10.1Modeling Income Inequality
In the literature the general model evaluating the effects of different variables on income inequality, say at the state level, is defined as:
Inequalitys = Xsβ + es(9.1)
where Inequality represents a measure of income inequality in state s (usually the Gini coefficient or the variance of the log of income), Xs is a set of variables assumed to affect income inequality (such as demographic, economic and human capital factors), and e s represents the error term. The sign and significance of the coefficients of vector β express the effect of a given independent variable on income inequality. A positive (negative) sign is associated with a variable that increases (decreases) income inequality.
The Gini coefficient is not a perfect measure of inequality,12 but it is well known and has been computed at the state level for each decennial census of the U.S. population since 1950. This index represents the proportion of total income that must be redistributed in order to achieve perfect income equality among classes of income population. It varies between 0 and 1, from the lowest to the highest degree of inequality, respectively. Figure 10.1 shows a detailed description on the construction of the Gini coefficient. This study covers the 48 contiguous states in 1990. The values of states’ Gini coefficients for the year 1989 are taken from LRP. As stated by the authors, the Gini coefficients for 1989 are matched with 1990 values of the explanatory variables “because the Bureau of Census obtains income information on families for the year prior to the year the census is conducted.”
In order to replicate approximately their results, this analysis uses several of the independent variables that LRP considered as explanatory variables. First, the industrial mix of a state’s workforce may influence the distribution of income. For several reasons such as a strong union power, less skilled workers may earn higher wages in mining, construction and manufacturing industries. Income inequality may then be lower in states that have a high share of their working population in these industries.
Second, LRP cited Kuznets’(1955) theory, which stated that income inequality may be related to the level of economic development, and “income inequality will increase as income becomes concentrated in the hands of the owner/capitalist class.” However, after some threshold, the level of inequality should decrease. In order to account for this possibility, variables measuring income and income squared are introduced.
Third, demographic factors such as the labor force participation rate or the racial composition of a state may influence income inequality. Indeed, because wages and salaries are the major component of money income, income should be more equally distributed where the labor force participation rate is higher. On the other hand, racial discrimination may increase income inequality. This effect can be captured with a variable representing the state percentage of non-white population. LRP used the percentage of the population that is black, but I think the percent of non-white is a better measure since segregation may exist for all non-whites, not only blacks.
Human capital is also an important factor influencing income inequality. LRP considered two measures of educational level for the states. First, the percent of the 25 year-old population that graduated from high school (but not from college), and the percent that graduated from college. The literature on income inequality has related the importance of human capital in reducing income inequality. Indeed, as the labor force gets more educated, workers get more skilled, which allows them to earn higher salaries. This will reduce the gap between low-paid and high-paid jobs, thus reducing income inequality. However, the effect of the percent of college graduates is less obvious. On one hand its increase can contribute to the reduction of income inequality, as it is the case for the percent of high school educated workers. On the other hand, its increase can aggravate income inequality simply because it increases the upper bound of the wage distribution, because highly educated people earn higher wages. Therefore the effect of an increase in the percent of college graduates on income inequality is ambiguous. All these variables are defined formally in Table 10.1, described in Table 10.2, and will be used to replicate approximately the results obtained by LRP. Then, the most significant variables will be retained, and some IT variables added to the model in order to analyze the effects of IT on income inequality.
As stated in chapter 2, the increase in IT activities over the last two decades has paralleled the increase in income inequality. Figure 2.11 reports similar trends in the Gini coefficient and the IT capital stock between 1977 and 1997. For various reasons, the demand-side explanation suggests that the increase in IT has somehow deteriorated the situation regarding income inequality. This is also suggested by Figure 2.12, which shows a scatter plot of the Gini coefficients and the stock of IT capital stock for the 1977-1997 period. The main reason may be the substitution and complementary effect of IT. Indeed, IT capital might substitute for unskilled workers, and be a complement for skilled employees. Still, the causality between IT and income inequality is not clear and must continue to be tested empirically.
In this analysis, I will focus on IT as a characteristic of the labor force, not the capital stock. First, I will consider IT employment and non-IT or “traditional” employment as the main variables for which I want to analyze the effect on income inequality. IT employment is defined as the number of employees in 16 types of 2-digit SIC industries that are considered as IT industries, as defined in chapter 7. Employees in these industries are believed to deal with information and knowledge more than any employees in the other industries. Furthermore, IT employment refers to jobs that are usually paid better than non-IT employment. The reason is that the economy moved from an “industry based” to “information based” paradigm. Today, the term “new economy” refers to an economic system where knowledge and information are the new “raw materials.” Some authors relate this change as the passage from a “hard” to a “soft” economy. Hence, workers who are able to create, process, transmit and analyze information will have a higher marginal product than non-IT workers. As a consequence, salaries of IT employees are usually higher than those of non-IT employees. On the other hand, since the ability to deal with information technology has become the scarce skill in the job market, workers that do not possess such knowledge have a lower marginal product. Therefore, states with a high percentage of IT employees are most likely to exhibit higher income inequality, since the “IT knowledgeable” workers get richer and the “IT ignorant” workers poorer.
Finally, the density of IT employment may also affect income inequality. Indeed, information is subject to externalities that increase with density. A university for instance, is a place where people exchange ideas and benefits from knowledge spillovers due to interactions with one another. Other IT workers could benefit from being close by. These externalities should increase with the agglomeration of IT employees. Therefore, the marginal product of IT employees should increase with the density of this type of employment, aggravating income inequality even more.
After having defined the models measuring the impact of IT employment on income inequality, the next section will now describe the variables and data needed, as well as some descriptive statistics.