ZIP-code level analysis reveals where being middle class feels rich or poor – and where in the U.S. it feels more like a developing country
The great recession has brought substantial changes in wealth and income distribution to U.S. households. This study analyzed the geographic distribution of income across U.S. zip codes. While we expected to see differences in where the rich and poor live, we were surprised by how stark this geographic divide is: the rich and the poor live geographically separated lives. In particular:
(1) The highest 10 percent of income earners live in cities, while households below the poverty line live in the countryside and in some inner city agglomerations.
(2) The concentration of high income earners in cities makes many city neighborhoods quite homogeneous: again, the rich only live with the rich. Income inequality within cities tends to be lower than in the countryside.
(3) The highest 30% income owners concentrate in cities so much that city household dwellers earning the U.S. median income fall into the lowest 30% of income earners locally in these cities. In other words, while a household earning the median income in the U.S. is by definition of the median poorer (and at the same time richer) than 50% of the households in the U.S., in its local zip-code more than 70% of households earn a higher income. Thus, city dwellers earning the U.S. median income have every reason to feel poor.
(4) Many middle income earners seem to avoid cities. Instead, they locate themselves evenly distributed in the countryside. Upper middle income households can be found in suburbs, while lower middle income households can be found in less affluent city areas, well separated from affluent neighborhoods.
The data suggest analyzing the U.S. as two economic – and social – entities: affluent large cities versus the less well-off areas outside these cities. Consequently, policy analysis and prescriptions should at least separately address the existence of these two entities within the U.S.
Our analysis differs in three ways from previous efforts and reports issued by the Census in February 2012 (http://www.census.gov/prod/2012pubs/acsbr10-18.pdf). First, we use year 2013 estimates (forecasts) instead of historical data. Second, we proceed with our analysis on the ZIP-code level, which helps us unmask otherwise hard-to-see separation. Third, we order and sort U.S. households into ten equally sized groups (deciles) and provide analysis separately for each group.
Income and wealth have made major news headlines since it became apparent that top income earners have benefitted the most from the economic recovery. Evidence is mounting about possible consequences, the most recent addition being the so-called Gatsby Curve (http://www.whitehouse.gov/blog/2013/06/11/what-great-gatsby-curve). This curve implies that it is hard for poor U.S. children to climb out of poverty as adults, and the more unequal this income distribution gets, the harder the struggle becomes. Combining this effect with reduced mobility across geography after the great recession may cement regional disparity if large income differences exist across regions. We therefore analyze and portray current income differentials in the United States.
Our analysis proceeds in four steps. First, we analyze where there are substantial concentrations of rich and poor. Not surprisingly, the rich concentrate in major cities, particularly on the East Coast, and the poor tend to live in the countryside – and some inner city clusters.
To better examine the concentration of high income earners in cities, we determine in which zip codes the US median income household provides high or low income in relation to all other income earners in the same zip-code. In other words, where does the U.S. median income make a household “feel” rich or poor relative to their neighbors? If the U.S. median income places a household into a low income category locally, this implies that more affluent households dominate the zip code.
We find that the rich concentrate so much in cities that households making the US median income in these cities actually may end up feeling poor. These median income households would have to move to the countryside to feel relatively rich.
Next, we asked how this crowding out of lower income earners in cities influences local income distribution. We find that it makes many city neighborhoods quite homogeneous: income inequality in city zip codes tends to be lower than in the countryside, as affluent households generally live next to similarly affluent ones.
Finally, we wanted to learn about the geographic distribution of middle income earners. Our findings suggest that they tend to avoid cities, but are otherwise quite evenly distributed in the countryside. Upper middle income earners concentrate more in the suburbs, while lower middle income earners concentrate around less affluent city centers to a considerable extent.
The combined analysis paints a stark picture of American income distribution across the US geographic space. The upper 30% of income earners live in city neighborhoods with low income differences that are separated from the rest of the population. Town and rural areas and certain city enclaves are dominated by the lower income segment. The middle income segments also avoid big cities, but they are more evenly distributed across the countryside. It is this large city versus town and countryside divide that coincides with the top 30% versus bottom 70% household income line that makes the U.S. appear like two separate socio-economic entities.
Our analysis differs in three important ways from earlier attempts to study US income distribution. First, instead of using historical data only, we approximate current income levels by interpolating the most recent available income estimates from ESRI for the year 2011 with their five year forecasts. Second, we proceed with our analysis at the zip-code level, which helps us unmask otherwise hard to see separation. Even zip codes may harbor quite a large number of neighborhoods with different income characteristics, but much less so than counties, where extreme cases may counter each other in the process of aggregation. Third, we order US households by income and sort them into ten equally large groups (deciles) to provide a brief analysis for each group. For ease of comparison, we present maps of the geographic concentration of each group in a short animation that is available on our website.
