Special Studies, October 13, 2007
By Natalia Siniavskaia, Ph.D.
Report available to the public as a courtesy of HousingEconomics.com
As the housing market downturn persists, uncertainty about when it will turn around grows. A measure that could be used to judge trends in both the national and local real estate markets is the share of buildings that are vacant at a particular moment. National data on vacancies still shows major supply overhangs in both the for-sale and for-rent housing markets, suggesting that markets remain in a significant cyclical downturn. However, vacancy conditions vary considerably from area to area. There are several sources of data on local vacancy rates that can provide clues to assess stages and turning points in local markets.
To be most useful, vacancy rate data should be timely, localized, accurate, and target a specific type of building. It would also be helpful to have vacancy data over an extended period of time, to judge if a current vacancy rate is above or below its normal level. Home builders would like to know if residential vacancy rates are above or below their normal levels in a particular locality in order to judge whether it is time to ramp up production in that area. Sources of local vacancy rate data include the Housing Vacancy Survey (HVS), the American Community Survey (ACS), and the US Postal Service/HUD (USPS) data base. Each data source has its own strengths and weaknesses, but none perfectly meets all the criteria for an ideal source of vacancy data (see Table 1). Some of these data sources have been widely used for decades, and some of them have just recently become available. All are available to the general public on the internet.
Housing Vacancy Survey
The traditional source of vacancy data is the Census Bureau’s HVS. The HVS provides timely homeowner and rental vacancy data on a quarterly basis for the U.S. as a whole, as well as for the four principal census regions, and has been doing so since 1956. On an annual basis, it provides vacancy rates for all states, and for the 75 largest metropolitan areas. The greatest advantage of the HVS is its extended historic coverage. Having access to such long time-series allows analysts to evaluate not only short-term trends but also to figure out what vacancy rates can be considered normal for various geographic areas.
The US Census Bureau provides access to the HVS data free of charge. Conveniently, the HVS vacancy rates are already tabulated for all available geographies.
Table 2 presents the HVS vacancy rates for the 75 largest metropolitan areas in 2005 and 2006. The data show that during that time homeowner vacancy rates increased sharply in most metro areas, with rates more than doubling, tripling and quadrupling in some areas. Orlando, FL and Jacksonville, FL claim the highest homeowner vacancy rates in 2006, 5.2 and 4.9 percent respectively. Bridgeport-Stamford-Norwalk, CT have the third largest rate, 4.3 percent, that is s a result of one of the sharpest annual increases, with rates more than tripling from 2005 to 2006. At the other end of the spectrum, is the El Paso, TX metro area where homeowner vacancy rates not only remained one of the lowest in the nation but also managed to decline in 2006 to a record low 0.3 percent. Fresno, CA, Seattle, WA, Honolulu, HI, and Springfield, MA also maintained relatively low homeowner vacancy rates of less than one percent, even though these metro areas, with Seattle being an exception, experienced some increases in 2006, and Fresno, CA had the highest annual jump of 450 percent from 0.2 to 0.9 percent.
Unlike the homeowner vacancy rates, rental rates movements were not consistent, with rates moving in opposite directions across different metropolitan areas in 2006. Detroit-Warren-Livonia, MI and Indianapolis, IN both experienced significant increases in vacancies and, consequently, claimed the highest rental vacancy rates, 21.2 and 19.4, respectively. The most dramatic fall in rental vacancy rates took place in Baton Rouge, LA, where rates dropped from 15.9 to 7.2 percent in one year. That was most likely a result of post-Katrina relocation in the south. Los Angeles, CA and Honolulu, HI posted some of the lowest rental vacancy rates in the nation, hovering around 4 percent.
Table 3 presents state vacancy rates. In addition to the HVS data, it also contains tabulated vacancy rates from the ACS and USPS data sources. Unlike other sources, the HVS provides data for both 2005 and 2006, thus allowing annual change comparison.
According to the HVS, the highest homeowner vacancy rates in 2005 were in Nevada (3.0 percent) and Georgia (2.7 percent) followed by Michigan and Colorado. The lowest were in Vermont (0.5 percent) and Hawaii (0.6 percent). The following year, as a sign of market weakness, vacancy rates rose significantly in most states. Only ten states experienced small declines in homeowner vacancies. In Arizona, Florida, and Vermont homeowner vacancy rates more than doubled. Despite this, Vermont’s homeowner vacancy rate remained well below the national average of 2.4%.
Historically, vacancy rates have been much higher for rental than for owned housing. According to the HVS, the national rental average in 2006 was 9.7 percent. At the state level, rental vacancy rates showed no consistent trend states between 2005 and 2006, with roughly half states experiencing declines and the other half increases. Michigan saw the biggest jump, from 13.1 to 18.1 percent (the 2006 state record level), while Arizona, Colorado, Montana, and Wyoming all experienced rental vacancy declines of 20 or more percent.
Aggregating vacancy data over the four Census regions shows that there are also significant regional differences in vacancy rates with rates being substantially higher in the South and Midwest than in the other two Census regions (Figure 1).
The HVS clearly shows that both rental and homeowner vacancy rates are much higher in the Midwest and South. Rental vacancy rates in these regions (12.4 and 11.6 percent), are almost double those of the West and Northeast (6.75 and 7.10 percent) respectively. The homeowner vacancy rates are also higher: 2.8 percent in the South and 2.6 percent in the Midwest, compared to 2.1 in the West and 1.70 n the Northeast. These variations are likely to reflect both cyclical and more permanent structural differences in regional vacancies, such as higher mobility of population and share of vacation/seasonal properties in the South, population loss to outmigration in Midwest, etc.
