Every year, the US Department of Housing and Urban Development (HUD) estimates fair market rents (FMRs), which are used to determine payment standards and rental assistance levels for housing programs. Accurate FMRs play a critical role in whether voucher households can secure safe, decent, and affordable housing through the private market. If HUD inaccurately calculates FMR estimates and FMRs do not keep pace with rising rents, families may face limited housing options, have to devote a large portion of their income to housing costs, or fail to find a decent unit to lease with a voucher. Therefore, accurately estimating FMRs is vital to ensuring that renters with vouchers can maintain access to stable and decent housing. In places where rents rise more rapidly than FMR estimates, the negative implications of low or inaccurate FMRs may be even more severe.

Our research examined avenues for identifying areas with rapidly rising rents and providing viable approaches to improving HUD’s FMR calculations. We explored models that leverage local data to forecast whether a county is likely to see rapidly rising rents in the future. We also used local data to improve the predictive ability of time series models, building on prior work and HUD’s current methods. In our analysis, we found that incorporating timely data on vacancies, housing starts, population growth, unemployment, home values, and interest rates can improve models’ performance in forecasting rents at the metropolitan and regional levels. We then applied these models to county-level data and calculated six sets of alternative FMRs. When compared with historical data, the alternative FMR estimates tended to outperform HUD–produced FMRs from the same period.

Findings

One overarching question drove our research: what method should HUD use to calculate FMRs in markets with rapidly rising rents so that families using housing choice vouchers (HCVs) can find and lease suitable housing? Although the agency’s current estimation approach accounts for regional, metro-level, and national trend factors, our research examined whether incorporating timely county-level data can improve FMR estimates in markets with rapidly rising rents. We found the following:

• Between 2009 (2009–10) and 2019 (2018–19), an average of 19.5 percent of counties saw rapidly rising rents from the previous year. These counties are located across the country but are concentrated in the West—particularly in Montana, North Dakota, Alaska, and Colorado—with fewer counties in the industrial Midwest. We also captured rising rents in urban areas along the Pacific Coast.
• We explored models using local area unemployment rates, vacancy rates, the share of building permits per 1,000 housing units, and growth in the number of housing units to predict which counties will experience rapidly rising rents before rental data become available. Our models showed relationships between vacancy rates and unemployment and the likelihood of rapidly rising rents. However, in a validation sample, the models were unable to accurately predict which counties had rapidly rising rents.
• We found that an ARIMAX model that incorporates data on vacancies, building permits, housing unit growth, unemployment, home values, and mortgage interest rates tends to provide better predictions of future changes in gross rent levels across metropolitan areas or for regional aggregates of midsize cities than a pure time series ARIMA model.
• We saw promising results when we applied six alternative models to county-level data to generate FMRs and assessed their performance. We calculated six sets of alternative FMRs for US counties that use local data to create unique forecasts for each. We designed these alternatives to maximize accuracy of forecasts by county and did not include adjustments to increase FMRs above statewide minimums or 90 percent of the previous year’s FMR. Looking across counties in 2018, five of our six alternative models were more accurate than HUD’s FMRs according to root mean squared error. However, when FMRs were more than 10 percent off from actual rents, HUD more frequently set FMRs too high while our models more frequently forecasted FMRs that were too low. Relative to HUD’s FMRs, our six sets of alternative FMRs performed similarly in counties with rapidly rising rents.

The importance of accurate FMRs cannot be overstated. FMRs that fail to keep up with rising rents can limit housing choice, hinder the ability of new voucher holders to find eligible rental units, and increase housing instability. Our research shows that rents have risen rapidly in many US counties; however, it is difficult to forecast which counties will have rapidly rising rents in any given year. Additionally, the county-level FMRs that we developed performed relatively well in comparison with HUD’s FMRs, including the 2020 FMRs that calculated local and regional trend factors. Taken together, the best path to improve FMR calculations in areas with rapidly rising rents appears to be improving FMR calculations overall. HUD could take steps to improve FMR calculations by using county-level local data, as we did, although further refinements to our methods are needed.

Alternatively, HUD could incorporate local data and generate more local trend factors while continuing to stay within the basic framework of its current FMR calculation process. Our research shows that using more precise local data and focusing on smaller geographies could improve FMR calculations. Additionally, HUD’s shift toward using small area fair market rents (SAFMRs) makes such local data even more important. Many of the datasets we used in this study, for example, provide county-level but not ZIP code–level information. But as housing searches, job searches, and permitting continue to move online, new types of data collection could help develop more accurate FMR calculations in the near future.