TL;DR
This paper proposes a novel method for detecting real estate bubbles by analyzing the dispersion of property prices, rather than just their average. They find that during bubble periods, price distributions exhibit a "power-law tail," indicating increased inequality, which is linked to a lack of price arbitrage across regions. This approach is validated using historical land price data from Tokyo and U.S. housing data.
What Problem Does This Paper Address?
Identifying real estate bubbles is crucial for financial stability, as their collapse can trigger crises. Traditional methods often rely on price indices, but it's difficult to distinguish a genuine rise in fundamental value from a speculative bubble. The authors tackle this by proposing a new indicator: the cross-sectional dispersion of real estate prices. They hypothesize that during bubbles, price inequality among properties increases significantly, leading to a heavier "tail" in the price distribution, a phenomenon they term a "power-law tail." This increased dispersion, they suggest, is a more reliable signal of a bubble than just a rapid price hike.
Methods
The core methodology involves analyzing the statistical distribution of real estate prices. Specifically, the authors examine:
- Cross-sectional price distributions: They analyze how prices are distributed across many properties at a given point in time.
- Power-law tails: They look for evidence of power-law distributions in the upper tail of these price distributions, which signifies a disproportionate number of very high-priced properties.
- Lognormal distributions: They compare the observed distributions to lognormal distributions, a common model for many natural phenomena, and investigate deviations.
- Spatial heterogeneity: They explore whether differences in price distributions within smaller geographical sub-regions (compared to the entire area) can explain the observed power-law tails.
- Deviation from normality: They quantify the deviation of price distributions from a standard normal distribution using a statistical measure (D√N).
Datasets
- Japanese Land Prices: A dataset of land prices for approximately 30,000 sites across Japan, collected annually by the Land Appraisal Committee, covering the period 1974-2008. The focus was on the Greater Tokyo Area for residential land.
- U.S. House Prices: Data collected from Zillow.com for houses for sale as of September 2011, including price information and attributes like size, address, and year built. This data was analyzed by state.
Key Insights and Ideas
- Dispersion as a Bubble Indicator: The most significant insight is that increased price dispersion, not just price increases, is a hallmark of real estate bubbles. This suggests that bubbles create greater inequality in property values.
- Power Laws and Arbitrage: The emergence of power-law tails during bubbles is linked to a lack of price arbitrage across regions. This implies that during speculative booms, market inefficiencies prevent prices from converging, leading to extreme values.
- Spatial Heterogeneity: The paper provides strong evidence that during bubble periods, price distributions become spatially heterogeneous. While prices might be lognormally distributed within small, homogeneous regions, the means and variances of these distributions differ significantly across regions, creating the overall power-law tail.
- U.S. State-Level Bubbles: The U.S. data reveals that states experiencing housing bubbles show a heavier tail in their house price distributions, supporting the generalizability of the findings.
- Foreclosure Correlation: The deviation measure from normality (D√N) shows a statistically significant correlation with foreclosure rates, suggesting it could be a predictive indicator of post-bubble financial distress.
Key Results
- Tokyo Land Prices: During the late 1980s bubble in Tokyo, land prices exhibited a power-law tail, while before and after the bubble, the distribution was close to lognormal.
- U.S. Housing Prices: States that experienced housing bubbles showed a heavier tail in their house price distributions.
- Spatial Homogeneity vs. Heterogeneity: In non-bubble periods, land prices in Tokyo were found to be spatially homogeneous (close to lognormal even across larger areas). During the bubble, prices became spatially heterogeneous, with significant differences in price distributions across regions.
- Foreclosure Prediction: The proposed deviation measure (D√N) correlated significantly with foreclosure rates, suggesting its utility in assessing potential financial fallout from housing market downturns.
Limitations and Future Work
- Data Limitations: The authors acknowledge that the Japanese land price data consists of appraisal prices, not actual transaction prices, which is a potential limitation compared to the U.S. house price data. The U.S. data is also a snapshot from 2011, lacking historical time-series analysis for that market.
- Finite Size Effects: In the analysis of spatial heterogeneity, the authors note that "finite size effects" might influence the convergence to normal distributions, particularly for smaller region sizes, and suggest this warrants further investigation.
- Mechanism of Heterogeneity: While the paper proposes spatial heterogeneity as the source of power-law tails during bubbles, the exact mechanisms driving this heterogeneity (e.g., liquidity shortfalls) are areas for further exploration.
- Predictive Power: While the deviation measure shows promise in correlating with foreclosure rates, further research could explore its predictive power for other financial indicators or for forecasting future bubble formation.
Summary updated