چرا تعداد کمی از زنان در نیویورک (و بسیاری در مینیاپولیس) کار می کند؟ عرضه نیروی کار زنان متاهل در سراسر شهرستانها ایالات متحده

This paper documents a little-noticed feature of US labor markets—very large variation in the labor supply of married women across cities. We focus on cross-city differences in commuting times as a potential explanation for this variation. We start with a model in which commuting times introduce non-convexities into the budget set. Empirical evidence is consistent with the model’s predictions: Labor force participation rates of married women are negatively correlated with the metropolitan area commuting time. Also, metropolitan areas with larger increases in average commuting time in 1980–2000 had slower growth in the labor force participation of married women.

Women’s labor supply has, for good reason, been the object of extensive empirical study. After all, the dramatic rise in female labor force participation that occurred over the past 60 years in the US (and in many other countries) has been the most visible and important shift in the labor market. Also, women’s labor supply is often the margin of adjustment in households’ responses to policy shifts, e.g., changes in the taxation of household income or welfare entitlement programs, and thus holds the key to proper policy evaluation. Although many empirical studies of female labor supply have been conducted, it appears that an interesting, potentially important feature of the US markets has gone largely unnoticed: There is wide variation in female labor supply across metropolitan areas in the United States. Consider, for example, one large group of women: married non-Hispanic white women aged 25–55 with a high school level education (found in the 2000 US Census). In Minneapolis 79 percent of such women were employed, while in New York the proportion is only 52 percent. The cross-city variation in female labor supply within the US that we document in this paper is as large as the well-known and widely studied variation across OECD countries in female employment rates. For instance, by way of rough comparison, one might look at employment rates among women with “upper secondary education” from selected OECD countries: United Kingdom, 80 percent; Sweden, 78 percent; Netherlands, 74 percent; France, 71 percent; Canada, 69 percent; US, 66 percent; Italy, 64 percent; Japan, 59 percent; and Germany, 52 percent. (These statistics, for women aged 25–64, are from OECD, 2007.) In an effort to make sense of international comparisons, analysts typically focus on policy differences across countries (in paid parental leave, marginal taxes, employment protection, welfare benefits, etc.).1 Such policy differences, of course, are much smaller across locations in the US than OECD countries. The cross-city variation in female labor supply in the US is apparently generated by characteristics of the local markets themselves. Furthermore, while the labor supply of women has increased substantially in all cities in the US over the past 60 years, there have been big differences in these cities in the timing and magnitude of the increase. Fig. 1 illustrates, for 1940 through 2000, the well-known large increases in the labor supply of married non-Hispanic white women generally, and shows also how different the paths are for two particular urban locations, New York and Minneapolis. In 1940 the labor supply among women was lower in Minneapolis than in New York, but the subsequent growth in female labor supply was much more rapid in Minneapolis than in New York, leading to the large disparities observed in 2000. These results are especially interesting in light of the ongoing discussion about the possibility that the US labor market has now achieved a “natural rate” of female labor force participation.The goals of this paper are to carefully document the cross-city variation in married women’s labor supply across US labor markets, to explore potential economic explanations for observed cross-city variation in married women’s labor supply, and to examine the implications for the study of female labor supply generally. We believe that many factors are at play in producing the large observed local variation in female labor supply across the US, but, we argue, one explanation stands out: Married women, particularly married women with young children, are very sensitive to commuting times when making labor force participation decisions. In building our argument about the importance of commuting cost, we start with the theory of labor supply when there is a fixed cost of participation (i.e., commuting time). The introduction of a fixed cost of participation introduces non-convexities into the budget constraint. This complication is easily handled in a one-period case for a one-person household: Assuming leisure and consumption are normal, and assuming also that initially the individual is at an “interior solution,” an increase in the fixed cost reduces both leisure and labor supplied, up to a threshold at which the individual moves to a “corner solution” of supplying zero labor. Matters are more interesting in a model in which a two-person household takes a “collective” approach to labor supply. In this case, increases in the commute time can induce one partner (traditionally the wife) to move out of the labor force while inducing the other partner (the husband) to work longer hours. As mentioned above, there are many studies of women’s labor supply. Blundell et al., 2007 and Blundell and MaCurdy, 1999 provide valuable discussions of key issues in this literature, and Killingsworth and Heckman (1986) overview earlier results. Most studies use national data, with results aggregated at the national level, and no attention is given to the possibility of meaningful local variation. A small body of work in economic geography does provide some evidence about cross-location variation in labor supply (e.g., Odland and Ellis, 1998 and Ward and Dale, 1992), but this work does not seek to provide an explanation for the observed variation. Specifically, we know of no work that posits the importance of fixed commuting costs for explaining cross-city variation in labor supply and then evaluates predictions empirically. 3 There are a number of papers in urban economics relevant to the arguments that we develop below. Most prominently, many papers examine commuting generally,4 and the possibility that commuting plays a significant role for women specifically. For example, Rosenthal and Strange (2012) give evidence from 2000 US Census data that women commute shorter distances than do men—arguing that women with children might be particularly likely to want to work closer to home. They suggest that the preference for relatively shorter commutes by women in turn drives gender differences in business locations of male and female entrepreneurs; specifically, this provides an economic rationale for an observed pattern in which women entrepreneurs locate businesses in less agglomerated locations than do their male counterparts. As Rosenthal and Strange (2012) note, their work draws on earlier work on gender differences in commute times, e.g., Madden, 1981 and White, 1986. Against this backdrop, we carry out our analysis in five additional sections: Section 1 provides the basic facts about the city-specific employment rates of non-Hispanic white married women in 50 large US metropolitan areas from 1940 through 2000 using Public Use Samples of the US Census. We document significant variation across cities in current levels of women’s employment, and also substantial variation across cities in the magnitude and timing of the increase in female labor supplied over the past 60 years. Section 2 is a discussion of economic forces that might serve as potential explanations for the observed cross-cities variation in women’s labor supply. We argue that the variation in observed employment rates are unlikely to be due primarily to differences across cities in labor demand. Section 3 contains the primary economic contribution of this study. We develop an argument about the effect of cross-city differences in commuting times (owing, for example, to differences in congestion across cities) for labor force participation. Our model allows us to examine the effect of commuting time on individuals’ and households’ labor supply. Section 4 presents empirical evidence concerning the predictions of the model. The cross-sectional evidence indicates that in cities with longer average commuting times, female labor force participation rates are lower. Women with young children are particularly sensitive to longer commute. We try a simple IV strategy (using location of birth as an instrument) in an effort to deal with potential endogeneity of location, and find that this does not change our key empirical findings. Also, results are similar for a specification that looks at differences over time (from 1980 to 1990 and 1990 to 2000). These results are all consistent with the theory presented in Section 3. Finally, Section 5 provides a conclusion and discussion of directions for future research.

