A Companion to Urban Economics - Arnott and McMillen - Chapter 2

24 A Companion to Urban Economics Edited by Richard J. Arnott, Daniel P. McMillen Copyright © 2006 by Blackwell Publishing Ltd C H A P T E R T W O Human Capital Externalities in Cities: Identification and Policy Issues Gilles Duranton 2.1 INTRODUCTION The case for corrective economic policies as stated in Econ101 is relatively straight-forward. In some instances, there is a wedge between the private costs (or bene-fits) resulting from some choice made by an economic agent and the social costs (or benefits) accruing to society. Such a wedge occurs when markets fail to medi-ate properly some economic interactions; that is, when there is an externality. In such a case, the privately optimal decision made by the agent does not lead to a socially optimal outcome. For instance, investors underinvest when they cannot appropriate all the positive returns from their investments. Firms overproduce goods, whose production damages the environment at no cost to them, and so on. Pigou’s (1920) ingenious solution to this type of problem is to impose a tax (or a subsidy) so that the private and social costs are made equal. In other words, whenever there is an externality, an appropriately chosen Pigovian tax or sub-sidy can make agents internalize the external effects of their choices so that the privately optimal decision leads to a socially optimal outcome. After exposing the details of this argument, the Econ101 textbook typically points out that such taxes and subsidies are difficult to implement empirically because of their informational requirements. Then the textbook usually ends the discussion on this topic and turns to something else. From a policy perspective, HUMAN CAPITAL EXTERNALITIES IN CITIES 25 however, this is where the real work ought to start. In this chapter, the objective is to show how in practice economists attempt to identify such externalities. With this in mind, this chapter will analyze the case of one set of externalities: those pertaining to the market for human capital and education. Human capital externalities are of particular interest to urban economics, as shall be made clear below. The rest of this chapter is organized in the following way. Section 2.2 shows why human capital externalities in cities matter. Section 2.3 outlines a simple theoretical model and shows how it has been estimated empirically. Section 2.4 highlights a first set of criticisms to the standard approach. Section 2.5 discusses crucial issues of model identification and shows how they matter for policy pur-poses. Section 2.6 describes alternative approaches for the estimation of human capital externalities in cities. Finally, section 2.7 concludes. 2.2 WHY STUDY HUMAN CAPITAL EXTERNALITIES? WHY IN AN URBAN CONTEXT? Among all the externalities that economists have been thinking about, human capital externalities are “special” in two respects. First, they are potentially of formidable importance for a number of reasons. Such externalities provide a strong justification for subsidies to education. If the private returns to education are only half the social returns, the optimal Pigovian policy is a 50 percent sub-sidy to education. If, instead, the social returns are essentially equal to the private returns, no subsidy is needed. Given that most developed economies spend up to 10 percent of their income toward education and training (broadly construed), the numbers at stake are very large. It is worth noting, however, that human capital externalities are important beyond the issue of the optimal split between public and private expenditure for education. As argued by Lucas (1988) and his followers, human capital externalities could constitute the fundamental engine of growth and development. If this were the case, governments could draw on educa-tion policy to speed up economic growth. Moreover, since Marshall (1890), human capital externalities are also accepted as one of the main reasons to justify the existence of cities. This is because human capital externalities may arise pre-dominantly from direct (or face-to-face) interactions between people, which are themselves expected to be highly distance sensitive. The second reason why human capital externalities are special is that they are particularly difficult to identify. To repeat, we speak of an externality when market prices fail to reflect the true social costs and benefits of an action. Unfor-tunately, most of the data collected around the world consist of simple measur-able characteristics of economic agents or of recorded market transactions. Since prices and quantities only reveal private costs and benefits, externalities, by their very nature, leave no obvious paper trail by which they can be tracked or measured. This being said, with some externalities, social costs and benefits may be meas-ured indirectly without too many conceptual difficulties. For instance, the costs of urban congestion can be measured by counting how many people are stuck in 26 G. DURANTON traffic jams and estimating how much time is wasted there. The mechanism at play is relatively simple: by taking the road at peak hours, I slow down every-body else and I do not take this into account when entering my car. The social cost of traffic congestion is the cost of the time wasted by people stuck in traffic jams plus that of an increase in pollution. Admittedly, measuring these costs precisely is by no means an easy task, but it is still feasible. Urban congestion is a reasonably well-circumscribed problem, with regard to which transport eco-nomists have made very significant advances. Human capital externalities are much more problematic, because the mech-anism at play is far less obvious. As shown below, there are many mechanisms that can generate human capital externalities. These mechanisms call for different policy prescriptions. In some extreme cases, subsidizing education may even be counterproductive. The fact that the social costs and benefits of education have many dimensions complicates the matter even more. Human capital externalities can be thought of having a positive effect on productivity and wages as well as criminal behavior or even voting outcomes. In what follows, the discus-sion will be restricted to the effects of human capital externalities on wages and earnings. To look at human capital externalities, the starting point of existing research is the following. Positive human capital externalities imply that measures of aggregate human capital should matter in the determination of outcomes over and above individual characteristics. In the absence of experiments on the issue, there are two main avenues for research: cross-section or time series analysis. Time series analysis does not appear to be particularly appropriate. Isolating the effects of an increase in education on aggregate output and confronting this to the private returns to