#1108 1987 · Karl Case and Robert Shiller · Real estate economics
Case and Shiller measured home prices by comparing each house only to its own past self
the problem
Median home-price stats swung wildly with the mix of houses sold, hiding whether prices actually moved
background
Through the 1980s the only widely available gauge of how home prices were moving was the median or average sale price reported each month by real-estate boards. That number was useless for tracking true appreciation, because it mixed together completely different houses: if more starter homes sold one month and more mansions the next, the median would leap even if no individual house had changed in value at all. Economists and mortgage investors had no reliable way to separate real price movement from this compositional noise, which meant nobody could measure a housing bubble, price a mortgage security against local risk, or build a futures contract on home prices — there was no trustworthy number to write the contract against.
Karl Case, an economist studying the Boston market, and Robert Shiller, who had spent years arguing that asset markets could be driven by psychology rather than fundamentals, needed an index that isolated pure price change from the shifting mix of what sold. Simply collecting more sales data wouldn't fix it, because the problem wasn't sample size — it was that no two houses in a given month's sample were the same object being measured twice.
what everyone would do
Everyone measuring home prices simply averaged or took the median of whatever homes sold in a period, or tried to build a hedonic model adjusting for square footage and features — both leave a gap between the number and the thing anyone actually wants to know, which is what a given house's value did.
what they saw
The noise in home-price averages wasn't measurement error, it was a category mistake: no two houses are the same unit, so comparing this month's sales to last month's compared apples to oranges.
the move
Case and Shiller threw out cross-sectional comparison entirely and tracked only houses that sold more than once, pairing each property's earlier sale price against its later one. Weighting and aggregating thousands of these repeat-sale pairs produced an index that measured how the value of the same physical houses changed over time, with the mix problem structurally impossible because every comparison was a house against its own past self.
why it works
By restricting the index to repeat sales, every data point holds the physical asset constant and lets only price float, which is exactly what an index is supposed to isolate; compositional shifts in what's selling simply can't leak into the number because nothing is being compared across different properties. That structural guarantee is what let mortgage investors and later a futures exchange trust the index enough to write contracts against it.
the payoff
Their method became the S&P/Case-Shiller Index, cited by the Fed and Wall Street, and the basis of the first housing futures market.
where it breaks
The method only works where enough properties resell within a usable window; it undercounts markets or property types that rarely turn over, lags true conditions because it needs a second sale to register any move, and says nothing about homes that never sell twice, such as new construction.
what came after
The repeat-sales method is now the default technique for real-estate price indices worldwide and gave researchers, including Shiller himself, the first solid empirical evidence to identify the 2000s housing bubble while it was still inflating.
references
- [1]Prices of Single Family Homes Since 1970: New Indexes for Four CitiesNational Bureau of Economic Research, 1987nber.org