Skip to main content
A housing construction project is seen in San Francisco, California.
Report

The early results of states’ Opportunity Zones are promising, but there’s still room for improvement

Eighteen states have submitted their selections of local neighborhoods that will qualify as “Opportunity Zones” under a new tax incentive created by Congress in last year’s tax bill.  These selections—and the characteristics of the neighborhoods themselves—will be important determinants of the ultimate success of the program. As I wrote in February just before the state selections were due, poor choices by states could turn a program meant to benefit residents of poor neighborhoods into a tax break for developers investing in already-gentrifying areas. With information in hand from 18 states, I describe the characteristics of neighborhoods that have been selected so far. (That information is also available in this file (.xls) for the remaining 32 states who have yet to select their Opportunity Zones).

Looking at the 18 states that have submitted, the good news is that most states designated deeply impoverished places for the new subsidy, with Georgia and California standing out for allocating most of their picks to their most distressed neighborhoods. Nevertheless, 22 percent of selections were for areas with relatively low poverty rates (below 20 percent) and an additional 19 percent were in already-gentrifying areas (areas with the highest rates of home price appreciation).

Opportunity Zones (hereafter referred to as OZs) offer favorable capital gains treatment for taxpayers who invest in designated low-income neighborhoods. States were given a list of eligible neighborhoods produced according to the law by the Department of Treasury—a broad list that actually includes 57 percent of all neighborhoods in America, not all of which are distressed—and were allowed to select one in four of them to be Opportunity Zones.

If states select areas that are relatively well off or already gentrifying, most of the benefits will accrue to investors and developers for investments that would have happened anyway.

The choices matter. They will help determine who benefits from the subsidy and the cost effectiveness of the program. If states select areas that are relatively well off or already gentrifying, most of the benefits will accrue to investors and developers for investments that would have happened anyway. If states select deeply impoverished areas that are starved for investment, the subsidy is more likely to encourage new activity—and, hopefully, benefit local residents.

We now have information on neighborhoods designated as opportunity zones from 18 states, giving a sense of the characteristics of the places selected. There is good news and bad, and substantial differences across states. As the remainder of the states select their zones, they can benefit from these early examples.

A look at the 18 states—who did well?

The good news is that, on average, states have selected deeply disadvantaged areas for their OZs. The average poverty rate of selected zones was 30 percent (in 2016), compared to an average poverty rate nationwide of 15 percent and an average rate within qualifying low-income communities of 25 percent. Across a range of indicators like child poverty and educational attainment, states’ selections are, on average, more disadvantaged than the low-income tracts they did not select, and selections had larger minority populations. (See Table 1.)

Table 1

To summarize states’ choices, we formed an index of economic distress based on poverty rates (adjusted for the number of university-student residents, who are often misleadingly counted as living in poverty), child poverty rates, educational attainment (the fraction with at least some college education), home prices, family income growth (from 2012 to 2016), and changes in the poverty rate (again 2012-2016) for each tract within each state.1 A state’s most distressed neighborhood has a rank of 1 and the least distressed neighborhood a rank of 0. For each state, we then compared how distressed are the areas states selected compared to low-income communities that were not selected.  Figure 1 illustrates the broad result.

Among the 18 states that have already submitted their selections, most selected relatively more distressed areas, with Georgia and California standing out for having picked some of their most disadvantaged places. In these states, the average neighborhood ranked at the 82nd and 81st percentile statewide for distress, respectively,—much more distressed than the eligible neighborhoods  they passed over, which ranked in the 63rd percentile and 69th percentile. Vermont and Colorado were close behind, followed by Wisconsin, Texas, Kentucky, New Jersey, South Carolina, and Oklahoma. In several of the other states—Ohio, Arizona, Alabama, Nebraska, Michigan—the state’s selections were, on average, the same as other LICs that were not selected.

Who could have done better?

But in South Dakota, Idaho, and Mississippi, neighborhoods designated as OZs were actually better off, on average, than eligible communities that were not selected. In Mississippi, in particular, the state selected tracts that had lower average poverty and child-poverty rates, were better educated, and had smaller minority populations than the low-income areas they skipped over.

For states that have not yet submitted their picks, we’ve ranked qualified areas by the same index economic distress in the attached XLS file and highlighted those in the most distressed 10 percent.

Why the range of outcomes?

