NAR_grey_logo-01

Q&A on Re-Benchmarking of Home Sales

NAR takes its role as a leading source for housing information very seriously, and toward that end, NAR research will announce results of a year-long re-benchmarking process for existing-home sales on Wednesday, December 21. Here are some answers to commonly asked questions about the re-benchmarking process. You can also view this video for more details.

Q. Existing Home Sales figures are being downwardly revised, based on new re-benchmarked figures. What is meant by re-benchmarking and why go through with it?
A. All economic data should undergo third-party validation, if possible. In the case of home sales, there are other metrics that are closely related. For example, a home sales trend could be gauged from courthouse deeds information or from Census information regarding the movements of people. Given that these other data sources are available and which, by the way, are not based on REALTORS®’ data entries through Multiple Listing Services, we should occasionally run a validity check and compare these other data to ours.

Q. Is the re-benchmarking issue unique to home sales measurement, or is this process also common in other economic data?
A. Revision and re-benchmarking are very common in economic data. Because nearly all measurements make assumptions, any changes related to those underlying assumptions will lead to a drift in the measurement. One of the most important economic data is GDP (gross domestic product), produced by the Department of Commerce, and this data undergoes frequent revisions. For example, it was found in 2010 that the economic downturn of 2007 and 2008 was much sharper than initially thought.

Jobs figures also get revised from time to time. In fact, the Bureau of Labor Statistics provides two separate employment data sets applying different methodologies, while noting the strengths and weaknesses of both methods. For example, there are 140.8 million people with jobs in one survey while there were 131.7 million people with jobs in the second methodology as of October 2011. The measurement of ‘true’ unemployment is said to be even more difficult.

Q. What is getting revised related to Existing Home Sales and why?
A. Home sales figures are getting revised because there appears to be a notable upward drift in the data compared to other data measurements such as courthouse deeds records. There are no changes to median home prices across the regions. NAR’s monthly home sales data are based on feeds from Multiple Listing Services across the country because they provide very timely information about recent trends. People want to know what is happening in the market as quickly as possible, rather than waiting several months or even several years, which would be the case if one wanted to gauge housing market trends purely from courthouse records or the Census data. However, there is a drawback to relying on home sales data only from the MLSs because there are home sale occurrences that are not listed on MLSs.

Q. What are some examples of home sales not going through MLSs and why does that matter in the home sales figure?
A. Let’s say there are 100 total home sales in a community in a certain year. 80 of the homes were sold by REALTORS® via MLS, while 20 were for-sale-by-owner (FSBOs). Assume the next year the total home sales did not change and remained at 100. But this time, more homes were sold through MLSs, say 89, with only 11 being done as FSBOs. From the total market point of view, there is no increase in home sales. However, from the MLS point of view, home sales rose 11 percent. Also from the NAR point of view, based on the MLS data feed, it appears as if home sales rose 11 percent from that region. From 2007onwards there have been substantial declines in FSBO sales.

Another example is when homebuilders sell homes by themselves without seeking REALTOR® help. But in the recent tough housing market years, homebuilders have been listing homes on MLSs and providing commission income to REALTORS® who bring clients. Some MLSs are able to clearly mark this kind of home sale as a new property rather than coding it as existing property, and NAR would not include these new home sales into the tally. But other MLSs do not provide a distinction, so inevitably some of the recorded home sales are likely to be newly constructed homes, even though NAR is trying to capture and report sales activity only for existing homes. (The Census Bureau reports activity on the new home sales market).

Q. Using the above hypothetical example, what’s the big deal about reporting home sales as having risen since REALTORS® did more business than before?
A. From a REALTOR® point of view, perhaps only the trends in REALTOR®-assisted home sales matter. But from a public policy point of view, total home sales and not just REALTOR®-assisted sales matter. Issues like changes to mortgage interest deduction or the homebuyer tax credit stimulus gauge the impact on all home sales and not just REALTOR®-assisted sales. Even the general inventory absorption patterns are determined by overall home sales and not just REALTOR®-assisted sales. So reporting on only the MLS home sales count can bring about misleading impact analysis, which is the reason for re-benchmarking from time to time.