Using Census Public Use Micro Sample (PUMS) data for 1940 through 2000, we find wide variation in labor market participation rates of white married women in 50 large US metropolitan areas. This wide variation is found in all years and appears for women with different levels of education, as well as for women with and without children. Among a number of explanations for observed cross-city differences in female labor supply one emerges as particularly important: Married women’s labor force participation decisions appear to be related to MSAs’ commuting times. While the possibility that fixed employment costs might matter for female labor supply was raised in the literature many years ago (e.g., Oi, 1976 and Cogan, 1981), we know of no work that has explored implications for the cross-city variation in labor supply. With this in mind we have undertaken an analysis that, first, sets out a simple theory of labor supply. In our theoretical exploration, we emphasized that an increase in the commuting cost c can be expected to reduce labor force participation of married women, while at the same time increasing the number of hours worked by working husbands. Cross-section evidence is consistent with the theory. We find a negative association between commuting time and women’s labor force participation rates in the three decades for which we have commuting time data (1980, 1990, and 2000). Similarly, the negative correlation between commuting times and participation appears also in a differences-in-differences analysis; metropolitan areas that experienced relatively large increases in average commuting times between 1980 and 2000 experienced slower growth of labor force participation of married women. For women, the effects of commute times are quite large. From Table 5 we see that a 1-min increase in an MSA’s commute time is associated with an approximately 0.3 percentage point decline in the labor force participation of women with a high school education. 38Table 4 indicates that the difference in commute times, from the shortest-commute MSA to the longest-commute MSA, is 33 min. Taking our estimates at face value, this variation might be expected to lead to a 10 percentage point difference in participation across cities. Commute time differences across MSAs thus plausibly “explain” a fair amount of the cross-MSA variation in participation observed in Table 1. It might be reasonable to ask, indeed, if our estimates are “too big”. One might reason that a 30-min increase in commute time presents an opportunity cost of only $10 daily for a woman who earns $20/hour. Could such a modest factor plausibly have such a large impact on behavior? In thinking about that issue it is important to recognize that the cost to a typical mother of living in a high-commute MSA extends far beyond the increased time she spends commuting if she works. As we noted above, in congested cities there will be additional time required for travel to the grocery store, piano lessons, or little league baseball. If she chooses to commute a long distance to work, a mother will likely be farther from her children’s school, which can make life difficult when she needs to be available for a parent-teacher conference or a class play. Our estimates are not picking up the pure effect of the commute time of the participation decision, but the total impact of congestion on the participation decision. One implication of our work concerns the century-long increase in the female labor force participation—the increase in married women’s participation from only about 7 percent in 1900 to current rates (near 70 percent). There are doubtless many factors contributing to this trend, many of which have received careful examination in the literature. Little attention has been given, though, to the possibility that part of this trend is due to the reduction in commuting costs, owing to improvements in transportation technology—the expansion of modern public transportation, the introduction and continued improvement in automotive technology, improvements in roads, and so on—and changes in residential patterns. Our findings about the wide cross-city variation in the labor force participation of married women also introduce a new dimension to the current discussion about trends in the female labor supply. In particular, these findings complicate discussions about women having reached a “natural rate” of labor force participation. The issue is how close to 1 we can expect this participation rate to be. Goldin, 2006 and Juhn and Potter, 2006, and others show that labor force participation rates depend on a combination of demographic factors such as age, presence of children, education, and race. The “natural rate” of participation is expected to be different for different groups. Our research suggests that the maximum achievable rate of labor force participation for each group would also vary across cities (and also across countries) because of differences in commuting time. Of course, commuting times in local communities also depend on population density, the resources devoted to transportation, and local planning (e.g., zoning laws that may sometimes serve to isolate residential communities from job locations). Thus, from a public policy perspective, it may be that targeted actions that reduce commuting times would thereby increase labor force participation by women.39 Yet another open policy issue concerns the importance of variation in labor supply across cities for tax and welfare policy. It would be interesting to analyze how differences in the time cost of commuting affect labor supply responses to changes in such policies. Finally, we note that most empirical research in labor economics is conducted at the national level, with little attention given to the possibility that local labor markets differ in important ways. Our research points to the value of work that allows for the possibility of differences across locations in labor supply responses. More generally, there is surely a rich set of interesting issues yet to be examined around the interactions of urban characteristics and labor market outcomes.