اندازه گیری ورودی سرمایه انسانی در سراسر کشور: یک روش مبتنی بر درآمد کارگر

I propose the aggregate output divided by the wage rate of an industrial laborer in a country as a measure of the aggregate human capital input for that country. I use this method to compare the human capital inputs for 45 countries of diverse output levels. I find that human capital input differs between the lowest-income and the highest-income countries by a factor of 2.2 or 2.8, depending on the inclusion of outlier countries. This is significant but small relative to the results from the method based on years of schooling.

Lucas (1988) and many others have proposed the differences in human capital input as a major source of output differences across countries. In order to evaluate this view, a measure of human capital inputs across countries is necessary. Conceptually, human capital input is the labor input in the production adjusted for quality in terms of skills and health. The difficulty in its measurement lies in this adjustment. It is not easy to say how much more (or less) human capital input an hour of work by a doctor in the US is when compared with an hour of work by a laborer in Nigeria. In this paper, I propose a method of measuring human capital input that relies on the laborer's income and conduct a measurement exercise based on this method. Broadly, there are two approaches to measuring human capital. One, the cost-based approach, measures the cost of human capital investment. For an international comparison of the human capital, the most common measure of the cost is the years of schooling. The measurement exercises based on the years of schooling include Kyriacou (1991) and Barro and Lee (1993). In comparing human capital input across countries, it is assumed that the years of schooling embodied in living (and working) people are proportional to the human capital input supplied by these people. This assumption is implicit, for example, in the growth accounting exercise of Mankiw et al. (1992). The popularity of the measurement method based on the years of schooling seems to stem from the fact that it directly relies on educational investment, which is considered a key element for human capital formation. However, this method has some shortcomings. First, it does not measure the human capital acquired outside the school: skills acquired before schooling, in job training outside the school and in the workplace. A worker with no schooling clearly has a human capital to the extent that he is contributing to the production. Skills acquired in the workplace, especially, may differ greatly between the workers in low-income countries and workers in high-income countries. Second, this method does not measure human capital in terms of health, which is an important factor in labor productivity. Human capital in terms of health may differ greatly between low-income and high-income countries. Third, the measurement using the years of schooling implicitly assumes that the formation of human capital per year of schooling is the same in all countries. The quality of education may vary greatly across countries, especially between the low-income and the high-income countries, leading to different quantities of human capital formation per year of schooling. Fourth, the measurement using the years of schooling implicitly assumes that the formation of human capital per year of schooling is the same at all levels of schooling. One can conjecture that the marginal formation of human capital decreases as the duration of schooling increases and is the same as the marginal cost at the point when schooling stops. This conjecture is supported by the finding that the return to primary education is higher than the return to secondary education, which is higher than the return to tertiary education (Psacharopoulos, 1994). The other approach, the income-based one, uses the labor income differences across workers with various levels of human capital to measure human capital inputs. Income differences across workers are the differences in the market values of their human capital inputs and are largely determined by the differences in their human capital inputs. The differences in the human capital input could then be derived from the income differences by eliminating the part of the differences due to the factors other than human capital input. For example, Krueger (1968) classifies workers by education level, age and sector where they work (urban or rural) in a sample of 21 countries, with the assumption that two workers of the same type, across countries as well as within a country, supply the same human capital input. She derives the aggregate human capital input for each country by weighing the inputs of different types of workers by their average labor incomes in the US. This method, unlike the method based on the years of schooling, allows the differences in human capital formation per year of schooling across education levels since these differences would be reflected in income differences. However, this method still assumes that differences among workers in the skills acquired outside schooling and in health are zero and the quality of schooling is the same across countries.1 A more recent example of the income-based approach to measuring human capital input is by Mulligan and Sala-i-Martin (1997). In a study on the changes of human capital inputs across the states of the US over time, the authors propose as a measure of aggregate human capital input for a given state in a given year of the aggregate labor income divided by the average income of the workers with no schooling for that year in that state. The implicit assumptions here are that a worker with no schooling supplies the same human capital input across the states and years, but that for any state in any year, human capital input differences across workers are proportional to their actual income differences and not some cross-state or cross-year representative income differences. If we were to adopt Mulligan and Sala-i-Martin's method for international comparison of human capital inputs, it would require less data and is, therefore, easier to implement than Krueger's method. Also, the assumption that workers with no schooling supply the same human capital input seems more defensible than Krueger's assumption that workers with any given level of schooling supply the same human capital: workers with no schooling across countries are more comparable in their human capital input than workers with positive and equal years of schooling since the differences in school quality are irrelevant to the human capital input of the workers with no schooling. However, this method would still have the shortcoming of ignoring the differences in the skills acquired outside schooling and in health among the workers with no schooling.2 In this paper, I adopt Mulligan and Sala-i-Martin's method, but