This is a recent research by Byron Tsang of Virginia Tech and Xiaojin Sun of University of Texas at El Paso. The focus of the research is to find out if there is any “predictable” pattern following housing “bubbles”.
To answer this question, the two examine the house prices of 50 states (plus the District of Columbia) as well as 402 metropolitan statistical areas in the U.S. over the past 40 years. They define the “run-up” of a “bubble” episode as a cumulative real housing return of 15% or more in the past year as well as 20% or more over the past four years.
The research has two significant findings. First, a sharp house price increase usually predicts positive returns for about two years ahead and negative returns after. The black line in the figure below plots the average return in housing prices after the “run-up” episodes, separately for states and metropolitan statistical areas (MSA). As you can observe in the graph, a rapid rise in housing prices would not be followed by a crash immediately; the crash will happen two years later, on average.
The researchers further separated the post “run-up” price movements into “Crashes” and “Non-Crashes”. Crashes scenario defined as there is not much difference between crashing and non-crashing markets within the first year after an identified price run-up.
House prices start to drop in markets that will end in crashes right after the first year, whereas in markets that will not crash prices keep going up for another year. Eventually, prices fall in both types of markets, to a greater degree in crashing than non-crashing markets. There is no apparent difference between the state-level and MSA-level data.
The second investigation of Tsang and Sun’s research is the probability of a “crash” following a consistent “run-up” in housing prices.
Their result suggests that a sharp increase in real house prices “predicts” a substantially higher probability of a crash, as they found a clear pattern that the crash probability increases with the size of the price run-up.
At the state level, the probability of a crash following a 5% annual price run-up is only 13%. But then the probability rises rapidly to 35% if the annual run-up is 15%, and the probability increases further to 67% at past annual run-up of 25%. A similar pattern can be identified as well in MSA-level data.
Also, the magnitude of the drawdown increases with the size of the run-up in general. The average fall in price following a 5% annual price run-up is only 9% at the state level and 7% at the MSA level, but it increases to 20% and 22% at past annual run-up of 15% and further increases to 35% and 47% at prior annual run-up of 25%.
You can read the research in full here.