Brief Rationale for This Database
Why focus on the prime-age employment rate for particular place? A more detailed rationale is provided in the previously-cited Bartik report, but here is a brief summary: The rationale for focusing on the prime-age employment rate is that it is a key driver of economic and social well-being for a particular place. The prime-age employment rate roughly controls for age, by focusing on a particular age range, from ages 25-54. By focusing on this age range, the prime-age focuses on a group that is generally expected to work, and avoids most complications due to individuals being in school or retired. A higher prime-age employment rate will directly increase earnings per capita, by increasing employment, and will indirectly increase earnings per capita by putting upward pressure. Furthermore, a higher employment rate will reduce social problems, such as substance abuse, crime, and family breakups.
But why focus on places with low employment rates, distressed places? There are two distinct types of distressed places: “local labor markets”, and “neighborhoods”. The distressed place problem for “local labor markets” and “neighborhoods” are different in their causes and solutions.
“Local labor markets” are defined as multi-county areas that encompass most commuting flows. Because of extensive commuting, changes in labor market conditions in one part of a local labor market are transmitted relatively quickly across a local labor market.
In a distressed local labor market, with low employment rates, the problem is jobs are not readily available. The solution is to create jobs. In a distressed place, creating jobs will have greater social benefits, by increasing employment rates more. In contrast, in a non-distressed place, job creation will not increase employment rates as much, and will instead lead to more in-migration. Hence, targeting job creation at a distressed local labor market will have greater benefits per job created.
The United States does not have one agreed upon definition of local labor market. In this database, two definitions of local labor market are used, and prime-age employment rates are reported for both of these two definitions.
Definition 1 is “commuting zones”, which are a geographic area that was originally defined by staff at the U.S. Department of Agriculture, which classified all counties in the United States as part of a commuting zone. The definitions used here were updated by researchers at Penn State.
For political reasons, we further divide commuting zones at state boundaries, to form what we call “state commuting zones” or SCZs. This reflects the political reality that many policies to aid local labor markets will need to involve state governments.
Definition 2 is based in part on “core-based statistical areas”, which are metropolitan areas or micropolitan areas defined by the U.S. Office of Management and Budget. CBSAs are also multi-county areas that include most commuting flows, but tend to be defined somewhat more narrowly than commuting zones.
There are two problems with CBSAs. First, they stretch across state boundaries, which is politically inconvenient for targeting given the importance of state governments. To deal with this problem, our analysis here divides CBSAs at state boundaries. Second, CBSAs do not include all U.S. counties. To deal with this problem, any county that is not in a CBSA is defined as its own labor market. The resulting set of areas – CBSAs divided at state boundaries, plus individual non-CBSA counties – is labeled in this data base as “state-based local labor markets”, or SLLMs.
Both SCZs and SLLMs have the further problem that in general, they may not correspond to any political entity with any significant political power. To deal with this problem, this database also reports data on the prime-age employment rate for the state as a whole, and for each county in the state. The state is always a political entity, and in most states, counties are also a political entity. Furthermore, with the county data, a user of this database can combine counties to form their own definition of “local areas”, in a way that may correspond to areas used for state planning or administrative purposes, for example the areas used by local job training agencies.
In addition to “distressed places” that are local labor markets, the U.S. has distressed places that are “neighborhoods”. Although there is no universally agreed upon definition of “neighborhood”, it is often proxied by using data on census tracts, which are geographic areas that typically have a population of around 4,000.
“Neighborhoods” are not local labor markets, in that most jobs located in a neighborhood are not held by neighborhood residents, and most neighborhood residents do not work in their neighborhood. However, neighborhood distress due to low employment rates is still a problem. Research suggests that child development tends to be worse in neighborhoods with low employment rates, that is children growing up on such neighborhoods tend to have worse outcomes as adults. These worse outcomes may be due to low employment rates contributing to higher local crime rates and substance abuse, and fewer adults who are well-linked to job networks.
The neighborhood distress problem is that some neighborhoods have much lower employment rates than others, which worsens prospects for children growing up in such neighborhoods. As discussed in the Bartik report, such neighborhood problems are not best dealt with by creating new jobs in the distressed neighborhood, which is neither a necessary nor a sufficient condition for increasing the neighborhood’s employment rate. Rather, adults in the neighborhood should be helped by increasing “job access”, where “access” is broadly defined to include not only better transportation links to jobs throughout the local labor market, but also help with information on job openings, job training, and access to childcare and other support services.