with some modification, to measure human capital inputs across countries of diverse income levels. The modification is that I assume that the industrial laborer, as classified by the International Labor Office, rather than the worker with no schooling supplies the same human capital input across countries. I derive the aggregate human capital input for a country by dividing the aggregate labor income, which is assumed to be proportional to the aggregate income or output across countries, by the average income of the industrial laborers in that country. For international comparisons of human capital input, this method has advantages over methods based on schooling: methods using years of schooling and Krueger's and Mulligan and Sala-i-Martin's methods. By not using schooling as a way of comparing the workers altogether, this method avoids the problem of mismeasurement of human capital acquired outside schooling that is present in all methods based on schooling, the problem of international comparability of schooling quality that is present in the method using the years of schooling and Krueger's method, and the problem of comparability of a school year at different levels that is present in the method using the years of schooling. Partially for these reasons, the human capital inputs of industrial laborers across countries seem more comparable than those of workers with any given years of schooling, including workers with no schooling in Mulligan and Sala-i-Martin's method. Industrial laborers are workers who primarily supply their physical effort with little skill. Further, it is plausible to assume that these workers are physically fit to work in the industrial sector. Thus, in terms of health as well, industrial laborers seem comparable.3 One criticism against the method used in this paper may be that there may be factors other than human capital input itself that affect the wage rate of the industrial laborer relative to the other occupations, and these factors may differ across countries or years. The same criticism applies to the method based on the wage rate of a worker with no schooling in Mulligan and Sala-i-Martin (1997) since the method in this paper is an adaptation of theirs. However, some of this criticism is really against the concept of aggregate human capital and not against the methods of measuring it per se. For example, it is widely held that the technological change in the US over the last couple of decades has been skill-biased, raising the skill premium. This change would have increased the measured human capital input in the US under the method based on the laborer income or on the no-schooling worker's income even if there had been no changes in the supply of human capital input. Although this may very well be the case, it is not allowed under the assumptions supporting the concept of aggregate human capital. An example of a more legitimate criticism is the laws and regulations that make wage rates differ from the marginal products. The minimum wage law may raise the wage rate of no-schooling workers beyond the marginal product in some countries, leading to inaccurate measures of human capital input. However, the minimum wage law seems unlikely to be binding for the industrial laborer since his wage rate is not low relative to the other occupations. At any rate, the main concern of this paper is the difference in the human capital input between the low-income and the high-income countries and not a country by country comparison of human capital input. Thus, any factor that unduly affects the measurement would affect the main result only to the extent that it is systematically related to the income level. There do not seem to be such obvious factors. In Section 2, I present a model that shows the concept of human capital used in this paper and provides the basis for its measurement across countries in Section 3. In Section 3, I introduce the wage data set LABOCT from the International Labor Office (2000) and document the differences in the wage rate of the industrial laborer among 45 countries of diverse output levels. I then use these wage rate differences to derive the differences in the aggregate human capital input for these countries. I find that the human capital input differs between the lowest-income and the highest-income countries by a factor of about 2.2. This result contrasts with the human capital input difference between the lowest-income and the highest-income based on the years of schooling, which is about 3.5 for the same sample of countries. In Section 4, I show that in neoclassical output accounting, the human capital input difference between the lowest-income and the highest-income countries can account for the output difference between them by a factor of about 1.7. This is significant but small relative to their output difference, which is by a factor of about 30. Even if we add the physical capital input in the output accounting, the two inputs, human and physical, leave a large art of their output difference unaccounted for. In Section 5, I discuss the reasons why there are human capital input differences across countries as measured in this paper. Section 6 concludes.

In this paper, I proposed the aggregate output divided by the wage rate of the industrial laborer in a country as a measure of the aggregate human capital input for that country. The assumptions that support this method of measuring the human capital input are that the human capital inputs of the industrial laborers are the same across countries, that the human capital inputs of the workers within a country are proportional to their wage rates, and that the human capital input (labor) share of the aggregate output is the same across countries. The main advantage of this method over the methods based on schooling is that it takes into account cross-country differences in the skills acquired outside the school and in health status. I used the wage data set LABOCT from the International Labor Office to measure human capital inputs for 45 countries of diverse output levels. I found that, as one may expect, the low-income countries use less human capital input in the production and that the human capital input differs between the lowest-income and the highest-income countries by a factor of 2.2 or 2.8, depending on the inclusion of the outlier countries. This is significant but small relative to their output difference or compared to the results from the method based on the years of schooling. In neoclassical output accounting, this implies that the human capital input difference between the lowest-income and the highest-income countries can account for their output difference by a factor of about 1.8. Even if we add the physical capital input as an additional input of production, we need factors other than the human or the physical capital input to account for a large part of their output difference.