The Treasury

Global Navigation

Personal tools


Human Capital and the Inclusive Economy - WP 01/16

Part Two   Human Capital Policies and the Distribution of Income

Education, skills and capabilities are correlated with earnings

Data across time and countries show a strong positive correlation between individuals’ education, skills and capabilities, and their earnings. To the extent that governments are concerned about a widening distribution of income, or more particularly about the adequacy of incomes towards the bottom of the distribution, it is natural to consider the role that human capital policies can play.

If a widening distribution of earnings reflects increasing premia for skills[21] in short supply, then it might be possible to reduce dispersion by producing greater numbers of highly educated workers. On the other hand, raising the skills of the low skilled may be more effective in increasing the adequacy of their earnings, as well as in narrowing the overall distribution of earnings.

However, if, as some economists believe, the relationship between education and earnings reflects to a significant extent its value to employers as a signal of innate ability, rather than of acquired skills, then human capital policies might not be effective in improving distributional outcomes. Acemoglu (2001b) synthesises the theoretical and empirical literature on each of these questions.

Skill-biased technological change helps explain a widening dispersion of earnings

He concludes that:

  • While other factors are relevant (e.g. deunionisation, openness to trade, welfare provisions), skill-biased technological change and consequent increases in skill premia are a major contributor to increased earnings inequality in OECD countries[22].
  • Greater openness to trade does not appear to have been a major factor in the widening of the earnings distribution.
  • Standard supply and demand analysis suggests that an increased supply of skill will reduce skill premia, other things being equal. However, technologies adjust to the supply of skills[23], and, the historical record in the United States suggests (see figure 3) that the effect of increased supply on skill premia is likely to be small[24]. On realistic assumptions about the rate at which other labour will be substituted for college educated labour as its wage rises[25], Acemoglu estimates that the relative demand for the latter has increased by 200-300% between 1940 and 1990, having accelerated after 1970.
Figure 3. Relative supply of college skills and college premium (U.S)

Figure 3. Relative supply of college skills and college premium (U.S).

The behaviour of the (log) college premium and relative supply of college skills (weeks worked by college equivalents divided by weeks worked by non-college equivalents) between 1939 and 1996. Data from March CPSs and 1940, 1950 and 1960 censuses.
Source: Acemoglu (2001)
  • In any case, because the flows of skills are small relative to the stock, this mechanism would take a long time to have any effect on earnings dispersion, even without increased demand. Acemoglu calculates that an economy that doubles its output of college-educated workers would reduce skill premia by only 16 per cent after 10 years[26]. Similarly, immigration will have little effect on skill premia [27].
  • It might be expected that higher skill premia will be self-correcting through encouraging further schooling. However, as noted above, the effect of increased supply on skill premia is limited, and, moreover, in some data there is a surprising lack of correlation between education premia and education investments[28]. Nevertheless, Acemoglu acknowledges that the evidence on this is ambiguous.
  • As Figure 4 shows, the broad New Zealand data for the last two decades does suggest that rising returns to education investments were accompanied by an increase in tertiary education completion.
Figure 4. Supply of graduates and income premia for degree 1981-1996 (New Zealand)
Figure 4. Supply of graduates and income premia for degree 1981-1996 (New Zealand).
Source: Maani (1999)
  • Reducing the dispersion of skills (by raising those in the bottom tail of the distribution of skills) is likely to be the most effective human capital policy to address income dispersion and particularly adequacy issues[29].
  • However, if a significant part of the relationship between education and earnings were due to “signalling effects” then increasing educational attainment in the bottom tail might not be effective in reducing dispersion. Evidence from the United States, controlling for age and cohort effects[30], suggests, however, that changes in signalling effects are not very important in explaining changes in skill premia. Other evidence looks at whether social returns to compulsory schooling (measured through average earnings) are similar in magnitude to private returns, and concludes that they are[31]. This also supports the idea that signalling effects are small.
  • In the United States changes in the types of jobs that firms create (“good jobs” versus “bad jobs”) may have been important in shaping the wage distribution[32]. However, this is likely to have been driven by technology – and does not affect the conclusion about putting a top priority on raising the skills of those towards the bottom of the distribution of skills. In Acemoglu’s model, raising skills at the bottom can induce an increased supply of better jobs.


  • [21]The term “skill premium”, defined empirically in a number of ways, refers to the extent to which earnings of skilled workers exceed those of the unskilled.
  • [22]He notes that while increased returns to education account for some of the increase in earnings dispersion, there is evidence for increased returns to skills within education categories, which will account for more of the increase in earnings dispersion. He considers that the analysis based on premia for education will apply more generally to all skill premia (see Acemoglu 2001b).
  • [23]Acemoglu refers to his own work on this (1998, 1999) & argues that it applies as much to technology adoption as to technology creation. The essential point is that an increase in the supply of skills creates, in part at least, its own demand. (See also Nickell & Nicolitsas, 1997).
  • [24]Maani (1999) shows that returns to education in New Zealand increased strongly over the period 1981 to 1996, and were accompanied by rapid increases in participation in tertiary education, translating into a steady rise in the proportion of the working age population with tertiary qualifications – see Figure 4.
  • [25]Technically referred to as the elasticity of substitution between college educated and other labour.
  • [26]This result depends on empirical estimates of the rate at which other labour is substituted for college-educated labour, as wages of the latter rise. Acemoglu postulates an economy that starts with 25% of its workers college educated.
  • [27]He refers to U.S. research by Card (1990), and by Borjas, Freeman & Katz (1997), in support of this.
  • [28]Acemoglu & Pischke (2000) analyse college enrolments across U.S. states and find no correlation between wage inequality and changes in college enrolments. They argue that migration effects, which might explain this, are limited. They also look at cross-country evidence in the OECD and come to the same conclusion. They acknowledge, though, that the increase in enrolments in the U.S. during the 1980s is consistent with demand-induced increases in supply.
  • [29]Recent research using data from the International Adult Literacy Study shows that differences among countries in skill premia explain more of the cross-country variation in the distribution of earnings, than do differences in the distribution of skills. (See Devroye & Freeman, 2001; Blau & Kahn, 2001). This suggests that while making an important contribution, policies that compress the distribution of skills will not, by themselves, eliminate differences in the distribution of earnings across countries.
  • [30]This evidence looks at whether changes in skill premia, or the demand for skill occurred within groups of workers defined by particular cohorts and ages. For the period 1970 – 1990 there were increases in skill premia within these groups (see Table 2, Acemoglu 2001b).
  • [31]Acemoglu & Angrist (2000). The methodology relies on across state variations in compulsory schooling laws, as a means to identify the true effects of additional years of schooling.
  • [32]Acemoglu argues that there may be pecuniary externalities to more productive high-quality jobs, due to imperfectly competitive labour markets enabling employees to bargain to higher wages. Employers do not take these benefits to employees into account, which leads to an under supply of such jobs from an efficiency perspective.
Page top