Our analysis has some important limitations. First, it is based on point estimates (not historical measurements) which introduces some error into any analysis. Second, our ideal unit of analysis would be neighborhoods which combine households of similar characteristics. Zip codes are still very large approximations of neighborhoods and have been designed to optimize delivery routes, not for economic analysis. They are advantageous because everyone knows in which zip code they live in. This is not necessarily the case for smaller units of analysis such as census blocks or block groups, which generally also have higher margins of error. Finally, we only describe the data up to this point in time. We do not establish causal relationships and consequently do not suggest drawing specific policy conclusions from it yet, as future research will focus on causal connections between regional income disparity, education, and regional economic development.
Nonetheless, our findings establish empirical regularities about U.S. income distribution across geographic space. They suggest revisiting and furthering academic research in optimal policy design, as separate U.S. regions may be better addressed as different socio-economic entities. As such, policy analysis may yield better results if effects of a particular policy are analyzed for those two (or even more) entities separately.
Next we describe each step of our analysis in more detail and illustrate it with national level maps. Separate maps for our home-state California and specifically Southern California are available on ISEA’s webpage (http://isea.redlands.edu).
Where the rich and the poor live
In October 2011, the Congressional Budget Office (CBO) released a report (http://www.cbo.gov/ftpdocs/124xx/doc12485/10-25-HouseholdIncome.pdf) that showed that while the average American household income rose by 62% between 1979 and 2007, income growth for the top 1% of households was more than 15 times the income growth for the bottom 20%. This development increases the importance of the more recent finding of the so-called Gatsby curve (http://www.whitehouse.gov/blog/2013/06/11/what-great-gatsby-curve), which provides a bleak perspective for intergenerational income mobility. Significant income concentration in certain regions and low cross-regional mobility can have important implications for regional development. Here we analyze the issue of income concentration only and leave the issue of cross-regional mobility for future analysis.
As shown in Map 1, the rich and poor concentrate geographically in different locations. Rarely will one find significantly higher than average concentrations of rich (in this study defined as the top 10% income earners by household) and poor (defined as households living below the federal poverty line) in the same zip code. About 0.2% of all zip codes fall within this category (as represented by the color red on the map.)Those zip codes tend to congregate around regions where the top 10% are more concentrated, specifically inner-city developments that are proximate to richer suburban areas.
Probably the most striking feature of the map is the existence of “buffer zones” – zones that tended to separate the rich from the poor. As we will see below, these light-grey zones frequently cover suburbia where middle-income households dominate. Most of these buffer zones were geographically located in the Midwest of the country and comprise about 48% of all zip codes. With the shrinking of the American middle-class, the number of zip codes in these zones may decrease even more.
The rich live in the cities
Cities are engines of growth, and not surprisingly, many high income earners live in cities. To capture this feature in maps, we present highly populated places (more than 500 people per square mile) in solid colors and those in less populated places in translucent colors. As expected, higher income earners are concentrated in cities. Zip codes where the top 10% of income earners significantly exceed 10% of the population represent 15% of all zip codes and are concentrated in big urban metro areas such as the Southern California/Greater Los Angeles region, the Northern California/Bay area region, the DC/New York City/ Boston corridor, the Greater Chicago area, the Greater Minnesota area, the Greater Denver area, the Greater Houston and Dallas areas, and the Greater Atlanta area. There are almost twice as many zip codes where the top 10% income earners comprise more than 16% of the population (the dark yellow zones) as there are zip codes where the top 10% income earners are between 12% to 16% of the population (the yellow-green zones).
The poor live in the countryside
We had no direct data on poverty levels by zip code therefore we approximated it using data on average family size and prevalence of income levels. Our estimation yielded that about 16% of all households in the US live below the poverty line, which is consistent with US Census bureau figures. According to this approximation, surprisingly, poverty is geographically dominant in the United States, with 36% of zip codes having substantially above average levels of poverty. Regions where the percentage of households with incomes below the poverty line is between 19.0% and 25% (the purple zones) comprise 20% of all zip codes. Regions where the percentage of households with incomes below the poverty line exceeding 25% (the purple blue zones) represent 16% of all zip codes and include almost all of the southern states, New Mexico, the Dakotas as well as urban inner-city areas. Within these regions, on average one in every three households live below the poverty line.
