28 Nov. 2012 | Comments (1)
When companies decide on locations for new operations, they should look at measures of pay for the specific types of workers that they need to hire, and compare those measures across locations. However, information about pay is more easily available for entire localities than for specific occupations within localities. In this blog, we argue that if companies look only at average pay for a location, rather than taking into account the educational, occupational, and industry mix of those locations, the results are likely to be misleading for the types of workers they need to hire.
In the 2010 American Community Survey, there are 282 metro areas. In the left column of the table below, we rank selected metro areas based on average usual earnings. In the second column, we rank the metro areas according to average usual earnings after using regression analysis controls for age, gender, hours worked, education, occupation and industry. Regression analysis enables us to rank metro areas taking into account various differences across locations that might be associated with average usual earnings
In general, pay is higher in larger metro areas, but not always. We find that Stamford, Danbury, San Jose, Washington DC, and San Francisco are always among the five locations with the highest pay, even when we control for the characteristics above.
In some localities that are lower ranked without regression analysis, we find that applying regression analysis changes the results of the ranking. For example, Boston is the location with the seventh highest average pay. But this is partly a result of Boston having a relatively educated population and a large number of high paying occupations and industries. Once we control for these attributes, Boston drops to 14 in the rankings. Similarly, Minneapolis moves down the rankings from 13 to 22, Chicago from 22 to 36, Denver from 27 to 57, Ann Arbor from 29 to 116, Raleigh-Durham from 43 to 123, and Tallahassee from 130 to 261. If we look at the overall average pay without controlling for characteristics of localities, we would miss the fact that for a given occupation, industry, and education level, Tallahassee is actually one of the metro areas with the lowest salaries in the country.
There are also some locations where the population is relatively less educated and the share of low-paying occupations and industries is high. In those locations, once we control for characteristics associated with average usual earnings, we see the ranking go up. For example, Santa Cruz moves up from 25 to 11, Los Angeles from 57 to 32, Riverside-San Bernardino from 114 to 40, Las Vegas from 123 to 13, and Merced, CA from 269 to 58.
All of the examples above demonstrate that by solely looking at the overall average pay, one could take away misleading insights regarding the relative pay across metro areas.
In addition to age, gender, hours worked, education, occupation and industry, race and ethnicity is also associated with average usual earnings. Racial and ethnic minorities typically earn less than white workers with similar characteristics. For example, according to the American Community Survey, compared with white, non-Hispanic workers with similar characteristics, black workers earn 15 percent less, American Indian workers 12 percent less, workers of Chinese origin 16 percent less, and workers of Mexican origin 8 percent less.
As a result, metro areas with large minority populations tend to have lower average pay. Once we take into account the racial and ethnic composition of the workforce, the rankings change quite a bit (Column 3). For example, Atlanta moves up in the rankings from 89 to 65, Memphis from 137 to 93, New Orleans from 130 to 96, and Miami from 186 to 97.
Table 1 - Ranking of usual weekly earnings for selected metro areas among all 282 metro areas
Source – American Community Survey and authors’ calculations