by the PIAAC surveys, the average assessment score can be related both to the person’s level of education and to the number of years of post-reform education. As before, this analysis also considers other relevant characteristics, including the respondent’s gender, age and country of residence.

This analysis shows that people who undertook some or all of their years of education after the start of those transition reforms performed better in PIAAC tests (see Table 1.3). As expected, each extra year of post-reform education makes a positive and statistically significant contribution to the test score (of around 0.4 per cent of a standard deviation), adding up to a total increase of around 7 per cent of a standard deviation for a university graduate educated entirely after the reforms relative to a university graduate educated entirely before the start of the reforms. Similar results can be observed if one looks at the number of years in education completed since the start of those reforms as a percentage of the total number of years of education.

This analysis is able to take account of participants’ ages, as the total years of education vary across individuals, while the start dates of reforms vary across countries, implying that participants of the same age had varying exposure to post-reform education. However, most participants who undertook some or all of their years of education after the start of reforms are younger than those educated before the reforms, which gives rise to a concern that the results may primarily reflect cohort effects – that is to say, the continuous improvement in education enrolment rates and the quality of education in the 1990s and 2000s.

To alleviate such concerns, a placebo test repeated this exercise for the other countries in the PIAAC sample, which did not experience a transition from central planning to market-based systems. In this group of non-post-communist economies, start dates for fictitious price liberalisation reforms (ranging from 1990 to 1995) were assigned to countries on the basis of alphabetical order. Similar calculations were then used to determine the number of years of education that each individual had undertaken following those pseudo-reforms.

Reassuringly, the effect of those pseudo-reforms in non- post-communist countries is very small (and, if anything, negative), notwithstanding the fact that those countries are subject to the same potential cohort effects as post-communist countries.

In the early 1990s Estonia embarked on a comprehensive reform of its education system, moving towards the model employed in Finland. Those reforms, which spanned secondary and higher education and were rolled out over the period 1992-98, involved changes to curricula, student assessment and the retraining of teachers. Faster and more comprehensive reforms were carried out in Estonian-speaking schools, with a reform plan for Russian-speaking schools not being put forward until much later, in 1998. A plan to provide all education in Estonian was approved in late 2007.43

The situation in Estonia provides an opportunity not only to compare the performance of people who began studying after the implementation of those reforms with that of people who graduated before 1992, but also to look at differences between Estonian speakers (who were more exposed to reforms) and the Russian-speaking population in terms of improvements in test scores. Approximately 800 Estonians participated in the PIAAC assessment, 27 per cent of whom were Russian speakers. Of those educated after 1998, around 17 per cent identified themselves as Russian speakers. People who were partially educated during the reform period were excluded from this analysis.

Difference-in-difference regressions show that improvements in test scores following the implementation of reforms were indeed significantly larger among Estonian speakers, with the difference between Estonian speakers and non-Estonian speakers totalling 20 per cent of a standard deviation of test scores (see Chart 1.17). That difference is statistically significant at the 5 per cent level and can be attributed to the effect of Estonia’s education reforms in the 1990s.

A concern remains, however, that these results may reflect national language speakers’ inherently superior ability to improve their skills relative to speakers of other languages. For example, taking a PIAAC test in one’s native language is associated with a statistically significant improvement in test scores. In order to see if this could be the key driver of improvements in test scores over time, this exercise has been repeated for other countries, comparing differences between the scores of national language speakers educated before 1992 and after 1998 with the corresponding differences for non-national language speakers. Unlike in Estonia, the differential for other countries is very small and not significantly different from zero in the statistical sense (see Chart 1.17).

While skill scores are lower among older workers, the wage premium associated with improved skills increases with age. People’s abilities tend to change in the course of their lifetime, with a shift from learning skills to utilising acquired knowledge in decision-making. Through on-the-job learning, people may acquire various “soft” skills (such as the ability to understand the emotions and motivations of others, negotiate, pitch a product or manage a team), as well as occupation-specific skills. These are distinct from the skills that are assessed in numeracy or literacy tests. And yet, test scores and broader skill sets are likely to exhibit a strongly positive statistical relationship, not least because basic skills such as literacy and numeracy help with the development of other role-specific skills.

In other words, the skills that are assessed in PIAAC tests may act as a useful – albeit imprecise – measure of the broader skill sets accumulated by adults. To account for this possibility in the regression analysis, the average PIAAC test score is instrumented with the number of books the respondent had at home at the age of 16. The number of books at home is a strong predictor of an individual’s taste for learning and ability to acquire skills, but it should not directly influence an individual’s wage or other labour market outcomes, nor can it be influenced by them. Within the EBRD regions, the percentage of respondents who had at least 100 books at home at the age of 16 ranges from 2 per cent in Turkey to 34 per cent in Estonia.

Regressing the log of monthly wages on test scores and several other explanatory variables suggests that the wage premium associated with higher levels of skill is larger for older employees. The difference in wages that is associated with a 1 standard deviation increase in test scores rises from a few per cent for people under the age of 25 to 25 per cent for people aged between 55 and 65 (see Chart 1.18). Returns to skills for that oldest age group are more than twice the size of those observed for people between the ages of 25 and 34, and that differential is statistically significant at the 5 per cent level. While it may be harder to retain and update skills later in life, the financial returns may well compensate for the effort.

The age profile of skill premia also highlights the way in which population ageing may exacerbate income and wealth inequality. People in occupations where lifelong learning and development are the norm benefit from increasing returns on acquired human capital, whereas people in basic occupations with few such opportunities (as is the case, for example, with many jobs in hospitality or construction) may see their incomes stagnate or decline. As working lives become longer, those pay differentials between occupations may increase accordingly.44

Consequently, improvements in lifelong learning opportunities may go a long way towards mitigating income inequality. However, if access to education and entry-level jobs is deeply unequal, such improvements may actually exacerbate inequality instead.45