Q. Are there other reasons for MLS data drift that are not a good reflection of market changing trends?
A. There are two key additional factors that lead to upward drift in the MLS data. First, some of the local MLSs have enlarged their geographic coverage areas (like expanding into outlying counties). In these instances, it is possible to get an overstatement in the home sales count that is unrelated to more home buyers going to the closing table. If not properly informed by local MLSs of these changes, NAR would assume more home sales in this region because of more buyers, rather than because the local MLS is picking up home sales from new outlying areas not covered before. Second, there could be double or triple counting of a single home sale transaction. A home listed in Colorado Springs also gets listed in a separate MLS in Denver. From this single transaction, the Colorado Springs-based MLS is sending info about this transaction to NAR, but so is the Denver-based MLS. A genuine example of double-counting. Again, re-benchmarking will address this issue since home sales count will not be double-counted in theory using courthouse records or Census data about movements of people.

Q. Given the potential bias in home sales as discussed above, perhaps I should monitor other housing data? Which other housing data are good to follow?
A. There are pros and cons of all data. One data set many economists follow related to housing is from the Mortgage Bankers Association about mortgage applications for a home purchase. It is very timely because it is compiled weekly, and provides a way to gauge the rising or falling trends of people applying for mortgages. However, this data is only about applications and not approvals. Furthermore, it completely leaves out home sales that are all-cash deals. In addition, this data also has drift, in fact quite large, over the longer term. For example, from 1990 to 2005, the peak housing years, mortgage purchase applications for home purchases (excluding re-finances) rose by 500 percent. Yet, new home sales rose by only 200 percent and existing home sales by 100 percent. So there has been a massive upward drift in the mortgage application data going up. However, this data is not benchmarked because there is nothing to benchmark against. So most economists place value on the general week-to-week trends about market strength, but tend to not use this data for a long-term time series analysis.

The new home sales figure is another important housing data series that economists follow. This label is a bit misleading, however, because they record contracts to buy a newly constructed home and not actual closings. There is no separate information about closings. This data does not go through a re-benchmarking process so it is hoped that any data drifts – upwards and downwards – wash out over time. Still, it is considered to be reasonably reliable because it closely tracks with the new housing starts figures, which in turn track well with housing permits data from county sources.

If you are trying to gauge the housing market direction using Census or courthouse information, then you would be looking deep into only the rear view mirror because of long lags in data availability.

For home price information, generally a repeat price index will provide a better reading on price changes rather than the median price. Federal Housing Finance Agency, Freddie Mac, CoreLogic, LPS, and Case-Shiller are among the data sources for this repeat price index. One drawback to these measures is the time lag in getting the data. NAR’s median price provides good information about the type of homes being sold in a given month, though at times can be misleading if people view it as measuring price change trends. For whatever reason, if only the upper-end is moving, then the median price of sold homes will be quite high, and vice-versa if mostly foreclosed properties are being sold. Neither case would imply strong price appreciation nor strong price depreciation. Over several months, though, one would think that there would be a representative distribution of high-end and low-end home sales, so the median price does capture changing price trends. But given some weaknesses in the proper interpretation of median price, NAR has been exploring a way to create the most timely repeat-price index using MLS live feeds.

Q. Why should I trust NAR data?
A. NAR has been the source of key housing measurements for many years. The Federal Reserve, many government agencies, Wall Street firms, real estate practitioners, and consumers have looked to NAR to provide this important information about trends in home sales. As such, NAR consistently strives to provide the most accurate reading of the housing market conditions by undergoing necessary re-benchmarking processes as needed by working closely with housing economists from government agencies and private organizations to develop the latest methodology.

