Just How Accurate Is Housing Data?

Written by Michael Kling
on June 06, 2012 No Comments
Categories : Required Reading

11713_cloud_house Just How Accurate Is Housing Data? REQUIRED READING: How much home prices are changing depends on which data source you seek out. The Standard & Poor's (S&P)/Case-Shiller Home Price Index said home prices fell by 3.7% between November 2010 and November 2011. Santa Ana, Calif.-based CoreLogic reported a 4.3% year-over-year decline in home prices in November 2011, while Seattle-based Zillow reported a 4.6% decline for the same period.

Meanwhile, the home price index from the Federal Housing Finance Agency (FHFA) said prices declined by 1.8% for the 12 months ending in November, and the National Association of Realtors (NAR) said home prices fell 3.5% for that 12-month period.

How is it that five different entities measured the same subject and came up with five completely different answers? The answer is both simple and complex.

Home price indices calculate figures differently. They may collect prices from different sources, different types of properties and different geographic regions. How much researchers say home prices change depends on their methodology, or how they go about counting and analyzing real estate – for instance, if they use repeat sales or median sales, and if they count distressed sales.

Yet despite their different methodologies and results, real estate analysts say their overall results tend to be very similar over the long run.

‘Different methodologies commonly prompt different estimates for changes in home values, especially month to month,’ says Maureen Maitland, vice president of S&P Indices. ‘But you should see the same momentum in home prices. The trend over years will be pretty close.’

The S&P/Case-Shiller 20-city composite index is one of the most closely followed home price indices. This index collects its data from recorders' offices at local counties and municipalities and reports a three-month moving average of home prices, using weighted repeat sales. While covering just 20 major cities, the index covers 75% of the nation's residential real estate market by value, Maitland says.

Also, S&P/Case-Shiller uses repeat sales, or sale price changes of the same homes, which is more accurate than median sale prices used by groups like NAR, Maitland adds. It does not count new construction, condominiums, or other multifamily properties, or homes sold more than once within six months.

‘We exclude outliers,’ Maitland continues. ‘We exclude data we believe is in error or falls outside of our methodology. We don't change the data. We don't include sales between family members. We exclude anything extremely high or low for the market. We try to measure the true value of homes where transaction is arm's length.’

S&P also will not include sales of homes that have undergone substantial additions. ‘We do look for constant quality,’ she says.

Over at NAR, the association's presents its data by tracking the median sales price of homes sold through multiple listing services (MLSs). Unlike the S&P/Case-Shiller index, the Realtor group's median sales price data covers all 50 states and is not value-weighted. The latter point is important, NAR stresses, because it believes that higher-priced homes impact the S&P index more than lower-priced ones.

NAR also argues that monthly price changes are not valid because they do not compensate for seasonal changes in home-buying patterns. The trade group believes that only year-to-year changes offer a meaningful gauge of housing markets.

Although NAR's median sales produces timely results, it can be problematic when the composition of homes sold in a given time period changes – for instance, if more expensive homes are sold in a given month. Some home price indices use repeat sales, comparing sales of the same homes, to get around that issue.

‘Unlike a median sales price, the repeat-sales methodology is very complex and requires an extremely large database of home sales transactions that limits its application to the nation or to large metropolitan areas,’ according to NAR.

In comparison, the CoreLogic home price index uses data from county recorder offices and employs a repeat-sales methodology that tracks changes in sales prices for the same homes over time, which the company believes provides a more accurate ‘constant quality’ view of pricing trends than basing analysis on all home sales.

‘The two different approaches actually end up at a very similar spot,’ says CoreLogic Chief Economist Mark Fleming, adding that the similar results are an affirmation that the results are on target.Â

Zillow asserts that it overcomes the problem of changing composition of properties – which it says can skew both median- and repeat-sales methodologies – by estimating the sales price of every home, whether it is on the market or not. Because estimates are equally likely to be above the actual sales price as below it, he says, Zillow's monthly index closely tracks real values.

According to Stan Humphries, Zillow's chief economist, it would be ideal for the same set of homes to be compared from one period to the next. This approach of using a constant basket of goods is widely used, as seen with a commodity price index and a consumer price index. But unlike commodities and consumer goods, Humphries adds, sales of the same home cannot be compared, because not all homes are sold in every time period.

The FHFA creates a monthly home price index using purchase prices of houses backing mortgages that have been sold to or guaranteed by Fannie Mae or Freddie Mac, which are only conforming loans. The FHFA index uses data from all states, while the S&P index lacks data from 13 states.