American Community Survey
A second source of vacancy data is the American Community Survey (ACS). The ACS is designed to replace the decennial census long form and provide equivalent data on a more timely basis. It was implemented for the first time in 2001 but only for a limited number of areas. Currently, it is the largest household survey in the US, covering about 3 million addresses per year. Owing to its wide coverage of households, information, including vacancy rates, is available for many detailed levels of geography, including states, metropolitan areas, congressional districts, and a number of counties and cities. More geographic detail will become available after data is accumulated for five years. As of this writing, 2005 is the last year for which ACS data are available, although the 2006 data are due out soon. Because it was first implemented on a wide geographic scale in 2005, the ACS does not show patterns over time yet. Another disadvantage of the ACS is that, just like any other survey-based data set, including the HVS, estimates’ margins of errors become rather large for lower levels of geography.
The US Census Bureau provides access to the ACS data free of charge. The ACS publishes data profiles with basic housing statistics including homeowner and rental vacancy rates for most geographies. In addition, the ACS detailed tables provide summary counts of occupied and vacant housing units that can be used to perform custom analysis, such as examination of seasonal vacancies.
Table 3 compares state vacancy rates computed the ACS to those found in the HVS and USPS data sets. Even though somewhat lower on average, the ACS homeowner vacancy rates are largely consistent with those published in the HVS. According to the ACS, the national average in 2005 was 1.7 percent, roughly in line with 1.9 percent reported in the HVS. Similarly, the ACS reports Nevada, Georgia, and Colorado as the states with the highest homeowner vacancy rates and puts Hawaii, North Dakota, and Vermont at the bottom of the list. Discrepancies are somewhat larger between the data sources when it comes to rental vacancy rates. According to the ACS, the average rate is 7.7 percent, two percent below the one reported in the HVS, and the differences are even larger in cases of some individual states. As a result, it is advisable to pick and work with one data source when comparing vacancy rates across different years and geographies.
The great advantage of the ACS over the HVS is its wider geographic coverage. In 2005, the ACS covered only counties with population of 65,000 or more, but eventually it will report statistics for all counties, census tracts, and block groups using multi-year averages. Figure 2 displays county homeowner vacancy rates available in the 2005 ACS.
U.S. Postal Service Records
The newest source of vacancy data comes from the US Postal Services via the Department of Housing and Urban Development (HUD). HUD recently entered into an agreement with the USPS to aggregate and publicly release Postal Service data on vacant addresses on a quarterly basis. The USPS data cover the entire universe of all addresses in the United States and thus potentially can turn into a new timely “census-like” source of data on vacancies that provides comprehensive coverage at a very detailed level of geography, including states, counties and census tracts.
However, several features make the USPS data less attractive.
The USPS data set:
does not separate residential and commercial properties;
employs an unusual definition of vacancy - as urban addresses not collecting their mail for 90 days or longer;
has an awkward “no-stat” category separate from vacant, which includes: 1) addresses under construction not yet occupied; 2) rural addresses vacant for 90 days or longer; 3) urban addresses identified by a carrier as not likely to be active for some time (e.g., if a building is being demolished to be replaced by another building, the address is preserved and considered “no-stat”);
has only been available for a limited time; not yet possible to identify local trends or see if vacancies are currently above or below some normal rate.
The USPS definition of vacant addresses is quite different from those in long-established survey-based data sources such as HVS and Decennial Census/ACS. The USPS data combine residential and commercial vacancies and do not include rural vacancies in the “vacant” category. The 90-day rule is also unique to this data set. These differences make it difficult to compare the USPS data to other popular vacancy data sources.
The USPS data are also available to the public free of charge at HUD’s website. However, these raw data are not tabulated and require additional steps before they can be loaded into Excel or any other statistical package.
As mentioned above, comparing the USPS results to the ACS and HVS is not straightforward, since the USPS data set does not differentiate between rental and homeowner vacancies, and combines residential and commercial empty properties. But analysis of these data might still bring some insights into cross-country variation in overall vacancies rates. According to the USPS data, the national average overall vacancy rate in 2006 is 3 percent. Indiana and Kansas have the two highest rates in the nation, 5.1 and 4.7 respectively. Connecticut, Massachusetts, Maryland, Virginia, and California all have low overall vacancy rates that do not exceed 2 percent. California stands out for the lowest, 1.3 percent, rate in the nation. Unfortunately, it is impossible to judge whether this cross-country variation is dominated by differences in commercial or residential vacancy rates.
Aggregating the USPS vacancy data over the four Census regions shows a configuration similar to the HVS pattern of vacancy rates that are substantially higher in the South and Midwest than in the other two Census regions. According to the USPS data, the West and Northeast regions averaged slightly over two percent (2.2 and 2.3 respectively in 2006), while the South and Midwest had vacancy rates more than a percent higher (3.3 and 3.7).
One of the biggest advantages of the new USPS data set is that it provides vacancy data at a very detailed level of geography, including county and census tract. The Figure 3 below summarizes vacancy rates for all US counties.
As mentioned above, without having access to the extended historic coverage, it is difficult to decide whether these differences shall be attributed to cyclical factors or to variation in natural vacancy rates. However, as the USPS data are refined and the longer time series are collected, the dataset will become of a greater practical value to builders deciding whether and when to downsize or expand building operations in their locality.
Return to the Top
See previous Special Studies