Where earning the middle (median) income feels rich – or poor
Next, we sort American households by income and divide them into ten equally sized groups. We label the three groups with the highest income the upper income category, the next lower four groups the middle income category, and the last three groups the low income category. While these groups are not identical with the traditionally-conceived upper, middle and lower classes, they provide easy to follow reference points. We repeat the exercise separately for each zip code and split each zip code into three household income segments: the 30% highest, the 40% in the middle, and the 30% at the low end. We then compare the U.S. national median household income with the household incomes in each zip code to determine whether the national median household income falls into the highest, middle or low income category in a particular zip code.
We find that in all major cities, the U.S. median household income falls into the low income category locally (see red areas on map 2), which comprises about 7% of all zip-codes. This indicates that the U.S. upper income segment predominantly concentrate in cities and represent their single biggest resident base. City dwellers earning the U.S. median income (specifically on the East coast) will likely feel poor relative to their neighbors as they belong to the local bottom 30% of the income distribution. If they could move to the green areas in the map, which comprise about 17% of all zip codes, while preserving their income, they may feel rich as they would fall into the top 30% of income earners in these zip codes.
Switzerland or Swaziland: income inequality within US ZIP codes
If the most frequent habitants of urban areas are affluent, with the relatively poorer living in rural areas), we wanted to know how income is distributed within zip codes. We employed the frequently used Gini coefficient for which comparison measures on the county, state and country level are available. Due to the much smaller size of zip codes relative to countries, one would naturally expect income distributions to be more equal within zip codes. It turns out that this is the case only for about 42% of U.S. zip codes, and especially prevalent in urban areas. To our surprise, however, we found that the majority of zip codes in the U.S. are either small mirror images of the national income distributions or have even more unequal distributions. One fifth of zip codes have higher inequality relative to the national average with about 6% of zip codes substantially higher inequality (mapped in orange and red). While rare, inequality in a zip code can reach similar levels as Brazil’s inequality on the country level. Rich areas are not only richer, but they also have lower income inequality. Poor, rural zip codes tend to have higher inequality. The rest of the zip codes roughly mirror the U.S. as a whole, which is surprising in its magnitude. We suspected that this result was an artifact of less densely populated zip codes in rural areas harboring more diverse groups. However, we found the local median household income to be much more highly correlated (-0.53) with the local Gini-coefficient than population density (0.06). Thus, the level of income is more highly associated with inequality than the size of the zip code.
In our last step, we analyzed how the middle income segment behaved in terms of location choices. We therefore compared the four decile groups of the middle income segment to the ones of the upper and lower segments. As can be seen in map 4, we found the middle income segment to be spread out across the country – excluding cities.
A full geographic representation of which income groups concentrate where can be found in an animation in the online version of this report on the ISEA website. For complete characterization, we describe major features of all 10 income groups from highest to lowest next.
Upper three income segments
The highest income group concentrates in city centers, specifically on the East coast. They concentrate more than any other income group and can only be sparsely found in rural areas. The 2nd highest income group lives in cities and suburbia, but not at the same level of concentration as the highest income group live in the cities. They can also be sparingly found in rural areas. The third highest income segment concentrates even less than the 2nd highest, but their domain is the same region as the second highest income group.
The four middle income segments
The fourth highest income group mildly concentrates in extended, less populated suburban areas. It avoids the city centers. Each city has, however, its own geographic pattern: In Atlanta, they tend to live closer to the city center, in Chicago and the East coast they live farther out in the suburbs. In the fifth highest income group, we start to see strong avoidance of city centers; otherwise, we can barely find any concentration. The sixth highest income group is the least concentrated, but strong avoidance of the city centers is apparent in this group. The seventh highest income group is also characterized by avoidance of city centers, but some concentration in poor neighborhoods within cities (e.g. downtown Los Angeles, Atlanta, Dallas) starts to emerge.
The three low income segments
The third lowest income segment is more concentrated in rural areas and poor city center areas as all higher income groups, and it can rarely be found in other city areas. For the second lowest income group, we see the trends from the third lowest group continue. We also find some additional concentration in Texas near the border with Mexico. The lowest income segment displays a distinct pattern: we find concentration in southern rural areas, poor city centers (West Detroit, South Chicago, downtown LA), and low presence in zip codes where the upper 30% reside.
The animation of the maps is available on YouTube (http://youtu.be/rYzVkFc2X2w).
 We don’t use the term “significant” in a purely statistical sense, but rather in the more intuitive sense that we require a significantly higher share to be 20% higher than the expected share. For example, if the expected share was 10%, it is significantly higher at 12%. Given the large number of zip-codes, all our significantly higher or lower results are also statistically significant results.