OUP - Abstract

Evaluating Alternative Methods of Forecasting House Prices: A Post-Crisis Reassessment

Dissertation School: The George Washington University

Abstract:

The recent, dramatic declines in house prices have drawn attention to our ability to forecast house prices. In this essay, I directly address two questions: 1) could econometric forecasts have predicted the recent downturn in house prices before any declines in house prices were actually observed; and 2) when did house price forecasts that predicted house price declines first warn us that a decline in house prices was imminent?

There are two curious traits in the housing market data that suggest certain types of forecasting models may perform well. First, in the run-up to the recent house price declines, house prices were far above where they should have been given established long-run relationships among house prices, rental prices, and personal incomes. This suggests that error correction models grounded in economic theory, as suggested by Malpezzi (1999), Gallin (2006), and Gallin (2008) may hold the key to predicting turning points in the housing market. Second, a large, negative acceleration in house prices is observed in the last quarter of 2004. Random acceleration models and Hendry's (2006) differenced vector equilibrium correction model may capture this deceleration and forecast successive quarters of decreasing growth, ultimately leading to declines in house prices.

My dissertation research directly addresses several of HUD's stated strategic goals. Better house price forecasts will help reduce mortgage delinquencies, defaults, and foreclosures by ensuring that consumers know more about price risks when borrowing to purchase a home. Knowledge of turning points in the housing market can also be used by HUD and participants in the Home Loan Modification Program to reduce strategic defaults and improve program performance. Lenders, investors, and financial institutions would benefit from better house price forecasts by being able to better assess the default risk for individual loans and for their loan portfolios as a whole.

In general, homeownership is more sustainable and stable when both consumers and lenders enter into financial arrangements that are predictably beneficial and sustainable for both parties. Policymakers, able to detect bubbles and turning points in the housing market before they occur, would be able to implement counter-cyclical policies and prevent housing bubbles from growing larger, smooth market corrections, or prevent downturns all together. A healthier and less risky housing market with better informed policymakers would ultimately result in fewer foreclosures and less systemic risk, resulting in a more stable financial system.

To answer these research questions, I propose to estimate a number of different house price forecasting models using data up until the pre-crisis peak, and forecast over the following period of dramatic house price declines. I will evaluate and compare the resulting forecasts along a number of dimensions consistent with the forecasting literature, including mean-squared forecast error and bias. I will also perform parameter constancy tests and forecast encompassing tests in order to determine which models contain relevant and useful information. Preliminary results indicate that declines in house prices were forecastable before ever observing any actual house price declines. Multivariate error correction models based on economic theory and univariate models that incorporate house price acceleration changes offer the most promise of predicting and early-detecting turning points in the housing market.

Back to Search Result of DDRG Dissertations