education may be a hopeless task because of the incredibly large number of confounding factors that may affect changes in aggregate output over time. Instead, most research focuses on cross-section analysis conducted either at the cross-country or at the subnational level. Cross-country analysis is made very difficult by the large number of institutional factors that may affect the outcomes of different countries. The second major problem is that comparing education data across countries is also very difficult. Subnational analysis seems “easier” to conduct because in many countries labor-market data sets are of good quality and have become widely available. Such data typically records individual wages, education, and location for large samples of workers across a given country. Among existing subnational units, cities are of particular interest for two reasons. First, as highlighted above, human capital externalities may be at the root of the existence of cities. These externalities are thus expected to manifest themselves strongly at this level of analysis. Second, urban areas when properly defined provide economically meaningful economic units of analysis as opposed to arbitrarily defined administrative regions or states. Note, however, that by conducting the analysis across cities we give up on any attempt to measure country-wide human capital externalities. This may not be a serious issue when looking at wages or crime (which may, to a large extent, be determined locally), but it is potentially more problematic when we are interested in voting behavior. HUMAN CAPITAL EXTERNALITIES IN CITIES 27 2.3 THE STANDARD APPROACH TO THE ANALYSIS OF HUMAN CAPITAL EXTERNALITIES IN CITIES Consider an economy with workers (subscripted i or j) living in cities (subscripted a). The social output of worker i with human capital hi and living in city a is given by yi = (A + B)hi, (2.1) where A is a technological parameter independent of location and Ba is a city-specific parameter. At the same time, the earnings of this worker are wi = Ahi + Da. (2.2) A straightforward comparison of equations (2.1) and (2.2) shows that workers do not receive the full value of their social product. Worker i cannot appropriate the part of her social output given by Bahi. At the same time, however, this worker benefits from being in city a and receives D as part of her earnings. This latter quantity will receive different theoretical interpretations in what follows. For the time being, it is convenient to think of it as the part of the external output of the other workers in the same city that accrues to worker i. Put differently, there is a reciprocal externality within cities: workers do not receive the full surplus that they create, but instead receive part of the surplus created by the others. Note that the specification of equations (2.1) and (2.2) is additive rather than multiplicative. This is mostly for simplicity and does not matter here (it does, however, in empirical work, where multiplicative specifications are preferred because they fit the data better). It is also useful to bear in mind that empirically we can only observe the wage wi, some proxy for human capital hi, and a few aggregate variables relating to city a, but not the social output yi. To finish the description of the model, assume that the cost of human capital, hi, is Ci = cihi, (2.3) with a > 1. The cost shifter, ci, can vary across individuals to reflect their differences in intrinsic abilities. Note that in a static context this cost may be interpreted as both a cost of acquisition and maintenance of human capital. At the free-market equilibrium, the investment in human capital of worker i is chosen so as to maximize wi − Ci. After simplification, we find equilibrium invest-ment to be equal to the following: hi = A 1/(a−1) . (2.4) i The socially optimal human capital investment is, instead, such that it maximizes yi − Ci. It is given by 28 G. DURANTON h* = A + B 1/(a−1) . (2.5) i The optimal Pigovian tax is to subsidize the returns to human capital in equation (2.2) by offering worker i an amount B per unit of human capital. Equivalently, one may subsidize the cost of acquiring human capital by a fraction B/(A + B) so that the worker faces a cost of acquiring human capital equal to C = Ac ha/ (A + B) rather than equation (2.3). What are human capital externalities about in this model? The standard story about human capital externalities in cities is eloquently summarized by Lucas (1988): “Most of what we know we learn from other people. We pay tuition to only a few of these teachers, either directly or indirectly by accepting lower pay so we can hang around them, but most of it we get for free, and often in ways that are mutual – without a distinction between student and teacher.” Indeed, to write this chapter I have read and benefited freely from a nice survey on human capital externalities in cities by Enrico Moretti (2004b), from whom I borrowed the above quote. To write this chapter, I also built on previous work conducted with Sylvie Charlot (Charlot & Duranton 2004). Again, she did not receive any direct compensation, despite contributing to this chapter indirectly. Arguably, such external effects take place across the board, in many industries and not only academia. To be more precise about human capital externalities in an urban context, assume that worker i’s human capital directly benefits N other workers in the city by an amount bhi. This “interaction group” of N workers with whom worker i interacts is assumed to be a representative sample of workers in city a. At the same time, worker i also benefits from the human capital investment made by all other workers in the interaction group. With our notations, summing across all workers j who are part of the interaction group of worker i, this implies B = bN and Da = å bhj = bNJa, where Ja is the average human capital in city a and N is the (unknown) size of the interaction group. Equation (2.2) can thus be rewritten in the following manner: wi = Ahi + bNJa. (2.6) This equation (as well as many closely related specifications assuming different functional forms) can be estimated by means of regression analysis. The data needed for this exercise must be at the individual level. This data must contain the wage of each worker, a set of human capital characteristics (such as school-ing, but also labor-market experience, etc.), and possibly further individual con-trols. To estimate equation (2.6), a set of aggregate (i.e., city-level) characteristics is also needed. Average schooling (or the fraction of university graduates) in the city is of course of particular interest here. The size of the interaction group may be taken as constant across cites or may be expected to increase with city popu-lation. In this case, the coefficient on city population can also be informative about the extent of human capital externalities. ... - tailieumienphi.vn