Georgia and California used a disproportionate share of their designations to choose the most disadvantaged areas. Ranked by economic distress, 47 percent of Georgia’s picks went to areas with the top 10 percent of economic distress, as did 45 percent of California’s picks.  (See figure 2.)  In Mississippi, only 11 percent of picks were in the worse-off 10 percent.

Author

Indeed, the bad news is that, across all states, more than a quarter  of all picks (28 percent) were selections that were either not poor (had poverty rates below 20 percent), were college campuses (more than 1/3 of residents were enrolled in college or graduate school), or (very rarely) were areas where no one lived.  This fact highlights how important the legislation’s targeting requirements are. In many cases, states selected areas that were not especially distressed when they had the option. In Mississippi, for instance, while only 37 tracts in relatively good neighborhoods were eligible, 21 of them were selected.

One policy implication from these early examples is that the federal selection criteria really matter. While the program is intended to target only low-income communities—those with poverty rates above 20 percent or family incomes substantially below the state or city median—states actually have wide authority to select tracts that are less disadvantaged or are already improving rapidly. One reason is that the law also provides exceptions from the strict poverty or income thresholds for less-populous neighborhoods or for certain areas that adjoin qualifying low-income neighborhoods. Furthermore, the poverty measures from Census used to define low-income areas include college or graduate students, which means affluent areas like those surrounding Stanford, Harvard, or Georgetown qualify. And the Census data dates as far back as 2011, when neighborhood economic conditions may have been much worse than they are in 2018.

As a result, more than 57 percent of all neighborhoods in the states and DC are eligible for selection under the program, and as many as 82 percent of all tracts in Mississippi and 80 percent of West Virginia. And even within the set of qualifying tracts, poverty rates and economic distress vary widely. While most are deeply impoverished, 49 percent of eligible tracts had poverty rates in 2016 lower than the intended 20 percent threshold.

As the early selections show, states used these exceptions to pick places with relatively low poverty rates.

Gentrification

Finally, many states selected areas with higher-than-average home price appreciation—one indication of gentrification. Following Baum-Snow and Marion 2009 (whose work suggests that place-based subsidies do not increase housing investment in gentrifying areas), we  identify tracts whose house price appreciation is greater than 75 percent of tracts within their state over the period from 2012-2016. Selecting neighborhoods with rising home prices increases the value of the tax benefits to local developers, but reduces the effectiveness of the program. As above, we compared the tracts selected by states to low-income communities they passed over, to assess whether they were more likely to pick places where home prices were rising faster.

In most states, selected tracts had higher home price appreciation than LICs that were not selected.  Mississippi again stands out—about 40 percent of selected tracts were in high home-price-appreciation neighborhoods versus 22 percent of the neighborhoods they passed over.

The selection of appreciating areas will reduce the cost-effectiveness or bang for the buck of the program, because the OZ subsidies will tend to accrue to projects that investors would have pursued anyway. That’s certainly the experience we’ve had with other place-based policies, like the low-income housing tax credit (LIHTC). And it’s no surprise that many states selected gentrifying areas when they had the opportunity. Our experience with LIHTC also suggests that when local governments are tasked with allocating tax benefits, developers often get their projects approved in already-gentrifying areas (Baum-Snow and Marion 2009).

Summary

Of course, by focusing on data available from the Census Bureau, we leave a lot of local or more recent knowledge untapped. There are other reasons to select zones, like state and local development priorities and the desirability of areas for investment and development. So it’s not surprising—and not necessarily a sign of failure—that states didn’t confine their picks to their worst-off places.

Finally, it is important to recognize that while state selections matter for a lot of things, their choices are only one factor influencing whether poor residents will benefit from the program. Ultimately, that will depend on whether the program succeeds at increasing development in distressed areas, whether those investments have positive spillovers within the local community (like increases in jobs or neighborhood quality), and whether current residents are retained in their neighborhoods or are pushed out by new residents.

Reference:

Baum-Snow, Nathaniel and Justin Marion. (2009.) “The effects of low income housing tax credit developments on neighborhoods.”
Journal of Public Economics, 93(5-6): 654-66.

Footnotes

  1. Specifically, within each state we summarized each measure using a normalized z-score (top coded at an absolute value of 4) and formed the index based on the mean z-score across all 5 factors. To form the adjusted poverty rate, we assume half of enrolled college or graduate students are recorded as living in poverty. In general, the broad results are not sensitive to the modest changes in the construction of the index.
Get daily updates from Brookings