In Depth Analysis: CalculatedRisk Newsletter on Real Estate (Ad Free) Read it here.

Wednesday, May 23, 2012

Kolko: Dissecting the House Price Indices

by Calculated Risk on 5/23/2012 05:16:00 PM

CR Note: This is from Trulia chief economist Jed Kolko:

Dissecting the House Price Indices

Each month, several data releases track house price changes. Case-Shiller, CoreLogic, the Federal Housing Finance Agency (FHFA), the National Association of Realtors (NAR) and others report monthly sales-price trends, and the Trulia Price Monitor reports trends in asking prices, a leading indicator of sales prices. These indices often show different trends even for the same time period. Some of the differences among these indices are well-known, such as the fact that FHFA’s traditional index is based on transactions involving conforming, conventional Fannie Mae & Freddie Mac mortgages, while other indices (including the newer FHFA expanded-data index) cover a broader set of homes. But other, more technical differences help account for why some indices go up while others go down, including how they handle:

• The mix of homes listed and sold.
• Seasonal patterns in home prices.
• Weighting of homes and metros.

How much do these issues really matter for price trends? A lot, it turns out. In constructing the Trulia Price Monitor, we (1) adjust for the mix of homes listed, (2) adjust for seasonality, and (3) “weight” homes equally so that our national trend best represents what’s going on with the typical home in the largest 100 metros. Using this approach, we found that asking prices nationally rose 0.2% year-over-year and 1.9% quarter-over-quarter in April. Other price indices take different approaches, and mix-adjustment, seasonal adjustment, and value-weighting all have pros and cons. To see how much these issues matter, we used our data to see what the price trends would look like using different technical approaches.

Mix of homes listed and sold.

The price of a home depends on its size, location, and many other factors. For example, if larger homes or homes in more expensive neighborhoods happen to be listed or sold, the average (or median) listing or sales price will rise. That doesn’t mean that the typical home has increased in value – which is what most owners, buyers, sellers, and investors really care about. To know how the typical home’s value has changed, most price indices adjust for the mix of homes that are listed or sold, either by factoring in specific attributes of the home like its size and location (hedonic models, which the Trulia Price Monitor and FNC use) or by looking only at how prices have changed for the same home over time (repeat-sales models, which Case-Shiller, FHFA, and CoreLogic use). How much does the changing the mix of homes matter? The Trulia Price Monitor for April 2012 showed that prices nationally increased 0.2% year-over-year; this adjusts for the mix of homes. But without adjusting for home size, neighborhood, and other factors, the median listing price increased by 8.1% year-over-year. That’s a huge difference and that can partly be attributed to the fact that homes listed today are, on average, 6.2% larger than a year ago. They also tend to be located in slightly more expensive neighborhoods.

The shift toward larger homes on the market means that price indices that don’t adjust for the mix of homes are showing much larger price increases than what the typical home is experiencing. So why look at any price trends that don’t adjust for the mix of homes? Unadjusted price trends do reflect how typical transaction amounts are changing, which affects real estate commissions and the health of the real estate industry.

Seasonal patterns in home prices.

Home prices – both asking and sales – follow predictable seasonal patterns, dipping in winter and rising in spring and summer. (Other housing activities, like sales volume and construction starts, swing even more with the seasons than prices do.) Comparing home prices at the same time of the year takes out any seasonal effect, but quarter-over-quarter or month-over-month changes can be strongly affected by seasonal patterns.

The Trulia Price Monitor for April 2012 showed that prices increased nationally quarter-over-quarter by 1.9%, seasonally adjusted, but by 4.8% without adjusting for seasonality since the adjustment removes the regular springtime price jump.

Seasonal adjustment has its challenges. If the seasonal pattern changes over time – like if winters get warmer and cause housing activity to drop off less in winter – seasonal adjustment methods need to reflect those changes. Not-seasonally-adjusted trends are still useful because they show what buyers and sellers are actually experiencing in the market right now and can help them time when in the year to buy or sell. But to detect if and when housing prices are finally reaching a sustained turnaround, seasonal adjustment is needed to distinguish the underlying trend from regular seasonal patterns.

The Trulia Price Monitor, Case-Shiller and FHFA report seasonally adjusted price changes, even though Case-Shiller emphasizes the non-seasonally-adjusted trends. Most other indices only report non-seasonally-adjusted trends.

Weighting of homes and metros.

In the Case-Shiller and CoreLogic indices, higher-priced homes count more – they are “value-weighted”; in contrast, the Trulia Price Monitor, the FHFA index, and most other indices don’t put extra weight on higher-priced homes. Why give more weight to pricier homes? Higher-priced homes should get more weight if the purpose of an index is to assess movements in the value of a real-estate portfolio. If, for instance, a $1,000,000 real estate portfolio consists of two homes, one initially worth $900,000 and one initially worth $100,000, the change in the overall value of the portfolio depends a lot more on the percentage change in the value of the $900,000 house than the $100,000 house. In other words, weighting by home price yields an index that shows how the value of a dollar invested in real estate changes. The Trulia Price Monitor weights homes equally, regardless of price, in order to show how the value of a typical home is changing – rather than the value of a dollar invested in real estate. (FHFA doesn’t use value-weighting, either.)

Value-weighting potentially matters a lot for price trends if high-priced and low-priced homes in a market are trending differently. We tested the potential impact of value-weighting by comparing the year-over-year price change in several metro areas from the Trulia Price Monitor with the price change we would have reported for those same metros over the same time period but with value-weighting. The Trulia Price Monitor for April 2012 showed that prices in New York decreased 2.6% year-over-year, but with value-weighting prices decreased just 0.3%. In Los Angeles, the Trulia Price Monitor showed that prices decreased 2.8% without value-weighting but increased 0.7% value-weighted. In Phoenix, the Trulia Price Monitor showed that prices increased 15.8% without value-weighting, but increased only 11.1% value-weighted. In short, value-weighting can change the price trend either up or down by several percentage points – a big difference for what sounds like an obscure technical issue.

Finally, value-weighting can lead to expensive metro areas counting heavily in a broad home price index. The New York and Los Angeles metros together account for 48% of the Case-Shiller Composite 10 index and 35% of the Composite 20 index (based on weights in the published methodology) -- not only because those metros are large but also because they are expensive. At the same time, Houston and Philadelphia, which are among the ten largest metros in the US, are not included in the Case-Shiller Composite 20 – even though much smaller metros, like Charlotte, NC, and Portland, OR, are.

To sum up: home-price indices can disagree with each other by several percentage points depending on whether they adjust for the mix of homes, whether they adjust for seasonal patterns, and how they weight homes and local markets in the index. These technical issues help explain why different indices looking at the same market at the same time can tell very different stories.