Job Decentralization and Residential Location

This paper addresses an old and central question in urban economics: how does the spatial distribution of employment opportunities influence residential location? Over the past fifty years, both employment and population left central cities for the suburban ring. Between 1960 and 2000, the share of metropolitan Americans who lived in the suburban ring increased from 48 to 68 percent. Over the same period, the share of metropolitan residents who worked outside the city rose from 41 percent to 58 percent. The decentralization of employment and population has led economists to ask whether workers followed jobs out to the suburbs or jobs followed workers. Answering this question is complicated by the fundamental simultaneity of the location decisions of workers and firms.

We adopt a novel approach to disentangling the causal relationship between the location of employment and population. Our main focus is on a single industry—namely, state government—whose primary location is predetermined with respect to current residential patterns. State government is concentrated in capital cities. The choice of capital city was established long before the process of suburbanization began. In many cases, core state buildings, such as the state capitol and the state supreme court, were built in the historic central business district (CBD) well over a century ago and have never been moved. As a result, state workers in capital cities are disproportionately likely to work downtown compared with other workers in the area (75.6 percent versus 55.4 percent). However, state employees are also present in every major metropolitan area. State workers in noncapital cities are not much more likely than other workers in the area to work in the central city (57.9 percent versus 52.0 percent).¹

If job location is an important determinant of residential location, we expect state workers in capital cities to be more likely to live in the central city relative to either state workers in other metropolitan areas or private sector workers in the capital . This set of comparisons naturally suggests a difference-indifferences estimation strategy. The first difference contrasts state workers in capital and noncapital cities to control for socioeconomic characteristics and differences in the taste for urban living that may be unique to state employees. The second difference contrasts state and nonstate workers within metropolitan areas to control for any relevant structural differences between capital and noncapital cities or their residents.

Our main empirical investigation is based on individual census records from the 1980 Integrated Public Use Microdata Series, or IPUMS (Ruggles and others 2008).² We find that state employees in capital cities are 12 to 14 percentage points more likely than state workers elsewhere to work downtown, relative to others in their metropolitan area, and 3 to 4 points more likely to live downtown. We find no observable differences in age, gender, education level or wage rates that could account for this residential pattern. Furthermore, this residential gap is robust to controlling for the industrial distribution of state workers in capital and noncapital cities. These figures imply that adding 1,000 jobs to the typical central city would increase the number of working residents in the central city by approximately 250 people.

Similar to state capitals, other government or government-related employment is characterized by historically determined locations that are very difficult to alter. We extend our basic analysis by considering employment in defense-related industry and at the United States Postal Service. Shortly after the Second World War, Congress specified that defense contractors should locate outside of existing city centers as a countermeasure against conventional or nuclear attack (O'Mara 2006). We use census data to identify workers who are likely to have been affected by this policy and demonstrate that they were significantly more likely to both work and live outside of central cities in 1980.

In prior work, we used the location of postal employment to study the role of spatial mismatch in black employment outcomes (Boustan and Margo 2009). Central to our earlier analysis is the fact that postal processing and distribution plants were located in or near central business districts before...