Don’t Trust Case-Shiller Macro House Price Trends: Most of Real Estate Price Variation is *Very* Local

Last week, the think tank Bruegel hosted an event in Brussels on the use of micro data for the evaluation of the impact of investment at the macro level.

I gave a presentation at the panel. The main point is that macro stats on house prices such as the Case Shiller index explain very little of the overall variance of house prices. In practice a simple variance decomposition of house prices on metro area fixed effects, with transaction level data, will reveal that only about 35% of overall variance is captured by the Case Shiller. That means that 65% of the total variation of house prices is local.

Talking about prices in New York vs. prices in Montreal isn’t that informative compared to comparing prices in Brooklyn versus prices in Mile End.

Perhaps even more importantly, as prices rise overall in a metro, prices can fall significantly in neighborhoods. The most significant example is that of San Francisco, where my data shows that there was a compression of the price distribution during the house price boom of 2000-2006 (yes, there’s another one now).