Dynamic Correlation: A Tool for Hedging House Price Risk?

Article excerpt

Executive Summary.

Dynamic correlation models demonstrate that the relationship between interest rates and housing prices is non-constant. Estimates reveal statistically significant time fluctuations in correlations between housing price indexes and Treasury bonds, the S&P 500 Index, and stock prices of mortgage-related companies. In some cases, hedging effectiveness can be improved by moving from constant to dynamic hedge ratios. Empirics reported here point to the possibility that incorrect assumptions of constant correlation could lead to mis-pricing in the mortgage industry and beyond.

This study estimates bivariate dynamic correlation models for housing price indexes and financial market time series. The latter include Treasury bond rates, the S&P 500, and individual common stocks sensitive to defaults on residential home mortgages. The primary motivation for analyzing these correlations is twofold: first, to provide a quantitative description of time patterns in the linear relationships between housing market variables and financial markets; and, second, to look for potential cross-hedging instruments against housing price volatility in different regions of the United States. No attempt is made to estimate a structural model. Rather, the goal is to see whether correlations used in previous studies of hedging in housing markets are statistically constant with respect to time and, if not, whether dynamic correlations can be predicted (in a forecasting rather than structural sense) using highly liquid securities that are easily used as hedging instruments and thought to be structurally linked to home prices.

The consequences of homeowners' lack of access to insurance against declines in home values and the broader economic impact of sharp movements in mortgage default rates have been described by case, Shiller, and Weiss (1996). This paper relaxes their assumption that correlations between home values and other assets are constant with respect to time, a maintained assumption in nearly all the literature in this area. There would likely be shifts in policy for mortgage industry decision makers, whose job it is to price risk, if real and financial asset markets were found to have systematically time-varying correlation. Difficult-to-forecast dynamic correlation may also help explain why financial-market innovators have so far provided few practical hedging instruments for average homeowners.

The major stakeholders in developing new hedging instruments include homeowners, builders, mortgage holders, insurers, and mortgage-backed securities companies. Although major home finance firms such as FNMA and FHLMC have "risk sharing" operations, these transactions are all over-thecounter, creating significant transaction costs and illiquidity. Case, Shiller, and Weiss (1996) have suggested the establishment of futures or options markets for residential real estate prices, but liquid markets for options based on region-specific indexes seem unlikely to emerge in the foreseeable future.1

There is a possibility, however, that existing futures and options markets for the S&P 500 and U.S. Treasury Bonds might provide a partial solution because of their correlations with home prices. With the aid of dynamical correlation models, one hopes to discover cross-hedging strategies that could be built using relatively liquid instruments, such as the S&P 500 and T-bond futures and options, based on their predicted time paths and the relevant correlations. In this paper, these frequently studied hedging instruments are augmented by the inclusion of publicly traded real estate investment trusts (REITs) and the common stocks of firms in the homebuilding and mortgage insurance industries.

A small but growing collection of empirical studies on the relation between housing prices and the stock market indexes has emerged in recent decades. Research examining the relation between securitized real estate indexes (REITs) and the stock market has produced conflicting views. …