To focus on the “neighborhood distress” problem, this report includes data not only on a census tract’s prime-age employment rate, but also on its differential from the prime-age employment rate of its local labor market. The differential focused on is from the SLLM average, but each census tract also reports prime-age employment rate for the overlying SCZ, state, and county. This focus on the differential distinguishes between neighborhoods whose low employment rate is due to their local labor market’s overall lack of jobs, versus neighborhoods whose low employment rate is due to neighborhood residents differentially lacking access to the local labor market’s jobs.
These Data Versus Census Data
Despite the importance of the prime-age employment rate, the Census Bureau does not directly report this variable. The Census does report the ratio of civilian prime-age employment to the sum of the civilian prime-age population plus the military prime-age population, but this reporting inappropriately considers two different groups in the numerator and denominator. However, using available Census data, one can calculate the ratio of the civilian prime-age employment to civilian prime-age population.
Compared to Census data, these data also have the advantage of providing all the prime-age employment rates for different geographic unites in one database. The user can readily compare prime-age employment rates for census tracts versus counties versus SLLMs versus SCZs versus states.
The Years Considered
This database is for the years 2015-2019, and are derived from tables prepared by the U.S. Census Bureau, based on data from the American Community Survey (ACS).
A five-year period is considered because the Census Bureau only reports data for census tracts, and for ALL counties, for five-year periods. The 2015-2019 period is considered for two reasons: (1) These data are close to a business cycle peak, and hence are more likely to reflect persistent economic distress or prosperity, rather than temporary conditions due to a business cycle; (2) Due to problems caused by the pandemic, Census Bureau data collection for the ACS had some problems for 2020, and may still have some problems for 2021.
On the first point, available evidence suggests that DIFFERENTIALS in prime-age employment rates across places are quite persistent over time (see Bartik (2020), or Austin, Glaeser and Summers (2018).) . Even though the prime-age employment rate for a particular place may differ in some other five-year period from that reported for 2015-2019, the differentials from national averages will be highly correlated for at least a 20 year period.
In addition to reporting the prime-age employment rate, the database also classifies a geographic unit by distress level. For all geographic units (state, SCZ, SLLM, county, tract), each unit is classified as “severely distressed”, “distressed”, “marginally distressed”, “close to fully employed”, and “fully employed”.
These classifications are inherently arbitrary, in that they are based on particular levels of the prime-age employment rate, and there obviously is little real differences between two geographic units that are just above or below a particular cutoff for the prime-age employment rate. The cutoffs used are:
- “Fully employed” is a prime age employment rate greater than or equal to 82.8%.
- “Close to fully employed” is greater than or equal to 80%, less than 82.8%
- “Marginally distressed is greater than or equal to 76.5%, less than 80%.
- “Distressed” is greater than or equal to 72.3%, less than 76.5%.
- “Severely distressed” is less than 72.3%.
These cutoffs are based on population percentiles in the SCZ data. In the SCZ data, the percentages of the U.S. population in each category, from “fully employed” to “severely distressed”, are: 10%, 30%, 40%, 10%, 10%.But we apply these same cutoffs to all geographic units.
Users can of course create their own cutoffs. But these categories seem to correspond to meaningful distinctions. It seems difficult for local labor markets to have employment rats much above 82.8% at business cycle peak, as few people live in such labor markets. Rates of 80% or above are close to 82.8%. Rates below 76.5% are more than 5 percentage points below the 82.8% level. And rates below 72.3% are more than 10 percentage points below the 82.8% level.
In addition, we classify census tracts into three “relative job distress” categories based on their prime-age employment rate compared to their SCZ: “relatively severely distressed’, “relatively job disadvantaged”, and “relatively nondistressed”.
- “Relatively severely distressed” is a tract whose prime-age employment rate is 9.8% or more below that of its SCZ.
- “Relatively job disadvantaged” is a tract whose prime-age employment is at least 3.3% below its SCZ, but not as great as 9.8% below its SCZ.
- “Relatively nondistressed” is all other tracts.
The cutoffs for relative distress are chosen so that 10% of the U.S. population is in “relatively severely distressed” tracts, and another 15% is in “relatively job disadvantaged” tracts.
As should be apparent, a tract’s RELATIVE distress classification may not correspond exactly to its absolute distress classification.
Users can use other cutoffs for tract relative distress classifications, or can classify relative distress based on comparisons with the prime-age employment rate for the SLLM, the county, or the state.