Why Is Detroit so Empty?

This executive summary and research paper were written by 2016-17 Values & Capitalism Young Scholar Award recipient, Eric LaRose, a recent graduate of Hillsdale College, Class of 2017.

The decline of industrial cities in the Midwestern and Northeastern United States has been accompanied by significant population loss of inner cities as residents migrate to the suburbs. Cities such as Cleveland, St. Louis, and Detroit have less than half of the population they did fifty to sixty years ago and continue to lose residents; others, such as Chicago, Philadelphia, and Minneapolis, have rebounded in recent years but still have populations twenty to thirty percent below their peak levels. Since the housing stock of a city rarely declines at nearly the same rate as population does, heavy population loss tends to give rise to significant residential abandonment. Abandoned and blighted structures pose substantial problems for communities; they can serve as havens for crime and lower nearby property values, among other things. While population loss—and thus some levels of abandonment—may be inevitable, this study shows how bad public policy can exacerbate the already critical condition of many Rust Belt cities.

Detroit has been especially hard-hit by abandonment. Large swaths of the city have become “urban prairie,” in which most lots are empty, and nature has started to reclaim neighborhoods. Detroit’s city government has identified and recommended for demolition over 46,000 vacant structures or lots. Overall, slightly more than 22 percent of residential structures in the city are likely to be abandoned. It seems that population loss should easily explain Detroit’s abandonment problem: the city has lost 63 percent of its population since its peak in the 1950s. Yet, St. Louis which has also lost about 63 percent of its peak population, has an equivalent abandonment rate of just 9 percent. Why is Detroit so different?

Overall, a simple regression of abandonment rates against population loss percentages for 28 Rust Belt cities reveals that only about 43 percent of the variation in abandonment rates between these cities can be explained by population loss rates. Clearly, other factors are at play in fomenting the abandonment of residential structures. Hence, this paper uses census tract level data on abandonment in Rust Belt cities to examine the extent to which certain political, economic, and demographic factors, as well as population loss, contribute to abandonment.

Until recently, residential abandonment has been very difficult to measure. While individual city governments have on occasion attempted to count abandoned residential structures in their cities, differing classifications, and definitions of abandonment have made cross-city comparisons impossible. Luckily, over the past several years the United States Postal Service (USPS) has partnered with the Department of Housing and Urban Development (HUD) to create a nationwide count for the occupancy status of residential properties, conducted at the census-tract level. USPS mail carriers keep track of the occupancy status of homes on their routes; they flag homes which have not collected their mail for 90 days or longer as “vacant” and homes which are inactive as “no-stat.” The data collected is updated every quarter, which makes it possible for inter-city comparisons of abandonment rates at the census tract level.

Research Methods

To investigate the extent to which factors other than population loss contribute to abandonment in Rust Belt cities, I used USPS/HUD abandonment rate data from the second quarter of 2016 for census tracts in 28 Rust Belt cities. These 28 cities comprised of a set industrial cities in the Northeast and Midwest (including major cities such as Detroit, Chicago, and Cleveland, as well as smaller urban centers such as Gary, Indiana, and Youngstown, Ohio) that had a peak population of at least 100,000 and have lost at least 20 percent of their peak populations. Using the Longitudinal Tract Database, I gathered data on economic and demographic characteristics for each of these census tracts for every census from 1970 to 2010 (most of the country was not mapped into census tracts prior to 1970). Additionally, I used the Lincoln Institute of Land Policy’s Fiscally Standardized Cities Database to obtain annual data on citywide policies from 1980 through 2010. A city’s average property tax burden over this period was calculated as the average of each year’s property tax revenue per household divided by the city’s median home value. Likewise, the average other tax burden was calculated as the average of each year’s other tax revenue per capita divided by the city’s per capita income. Other variables reflecting citywide policies were obtained, including average property and violent crime rates as well as debt levels.


A series of regressions were then run at the census tract level using two separate measures of abandonment based on the USPS/HUD data. A simple linear regression of population loss at the census again shows relatively low explanatory power, with R2 values of .32 for one measure of abandonment and .26 for another. When city dummy variables were included, these values nearly doubled, suggesting that variations in city policies, which are captured in the dummy variables, largely explain variation in abandonment rates.

Expanding the model to include changes in demographic and economic characteristics reveals that neighborhoods with older housing stocks have higher abandonment rates, all else being equal. A one percentage point increase in the percentage of a tract’s homes older than thirty years in 1970 increases its predicted abandonment rate by roughly .08 percentage points. A tract’s median household income in 1970, its percentage change in median household income from 1970 through 2010, and the percentage of the population that was white in 1970 were all also statistically significant at the 1 percent significance level.

Further expanding the model to include variations in city policies reveals that a city’s average property tax burden is statistically significant at the one percent level, with very large coefficients. A one percentage point increase in average property tax burdens over the period from 1970 to 2010 increases abandonment rates by just over 13 percentage points using one measure or by just over 10 percentage points using the other. Variables for most other citywide policies either have very small coefficients or are statistically insignificant. These results, which are consistent with the findings of previous literature, suggest that Detroit’s high abandonment rate results in part from its much higher than average property tax rates as well as its relatively old housing stock.


Theme developed by TouchSize - Premium WordPress Themes and Websites