«Working Paper 2008-10-1 October 2008 Why Are Smart Cities Growing? Who Moves and Who Stays John V. Winters Georgia State University This paper can be ...»
W.J. Usery Workplace Research Group Paper Series
Working Paper 2008-10-1
Why Are Smart Cities
Growing? Who Moves and
John V. Winters
Georgia State University
This paper can be downloaded at: http://aysps.gsu.edu/uwrg-research.html
ANDREW YOUNG SCHOOL
OF POLICY STUDIES
Why Are Smart Cities Growing? Who Moves and Who Stays
John V. Winters*
October 8, 2008
This paper examines why smart cities are growing by exploring the relationship between the college share in a city and migration to and from the city. The results suggest that the greater in- migration to smart cities is almost entirely due to persons moving to pursue higher education.
Smart cities are growing because in-migrants often stay in the city after completing their education. The growth of smart cities is also mostly attributable to population redistribution within the same state and has little effect on population growth at the state level.
JEL Classification: R11, R23 Keywords: population growth, migration, human capital, college education, smart cities *Department of Economics, Andrew Young School of Policy Studies, Georgia State University, PO Box 3992, Atlanta, GA 30302-3992, Phone (404) 413-0268, Fax (404) 413-0270, E-mail email@example.com 1
1. Introduction A considerable body of literature has shown that the stock of human capital in a metropolitan area, measured as the share of the adult population with a college degree, is a strong predictor of future population growth.1 Berry and Glaeser (2005) also show that the share of the adult population with college degrees has increased more quickly in cities with higher initial levels of schooling.2 There is still no consensus, however, as to why “smart cities” are growing.
A popular hypothesis is that cities with high levels of human capital have higher productivity, perhaps in part due to knowledge spillovers. Several studies have shown that wages in highly educated cities are higher than in less educated cities, even after controlling for individual worker characteristics (e.g. Rauch 1993; Moretti 2004a; Glaeser and Saiz 2004), though the extent to which this represents knowledge spillovers is questioned by some.3 Another explanation for the connection between population growth and human capital is that an educated populace makes a city more attractive and people flock to the city for the higher quality of life (Shapiro 2006).
In this paper, I explore the reasons why smart cities are growing by looking at who moves to smart cities. Smart cities are often small and mid-size metropolitan areas containing flagship state universities. This suggests that students moving to pursue higher education may play an important role in the growth of smart cities.4 I investigate this hypothesis by examining the relationships between migration and the share of the adult population who are college graduates separately for individuals enrolled in higher education and for individuals not enrolled.
1 See, for example, Glaeser, Scheinkman, and Schleifer (1995); Simon (1998, 2004); Duncan and Henderson (1999);
Simon and Nardinelli (2002); Glaeser and Shapiro (2003); Glaeser and Saiz (2004); and Shapiro (2006).
2 Throughout this paper, I use the terms city and metropolitan area interchangeably.
3 For reviews of the literature on human capital externalities, see Moretti (2004b) Lange and Toppel (2006), and Henderson (2007).
4 De la Garza (2008) suggests that the growth of smart cities is not solely due to the growth of college towns, but his categorization of cities as college towns is a less direct way of examining the role played by enrollment than the approach taken here.
persons pursuing higher education. It seems likely, though, that many of those who move to pursue higher education will leave the metropolitan area after their education is complete, and I find that high human capital cities also have high rates of out-migration. On net, though, high human capital cities gain more people than they lose. The suggestion is that a large number of people relocate to pursue higher education and end up staying in the same city after finishing their education. This finding explains much of the population growth of smart cities in recent years.
I also find that the bulk of net migration to high human capital cities comes from within the same state. This has important implications for state policy makers. Population growth in high human capital cities does not equate to population growth at the state level. Instead, the growth of high human capital cities is largely due to population redistribution within the state.
2. A Closer Look at Smart Cities To gain a better understanding of why smart cities are growing, it is useful to understand why some cities have initially high levels of human capital in the first place. One possibility is that cities that surround major universities will have a more educated populace. Table 1 lists the top 20 cities (out of 331) by the share of adults (age 25 and over) with at least a bachelor’s degree in 1990.5 Iowa City, Iowa tops the list with an impressive 44.0 percent of adults with college degrees, while State College, Pennsylvania ranks 20th of the 331 cities with 32.3 percent.
Interestingly, a number of the most educated cities are relatively small metropolitan areas surrounding major public universities. Iowa City is home to the University of Iowa and State College is home to the Pennsylvania State University. Additionally, Boulder is home to the 5 Data come from the Department of Housing and Urban Development (HUD) State of the Cities Data System based on the 1990 Census of Population and Housing.
Kansas, Columbia to the University of Missouri, Bryan-College Station to Texas A&M, Gainesville to the University of Florida, Madison to the University of Wisconsin, ChampaignUrbana to the University of Illinois, Charlottesville to the University of Virginia, Bloomington to the University of Indiana, and Tallahassee to Florida State University.
The fact that many of the most educated cities are home to major universities suggests that much of the growth of smart cities may be attributable to the role smart cities play as centers of higher education. I explore this possibility in subsequent sections by examining the relationships between migration and the share of the adult population who are college graduates separately for individuals enrolled in higher education and for individuals not enrolled.