Like the S&P index, the FHFA uses repeat sales. However, the FHFA index is not value-weighted. And unlike other researchers that focus on sale prices, the FHFA has an index called the all-transactions home price index that includes both refinance appraisals and purchases, in addition to its purchase-only index.

Whether or not home price indices include distressed sales can also generate different results. S&P does include distressed sales, because it considers them part of the market. NAR includes some foreclosures, but not all. Zillow does not include foreclosure sales, which typically have pricing discounts between 20% and 40%. CoreLogic releases separate home price trends with and without distressed sales.

Many economists follow the Mortgage Bankers Association's weekly data announcements of mortgage applications for home purchases. While the weekly update makes the data timelier than the monthly indices, it only covers applications, not approvals, and it does not count all-case sales.

Shadow boxing

Then there is the question of the shadow inventory, which also seems to elude a statistical consensus among data sources. In last September's MortgageOrb series ‘The Fall And Rise Of The Housing Market,’ the shadow inventory was estimated at 1.7 million by CoreLogic, at nearly 7 million by Fitch Ratings and at 3.8 million by Moody's Analytics.

Homes in the shadow inventory – defined as the supply of loans 90 or more days past due or in foreclosure – have a high probability of going into default. But trying to determine what real estate owned (REO) property is listed for sale is difficult, because sales data and servicers' portfolios are not connected.

‘It gets very complicated very quickly,’ says Herb Blecher, senior vice president of applied analytics at Lender Processing Services (LPS), based in Jacksonville, Fla. ‘Tying MLS data back to servicing is not easy.’

Foreclosing quickly on homes and pushing REOs onto the market could cause an adverse feedback loop of falling housing prices. Yet LPS has not seen that happening, says Blecher, noting that his company sees a larger foreclosure backlog in judicial states. Â

CoreLogic estimates the shadow inventory by calculating the number of distressed properties not currently listed on MLSs that are seriously delinquent (90 days or more), in foreclosure and lenders' REO. CoreLogic determined there was a five-month supply of houses in the shadow inventory as of October 2011 – a healthy market, by comparison, should have an inventory of less than one month.

Number crunched

Real estate data, however, is not immune to mistakes, or what researchers call ‘data drift.’ For instance, in December 2011, NAR revised its home sales numbers going back to 2007, admitting that it overestimated recent home sales by 15% last year. As a result, its home sale reports had been overestimated by over 14.3% between 2007 and 2010 – meaning the real estate downturn was even worst than previously reported.

What went wrong? It seems that the problem was that NAR's home sales reports are based on MLS entries from throughout the country, and do not include for-sale-by-owners (FSBO) sales or purchases directly from home builders. The downturn crushed the FSBO market, and home builders increasingly went through Realtors to sell homes, according to Lawrence Yun, NAR's chief economist. FSBOs accounted for a 16% market share in 2000, but only 9% in 2010 – a record low, according to NAR.

In addition, local MLSs sometimes reported the same home sale as desperate sellers advertised on two or more services, prompting NAR to double- or triple-count sales. Plus, some MLSs expanded their territories without informing NAR, creating the appearance of more home sales in the same population.

‘The recent drift was partly due to data error accumulating over several years,’ Yun wrote in his NAR blog.

NAR corrected its numbers – using the word ‘re-benchmarked’ to explain the update – to match other housing data sources, such as U.S. Census Bureau data and court records. Although MLS numbers are not perfect, Yun asserted, they are timely.

‘People want to know what is happening in the market as quickly as possible,’ Yun wrote, ‘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.’

CoreLogic, which had questioned the Realtor group's data last year, is now satisfied with NAR's revised numbers. CoreLogic's Fleming says NAR's numbers are now ‘in the same ball park’ as his company's.

‘Last spring, we said their numbers were tracking significantly low, which prompted a lot of questions,’ Fleming says. ‘At the time, they said they were re-benchmarking their figures.’

But, ultimately, the NAR update did not seem to disrupt confidence in housing data. Celia Chen, senior director at Moody's Analytics, says the corrections to the widely publicized home sales figures do not impact predictions for the housing market's outlook.

‘The growth rate in home sales is still the same,’ Chen says. ‘The month-to-month change is pretty much the same. Home sales have been basically flat the last two years.’

Michael Kling is a former editor of Secondary Marketing Executive and a business journalist based in Stratford, Conn.

Register here to receive our Latest Headlines email newsletter




Leave a Comment