Human capital, education and health

Before the 1960s, economists treated labour as a given quantity of raw effort. Theodore Schultz’s 1960 presidential address to the American Economic Association, printed as Investment in Human Capital (1961), argued that people acquire skills and health that raise their productivity, that acquiring them uses scarce resources, and that the resources spent on it are investment. He thought this explained facts that the ordinary accounts of growth left over, such as output rising faster than measured capital and labour. Gary Becker’s Human Capital (1964) set out the theory and the first estimates of returns, and separated general training, which raises a worker’s productivity in any firm, from specific training, which is useful only in the firm that provides it.

Three features set human capital apart from a machine. It is embodied in the person and cannot be sold, only rented out through work; so a loan to finance it cannot be secured on the asset (a reason for the weak credit market for education, taken up in Lesson 13). It is built over a long time, so the return comes years after the cost. And its yield depends on health and on the length of working life, which is why health sits beside education as a form of capital.

The test question is which spending counts. Replacing typewriters with computers is physical capital; teaching workers to repair every machine in the plant, older workers returning to school, and a vaccination programme are all investment in human capital (JNU 2015).

Jacob Mincer (1958, 1974) derived an earnings equation from the investment idea and made it the most used regression in labour economics. With s years of schooling and x years of work experience, the log of earnings is

The coefficient r is the proportional rise in earnings from one more year of school, so with r = 0.09 an extra year raises earnings by e^0.09 − 1 = 9.4 per cent, not exactly 9 (the semi-log form compounds). The experience terms give a profile that rises quickly in the early years of work, flattens, and eventually bends down: workers learn on the job at a decreasing rate. The first chart below draws two such profiles, for workers with s and s + 1 years of school; the gap between them is the same proportion at every experience level, so it widens in rupees as pay rises. Across many countries and studies the estimated return to a year of schooling has been about 8 to 10 per cent; Psacharopoulos and Patrinos (2018), reviewing the literature, report a world average near 9 per cent a year of schooling, with higher returns in lower-income countries.

Two warnings. The estimate is a private return: it uses the worker’s own earnings and cost, and says nothing about spillovers to others (the social return). And more able or better-connected people may both study longer and earn more for other reasons, so the regression coefficient can overstate the effect of schooling alone (ability bias).

Becker’s decision rule is the one for any investment. A student weighs the cost of one more year, which is fees and books (the direct cost) plus the earnings given up while studying (the forgone earnings, an opportunity cost), against the stream of extra earnings for the rest of working life, discounted to the present. If the net present value is positive the year is worth taking. The private rate of return is the discount rate at which the net present value is exactly zero; the year is worth taking whenever the rate of return exceeds the student’s own discount rate (the rate at which they can borrow or would rather spend now).

Because the cost comes first and the gains stretch over decades, three things move the answer. A higher discount rate shrinks the later gains. Higher fees raise the cost directly, which is why cutting low-cost student loans makes college less attractive: the direct cost of attendance rises (JNU 2021). And a shorter working life after the year, as for an older student, leaves fewer years of gain to repay the cost. The calculator uses a worker who earns 100 a year, with a Mincer return of 9 per cent, fees of 10 and the year’s earnings forgone, over 40 working years: the net present value at a 6 per cent discount rate is about 25 and the rate of return about 7.8 per cent. Move the sliders to see which of the assumptions the answer is most sensitive to.

Michael Spence (1973) asked what an employer should do who cannot see a worker’s ability. Suppose there are two types, of low and high productivity, and that education is costly for both but cheaper per year for the high type. Employers can then use education as a signal. For the signal to separate the types, the level of education e must be high enough that the low type does not find it worth copying and low enough that the high type does. If the pay gap between the two types is G, the high type takes education if G is at least c_H e and the low type refuses if G is less than c_L e, so the separating levels satisfy

For example, if the high type earns 20 more and a year costs the low type 4 and the high type 2, any education between 5 and 10 separates them (exercise 4). In this equilibrium schooling does nothing for productivity and yet pays: private returns exist and the social return can be zero or even negative, since the schooling is a waste for the high type. Real schooling is surely both: it adds skills and sorts people. The two stories predict the same earnings gap and differ on policy: if signalling dominates, making everyone attend longer only raises the bar, while under the human capital view it raises output. Empirical work has looked for sheepskin effects (a jump in pay at the diploma) to separate them, with mixed results.

Health is capital in the same sense: it raises the number of years a person can work and how much they can produce in each. Two arguments link nutrition and work.

• The nutrition efficiency wage (Leibenstein, 1957): at very low incomes, extra food raises strength and work output, so a worker who is paid more is more productive. A firm may then pay more than the market-clearing wage, and unemployment can persist. Dasgupta and Ray (1986) built a model in which an unequal distribution of assets leaves the poorest unable to afford the food that would make them employable.
• The incentive efficiency wage (Shapiro and Stiglitz, 1984): if effort cannot be observed, a wage above the market level makes losing the job costly, so workers do not shirk; this holds wages above the clearing level and keeps some people unemployed. This is the answer to the opening question: it explains why urban formal-sector wages may exceed market-clearing levels even without wage controls, the same wage premium that the Harris–Todaro model of Lesson 4 takes as given.

Robert Fogel’s work on long-run records of height, weight and food supply (Nobel Memorial Prize, 1993) argued that better nutrition accounts for a substantial part of the rise in British output per head since 1790, through a larger labour force who could work harder. The survey of Strauss and Thomas (1998) summarises the micro evidence: better nutrition and health raise wages and productivity, with the strongest effects for the least well-off. The causation also runs the other way, since richer households can afford better food and care.

Four World Bank series let you compare countries. Gross secondary enrolment in India was 78.1 per cent in 2024, against 64.3 in Bangladesh (2024) and 48.3 in Pakistan (2024). A gross ratio divides all pupils enrolled at that level, of any age, by the population of the official age group, so it can exceed 100 where many pupils are over- or under-age (Brazil’s does); it is not the share of children in school. India’s adult literacy was 78.2 per cent in 2024. Government spending on education was 4.1 per cent of GDP in 2022, and current health expenditure, which counts private and public spending together, was 3.34 per cent of GDP in 2023.

Read them with care. The latest year differs by country, and the series differ in what they count (public education spending; total health spending, including what families pay themselves). A low share can reflect a low-cost system and a high share a costly one, so the figures are a prompt for questions, not a ranking of quality. The Right to Education Act of 2009, which makes elementary schooling a right for every child aged 6 to 14 (since April 2010), widened access; enrolment is not learning, which is why recent debate focuses on learning outcomes.

Sen’s argument (Lesson 1) is that treating people only as means of production, which is what human capital does, leaves out what makes health and education valuable to the people who have them. Education makes a person more efficient in production, which is human capital, and enriches life directly, through reading, communicating and informed choice, which the broader capability perspective values. The two are related as a set and a subset: capability is the larger concept, and human capital the part that operates through production. So education enhances both (UGC NET 2024). A second consequence is for public policy: a case for a school that rests only on measured returns is weaker than one that also counts what schooling does for a person’s freedom and for others.

Questions on human capital, returns to education, health and the capability link. Questions on the HDI and on Sen’s approach to development are in Lesson 2 and Lesson 1.