Above is a scatter plot of annualized excess geometric average returns (above the Fed Funds) against annualized standard deviation in returns, for housing prices using the Case-Shiller index. The index has monthly observations back to 1987, for 17 municipalities (some start in 1991), for low, medium, and high-priced sectors. Thus, every observation is for a sector/city, over the 20 year sample.
As you can see, the average annual return is about 0.5%. That's the annual excess returns. After expenses, which for houses include property taxes of about 1-2%, plus broker fees (about 6% to sell), I think its safe to say the returns are indistinguishable from zero. Residential real estate has generally been thought to be a superior investment over the past 20 years, but I think this is mainly due to selective anecdotal inference.
There does seem to be a small volatility premium, in that the more volatile series had slightly higher returns. The volatility was low, consistent with intuition, and higher vol does seem to go with higher return.
Here we see a housing price index from Case-Shiller, and a housing price index that subtracts the opportunity cost, as reflected by the Fed Funds rate, labeled the Tot Ret Index. The main reason why our intuition about home prices is much rosier than reality, is that the returns that seem so certain, includes a long stretch in the nineties where the cumulative return was minuscule while Fed Funds averaged about 5% annually. A benchmark is essential, because, like an equity fund manager making money in 1999, if you ignore the opportunity cost, many investments look like great, when in fact they were not.
But what is truly amazing is that we see the improbable event that are all to common for financial time series. Prices have fallen about 20% cumulatively over the past year and a half(these are excess returns). Thus, from a standard deviation perspective, it was one of those five-sigma events (annualized vol was about 4%). Fat tails are part of almost every financial time series.