According to the Economic Policy Institute, the top 1% of earners take home 21% of the total income in the United States, a share that has been steadily growing since the 1970s. A February report by the Federal Reserve Bank of St. Louis stated that the bottom 50% of households held only 2.6% of the nation’s total household wealth.

Understanding income inequality is important, but collecting reliable data over time can be a challenge. Colgate University Assistant Professor of Economics Young Park and Dong Gyun [Don] Shin, the A. Lindsay O’Connor Chair of American Institutions in the Department of Economics, have overcome that challenge and put together a large-scale study examining income differences in the United States at the county level.

Park and Shin analyzed data across five decades for about 3,000 counties to identify trends. Their study was published in September in the Journal of Applied Econometrics, and, according to Park, its sample size and timespan set it apart from other efforts.

In their analysis, Park and Shin found that countywide incomes have diverged noticeably since the 1990s — absent governmental subsidies. The portion of middle-income counties has decreased as they either reach upper-income levels or, more frequently, are pushed into lower-income brackets, a phenomenon economists call bipolarization.

According to Park and Shin’s data, counties with populations that pursued higher levels of education were more affluent, as were counties that hosted high-tech industries. “Usually, the education level is related to the industry composition of the county,” Park explains.

Park and Shin also found that richer counties tended to cluster together, as did less affluent counties.

One implication of income bipolarization is particularly worrisome for Park. Peaks in income distribution — with high income and low income on each side of the graph and a trough in the center — show that there are structural impediments to income mobility. Thus, it has become tougher for lower-income counties to get out of the lower-income tiers.

“It’s not easy to leap directly from the low-income group to a high-income group, so lower-income counties just stay there,” says Park. “Lower incomes mean more income uncertainty, lesser chance of mobility in the income distribution, lower quality of life, and other problems. In short, a county’s initial income position determines its fate. That’s why we have concerns about the income bipolarization we identified.”

Park says that government transfers are very effective at slowing the widening wealth gap and that one possible way to address county income bipolarity is for the government to subsidize educational institutions in poorer counties. “That could promote educational investment and help remedy the income differences between counties,” she says. “It is an easy and direct way to resolve the income inequality problem among counties.”

But by far the strongest tools the government has are healthcare subsidies. “We found that Medicaid and Medicare benefits were the most significant governmental transfer [subsidy] programs and have the biggest redistributive effect and depolarize incomes the most,” Park reports.

One of the reasons, Park and Shin report in their paper, is simply that governmental medical benefits have risen relative to other government transfers — like retirement benefits — since the 1970s. “This partly explains not only why medical benefits play the largest redistributive role, but also why the increase in their role is also the largest when looking across the full sample period,” they wrote.

Governmental subsidies directed at industry can help, but they are usually less efficient, as different counties tend to maintain different industries — depending on their environment and other factors — to achieve comparative advantages. “It can be really difficult and costly for the government to change the industry composition of a county,” Park says. “For example, in Montana, they specialize in agricultural products, but you cannot push Montana to create high-tech companies or something like that because it’s very ineffective in that economic environment.”

Down the road, Park would like more data on individual and household incomes over time to draw a sharper picture of the effects and characteristics of income inequalities.

“The conclusions you can make by looking at county data are not as granular as those you could draw from individual data, which is why individual data are so important,” she says. “We do need to understand the regional level income distribution, but we can’t directly explain the nature of individual distribution by looking at the regional effect. We need both for a complete picture of how regional income segregation plays an important role in explaining income inequality at the household level.”