3. Empirical Framework Most studies interested in the growth of smart cities look at population growth over decennial years. Very few studies, however, examine who is moving to high human capital cities.6 In this study, I hope to gain insight on why smart cities are growing by looking at who moves to smart cities. I look at migration directly, instead of population growth as is usually
done. More specifically, I estimate:
where Min,j is the rate of in-migration to city j, Sj is the share of the adult population with college degrees in the city, Zj is a vector of other variables found in previous literature to affect city population growth, and in,j is a mean zero error term. Following previous literature, timevarying explanatory variables are measured with a ten year lag so that they are not affected by migration during the period under consideration. Because high human capital cities are often 6 To my knowledge Berry and Glaeser (2005) is the only exception. In that study, they look separately at logarithmic changes in the college educated population and the non-college educated population, but they do not look specifically at migration.
with the rate of out-migration and estimate:
Mout, j = outSj + outZj + out, j, (2) where “out” subscripts the out-migration rate and its corresponding coefficients and error term.
A city will grow if the rate of in-migration exceeds the rate of out-migration. Thus, I also
estimate the determinants of the net migration rate, obtained by subtracting (2) from (1):
Mnet, j = netSj + netZj + net, j, (3) where “net” subscripts the net migration rate and its corresponding coefficients and error term.
I first estimate the migration equations for the entire population, but this tells us little about who moves to high human capital cities. I, therefore, next estimate the migration equations separately for persons age 16 and over by whether they are enrolled in higher education. Computations from the Census 2000 Integrated Public Use Microdata Series (IPUMS) reveal that roughly 16 percent of all persons age 16 and over who lived in a different city in 2000 than in 1995 were enrolled in higher education in 2000. If high human capital cities are growing because individuals move there to pursue higher education, then migration by those enrolled in college may constitute a disproportionately large share of the total in-migration to high human capital cities. To add further evidence to my story, I also estimate the migration equations separately by five year age groups. I further explore who moves to smart cities and who stays by estimating separate equations for individuals who move from within the same state and individuals who move from another state or country.
4. Data The migration data used in this paper were constructed from the IPUMS (Ruggles et al.
2008) data for the 1980, 1990, and 2000 Censuses (5 % samples). The data for the percent of
employment in manufacturing come from two sources. For 1980 and 1990, the data for these variables come from the HUD State of the Cities Data System and are based on 1999 Primary Metropolitan Statistical Area definitions, while the 1970 data come from the 1972 County and City Data Book archived at the Inter-university Consortium for Political and Social Research (ICPSR) and are based on 1981 Standard Metropolitan Statistical Area definitions. Data on temperature and precipitation come primarily from the 2007 County and City Data Book where metropolitan area values were assigned based on the values for their principal cities. For cities missing information in the 2007 County and City Data Book, information was obtained from the 2000 and 1988 County and City Data Books. Metropolitan areas that crossed regions were assigned to the region in which the major principal city is located.
One complication with my analysis is that the IPUMS data do not allow identification of geographic areas with populations less than 100,000. As a result, the lowest level of identifiable geography in the IPUMS data, PUMAs in the 1990 and 2000 samples (county groups in the 1980 sample), often include both metropolitan and non-metropolitan areas.7 I, therefore, assign each PUMA (county group in 1980) to a metropolitan area if more than 50 percent of the population of the PUMA (county group) is contained within the metropolitan area. Using this procedure, identifies 323 metropolitan areas in 2000, 298 in 1990, and 276 in 1980.
A person is considered a migrant if they lived in a different metropolitan area in the census year than they did five years prior. Gross in-migration to a city was computed by adding up the total number of migrants to a city using person weights. The same was done to compute gross out-migration from a city. Note, however, that persons who exit the country are not in the 7 Note also that PUMAs of previous residence often include more than one PUMA of current residence. To have consistent metropolitan boundaries, PUMAs of current residence were aggregated to correspond with PUMAs of previous residence.
robustness suggest that international out-migration has little effect on the results. Net migration was computed as gross in-migration minus gross out-migration. Gross in-migration rates are computed by dividing gross in-migration by the population of the city, defined according to PUMA (county group) boundaries consistent with the migration flows, five years prior to the census. The same is done to compute gross out-migration rates and net migration rates. When I split the migrant flows by enrollment status, age, and state of previous residence, I continue to use the total population of the city as the population base to allow for easier interpretation of each groups contribution to the overall flow.
5. Empirical Results I first estimate the relationship between the share of adults with college degrees in 1990 and in-migration, out-migration, and net migration between 1995 and 2000.8 Table 2 presents the results of the estimating equations both with and without additional controls. The first column of Table 2 presents the effect of the college share on the gross in-migration rate. As seen, there is a strong positive correlation between the in-migration rate and the share of adults with a college degree. The coefficient estimate of 0.522 suggests that increasing the college share by 0.1 increases the in-migration rate by.05. If, however, many of those who move to highly educated cities do so to pursue higher education, we would expect the share of adults with college degrees to be positively correlated with the out-migration rate as well. This is exactly what we find in the second column of Table 2.9 Absent differences in fertility and mortality, a city grows if more people move to the city than leave the city.10 Therefore, the effect of the 8 Corresponding estimates for 1975-1980 and 1985-1990 are qualitatively similar and are available upon request.
9 However, other hypotheses could also explain the high out-migration from high human capital cities.
10 Fertility and mortality rates may also differ with the local human capital level, but I ignore such differences in this paper and instead focus on migration to and from high human capital cities.