The recent drift was partly due to data error accumulating over several years. One way to lessen the drift is to do more frequent re-benchmarking and NAR will be re-setting the data every year moving forward, which should greatly lessen the drift. In the past, one had to rely on Decennial Census data (every 10 years). Today, with improvements in other Census data collection about movements of people and from better coverage from courthouse deeds, the annual re-benchmarking process looks to be feasible.

Q. Will there be a re-statement of home sales by local MLSs?
A. No. That is not required. The purpose of a MLS sales count at the local level is just that: how many homes were sold through the MLS. The data drift problem is at the national level where a re-benchmarking is required to assess total existing home sales to account for changes in FSBOs,adjusting for flipping, and such.
However, there will be a state-level sales revision since re-benchmarking can be done at the state level.

Q. Should consumers care about this revision and re-benchmarking?
A. What matters for consumers is the price of their home and there is no revision to prices. For home-buyers and home-sellers, the newly revised national data has no impact, because only very localized data matters to consumers making decisions. Local REALTOR® experts can help consumers with the local data. The national data is important for policymakers to measure the broad strength and weakness of the housing market as they impact the national economy. Like weather, the only thing that truly matters for consumers are outlooks on data from their local market area. One thing observed in the re-benchmarking process for consumers is that the FSBO method of selling suffered greatly during the housing market downturn.

Q. Does the re-benchmarked data maintain the characterization of market swings that we observed in the past?
A. Generally yes. The strong home sales surges that were observed in the final month of the home buyer tax credit availability and the subsequent sales slump in the immediate months afterwards are still present after the data revision. Other characterizations of market changes from month-to-month also will continue to exist after the re-benchmarking. So even with the upward drift, the information about falling, rising, or stabilizing trends still holds true and provides key information about market direction. The re-benchmarking process will just make the data much better in regards to not only the direction of sales but also to the level of sales.

Q. How are home sale flips handled? It would appear that house flipping is an artificial demand (and then a supply) that adds nothing to the market.
A. From MLSs, flipping of all kinds is captured. Commission incomes related to sales are also present. However, from the re-benchmarking point of view, a resale of a home within 12 months is not included. There are several reasons for this. First, the second count is not feasible in the Census data that is used for the benchmarking. Second, from a public policy point of view, the flips are not considered important. Third, from a market point of view, inventory is absorbed and then added later, resulting in a wash. One could argue that flips are important since many are related to investors buying up an awful-looking home and turning it into something nice. Value was added to the home as well as to the community. But the lack of Census information prevented us from including it ino the re-benchmarking.

Q. Will there be state-level sales revisions? Why do the regional re-benchmarked sales figures show some disparities?
A. NAR will come out with state-level re-benchmarked figures in February. There are no changes to MLS data at the state or local level. NAR’s state-level figures will be based on third-party data, mostly from the Census information about people moving. (In general, all private industry data that is considered an important economic indicator should go through a third-party validity check, if possible.) The Census-based re-benchmarked home sales figures contain non-MLS home sales, such as FSBOs. So NAR’s (Census-based) data and the MLS data will not match perfectly in terms of levels, though the market swing characterizations from quarter to quarter would largely be the same as before.

Q. I want to know more about the detailed methodology. Where can I find it?
A. The re-benchmarked historical data files and methodology can be found here, along with a video explaining the re-benchmarking process.

Lawrence Yun, Chief Economist

Lawrence Yun is Chief Economist and Senior Vice President of Research at NAR. He directs research activity for the association and regularly provides commentary on real estate market trends for its 1 million REALTOR® members.

More Posts

Comments
  1. No wonder the numbers are so high for the ratio of homes bought by investors as a whole during the boom years if the sale by the flipee is not counted. This would really skew the numbers. Assume counted sales of 100,000 units with 40,000 being counted as investors meaning a 40% investor to resident owner ratio. Again, assume that 50% of the investors flipped and sold in the same year or 20,000 units would not be counted. If they were, there would have been 120,000 sales and investor ratio would be 30% of total. That is quite a difference.