As investor purchases of single family homes have increased over the last several years, so have concerns about the potential negative effects on communities. Black communities in particular have become hotspots for investor activity. Using data from the Home Mortgage Disclosure Act (HMDA) that contains a record of over 90% of all mortgages, we can better understand this relationship.
We will use the share of home purchase loans for non-owner-occupied properties as a proxy for investor lending activity. Starting in 2018, HMDA began separating Non-Owner Occupied (NOO) loans into second home and investment properties. However, the terms NOO loans and investor loans will be used interchangeably unless noted otherwise.
Black-majority tracts have represented a consistently higher NOO share than other tracts for most of the last 2 decades. However, this gap flattened during the Great Financial Crash of 2008 from 9.2 percentage points (pp) in 2005 to .3pp in 2009-2010 and then steadily increased to 7.6pp by 2025. The near convergence between the rates towards the end of the recession suggests that the downturn had a particular effect on investor lending in these neighborhoods.
The chart below shows the share of investor purchases in Black census tracts compared to all other census tracts:
A regression analysis can allow us to better understand the general relationship between the share of Black residents in a census tract and the rates of investor lending.
We used a fixed effects model that compares the share of investor lending in census tracts within the same metro and in the same year. The main variable of interest is the Black share of the census tracts. The control variables (home values, household income, price-to-rent, renter share and the vacancy rate) were added in stages.
The chart below shows a 1% increase in the Black share corresponding to a .084pp increase in investor share. This means that a 100% Black neighborhood (census tract) has an investor share that is 8.4pp higher than a neighborhood/CT with no Black people.
However, when the variables are introduced, the relationship between the Black share and the NOO share is statistically indifferent from zero (-.2). In other words, after we account for certain neighborhood characteristics, investors are no more likely to buy property in a Black neighborhood than any other neighborhood. Because the controls themselves have such explanatory power, it is worth exploring how they correlate with investor lending and what that might mean for Black neighborhoods.
The size of the regression coefficient and its sign show how each corresponds to investor lending. Higher home values, price-to-rent ratios, renter share amounts and vacancy share rates are associated with more investor lending as shown in the chart below:
Compared to other tracts, Black-majority tracts have conditions that are favorable to investors. On average, Black-majority tracts are less expensive, renter-heavy, have lower incomes and have more vacant properties.
These characteristics make Black neighborhoods attractive to investors but this dynamic is not a matter of pure coincidence. They are a result of decades of disinvestment, including redlining, discriminatory banking practices and a substantial racial wealth gap. This has created favorable conditions for outside investors to buy up property in economically depressed Black areas. This dynamic is even more evident when you examine the relationship between investor lending and Black neighborhoods in different metros.
Similar to the national trends, without controls in nearly all major metros, an increasing Black share in a census tract is associated with higher levels of investor activity. After we control for neighborhood characteristics, only Memphis, TN and Cleveland still have a statistically positive relationship. Washington, DC and New York have negative relationships once controls are introduced. The contrast suggests that local conditions affect the way investor lending operates in these regions. Nevertheless, Memphis, TN and Cleveland are metros where deindustrialization has hampered economic growth.
The chart below shows the relationship between investor lending and Black neighborhoods across 12 major metropolitan areas:
So far, we have only examined the relationship between investor lending and Black neighborhoods nationally and across metros. However, investment dynamics might differ between upper-middle and low- and moderate-income (LMI) neighborhoods. To understand if investor lending differs between LMI and non-LMI neighborhoods, we ran the same models as before at both the national and metro level.
After controlling for neighborhood characteristics, there was no statistically significant relationship between investor lending and Black neighborhoods at the national level. However, when you separate LMI and non-LMI Black neighborhoods, we see that the null estimate really masked the Black-LMI relationship.
The chart below shows how investor lending in LMI and non-LMI neighborhoods changes as the Black share increases and how it corresponds to changes in the share of Black residents:
Looking at the entire 15-year period, an all-Black LMI census tract has 2.6pp more investor lending than an LMI CT with no Black residents. Conversely, as the share of Black residents increases in a non-LMI neighborhood, investor lending decreases by -1.6 pp.
Recall we have used non-owner occupancy as a proxy for investors because only the post-2018 HMDA series splits investment and second home/vacation properties. When we look at investment properties purchased during the 2018 to 2025 time period, the LMI-investor lending effect is several times larger than non-LMI-investor lending. It suggests that the LMI-investor lending effect for investment-only (excluding second home) loans for the entire 15-year period could be substantially larger than 2.6 pp.
The chart below shows the same 12 metros with the percentage point estimates and their significance for non-LMI and LMI tracts:
The relationship between the share of Black residents and NOO lending is highly metro-specific. In Memphis, TN, Cleveland and St. Louis, it’s positive and concentrated in LMI tracts. In contrast, in New York and Washington, DC, the relationship is negative. In Memphis, an all-Black LMI neighborhood has a 30 pp higher investor share than a LMI neighborhood with no Black residents; in Cleveland, it is about 15 pp.
A higher share of Black residents in a census tract is often associated with a lower NOO lending rate. Across the 12 major metros examined, most show no robust relationship after controls, but the differences between metros are statistically significant. One implication is that local dynamics we can’t observe might explain why investor lending interacts with neighborhood demographics across metropolitan areas.
The metro analysis was limited to the 12 metros with both a sufficient number of counties and a Black population high enough for a robust model. The inclusion of other metros like Houston and Charlotte, NC wouldn’t change the fact that investor lending’s impacts on LMI and Black communities are highly likely to depend on local conditions.
More broadly, lending for non-owner-occupied properties and investors’ motivations is but only one facet of understanding investor activity in the US single-family housing market. The other half of the investor market is all-cash buyers. Their share of the total sales is estimated to have been as high as 39% in 2025. Therefore, cash-buyers could account for two or more times greater loan volume than the financed portion recorded in HDMA.
For the analysis in this article, the effects of cash-buyers on investor lending and mortgage lending more generally remains unknown and could be considered an omitted variable. What we can say with some certainty is that among financed, non-owner-occupied home purchases there is a statistically significant relationship between them and Black-LMI neighborhoods. This effect is dependent on the local and metro-specific factors, some of which can be measured, such as the overall renter share and the vacancy rate.
Methodology: This analysis uses HMDA home purchase originations from the 2004-2025 time period that were restricted to first-lien, 1-4 family, owner-occupancy-coded loans. Every year was placed onto constant 2010 census tract boundaries. Race/ethnicity shares came from LTDB decennial files that were interpolated between census years. Data about home values, incomes, rent amounts, tenure and vacancies came from ACS 5-year tract estimates. A fractional-response logistic regression with NOO’s share as a bounded [0,1] outcome weighted by each tract-year’s loan count. The model employed metro—year fixed effects and standard errors clustered on the county. The controls were home price, income, rent-to-value yield, renter share and vacancy rate. Metro-level significance was adjusted for testing 12 metros at once.
Joseph Dean is the Junior Racial & Economic Research Specialist with NCRC’s Research team.
Photo credit: K from Pexels.
