Movements in real estate uncertainty in the United States: the role of oil shocks

ABSTRACT

In this paper, we analyse the role played by disaggregated oil shocks in driving real estate uncertainty (REU) over the monthly period of 1975:02 to 2017:12, based on impulse response functions generated from the local projection method. We find that the oil-specific consumption demand shock is statistically the strongest predictor of higher future REU, followed by the significant negative impact from the aggregate supply shock, especially for long-run REU. While the oil inventory demand shock has a short-lived positive impact on REU, global economic activity shock virtually plays no role in driving the same. Our results have important implications for policymakers and investors.

KEYWORDS:

• Oil shocks
• real estate uncertainty
• local projection model
• impulse response functions

JEL CLASSIFICATION:

• C22
• Q41
• R30

Acknowledgments

The corresponding author acknowledges support from the National Natural Science Foundation of China under Grants 71974181, 71774152, and Youth Innovation Promotion Association of Chinese Academy of Sciences Grant Y7X0231505.

We would like to thank two anonymous referees for many helpful comments. However, any remaining errors are solely ours.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ According to the Financial Accounts data (downloadable from: https://www.federalreserve.gov/releases/z1/20190920/html/b101h.htm) of the US corresponding to the fourth quarter of 2018, residential real estate represents about 83.7% of total household non-financial assets, 28.3% of total household net worth and 24.6% of household total asset.

² The MU and FU indices are available for download from the website of Professor Sydney C. Ludvigson: www.sydneyludvigson.com/data-and-appendixes.

³ The REU index is downloadable from the website of Professor Johannes Strobel: https://sites.google.com/site/johannespstrobel/.

⁴ The maximum length of forecast horizons is set to 24, which corresponds to 24-month forecast horizons. ${\gamma }_s\left(L\right)$ is a polynomial of order 12, which corresponds to 12-month lags for control variables.

⁵ Let us suppose that we want to generate an IRF for a shock to x_t on itself, i.e. we consider an univariate process. The first step is to choose the size of the shock u. At the time of the shock, E[x_t], the point estimate for the shock at the time of impact (which has been set as period t) is then simply x + u, where x is typically the mean. The next step is to choose, how many lags to include in the estimated autoregressive process. For the sake of simplicity of exposition, suppose we choose two lags, though in our case, we work with 12 lags. The next period of the IRF function is then obtained by regressing x_t on two lags of itself, i.e. x_t = α+ β₁x_t-1+ β₂x_t-2+ u_t. The IRF estimate for the period after the shock is then: E[x_t+1] = α+ β₁($\bar{x}+\bar{u}$)+β₂$\bar{x}$, and the confidence interval will be computed using the standard errors of the regression coefficients. At this point, the local projections method deviates from the traditional approach used in the VAR (as it does not rely on the iteration of older and possibly misspecified expectations). It forms E[x_t+2] and each subsequent period using a separate OLS regression. In order to obtain the IRF for t + 2 one needs to regress x_t on x_t−2 and x_t−3, i.e. i.e. x_t = α+ β₁x_t-2+ β₂x_t-3+ u_t, and again E[x_t+2] = α+ β₁($\bar{x}+\bar{u}$)+β₂$\bar{x}$, and it is straightforward to compute the confidence interval. However, the coefficients now take different values, with each period getting its own regression. Further technical details about LP IRFs can be found in Jordà (2005).

⁶ Based on the suggestion of an anonymous referee, we found that standard linear Granger causality test reveals strong influence of the OSS for REUs at all horizons, and EAS and OIDS significantly impacted REU12 and REU3 respectively. Complete details of these results are available upon request from the authors.

⁷ As a robustness check, we analysed the impact of the four oil shocks on the conditional volatility of nominal housing returns, often measured as housing market uncertainty (see, for example, Christidou and Fountas (2018)), derived using the exponential generalized autoregressive conditional heteroscedasticity (EGARCH) model of Nelson (1991) (since it produced the best fit among alternative symmetric and asymmetric GARCH models). Note the seasonally-adjusted housing price data is derived from Freddie Mac (with the data downloadable from: http://www.freddiemac.com/research/indices/house-price-index.page), and the model is estimated over the same sample period of 1975:02 to 2017:12 for the sake of comparability. This house price index provides a measure of typical price inflation for houses within the United States and is based on expanding database of loans purchased by either Freddie Mac or Fannie Mae. As observed from Figure A1 reported in the Appendix, the IRFs follow a general pattern as those reported for REUs, but are more volatile. Interestingly, in this case, the global economic activity shock is found to produce statistically the most pronounced negative impact on housing market volatility, with delayed statistically significant intermittent negative effects also observed under the aggregate supply shock. The oil inventory demand shock is found to produce delayed positive short-lived impact on housing returns volatility. Complete estimation details of the underlying EGARCH model is available upon request from the authors.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [71774181, 71774152].