Understanding the relationship between public and private commercial real estate markets

ABSTRACT

This paper provides a modelling framework to examine the very low correlation at short horizons and high correlation at long horizons between private and public commercial real estate returns. For this purpose, we use a correlated, unobserved component model with a common trend and Markov-switching heteroskedasticity. This model decomposes the public and private commercial real estate prices into a common trend and interdependent cycles. The proposed model is able to endogenously capture low and high volatility regimes in real estate markets. More importantly, our model shows that the low correlation observed at short horizons between the public and private real estate markets is mainly due to the absence of any correlation in low-volatility regimes. On the other hand, the cycles, or short-run movements, in these two markets are highly correlated in high-volatility regimes.

KEYWORDS:

• Commercial real estate
• unobserved component model
• Markov-switching

Acknowledgements

I am thankful to the participants at the 2017 Midwest Econometrics Group Meeting at Texas A&M University for valuable comments on the different versions of this paper.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. See for example, Morawski et al. (2008) and Boudry et al. (2012) among others.

2. Myer and Webb (1993) and Ziering and McIntosh (1997).

3. See for example, Hoesli and Oikarinen (2012) and Yunus et al. (2012) among others.

4. Morawski et al. (2008), Boudry et al. (2012), Hoesli and Oikarinen (2012) and Yunus et al. (2012).

5. The result that private and public real estate commercial real estate returns display a high degree of correlation in stressful times is also consistent with the ‘tail dependence’ literature in financial market. For example, Zhou and Gao (2012) and Hoesli and Reka (2013) find asymmetic correlation in local and international real estate markets during the financial crisis.

6. Pagliari et al. (2005), Pagliari and Webb (1995) among others.

7. There also some studies that have studied the long-run relationship between these two markets in international context. See Ong (1994, 1995) and Yunus et al. (2012) among others.

8. See for example, Bhatt and Kishor (2015), Chauvet (1998), Hamilton (1989), Kim (1993a, 1993b, 1994), Kim and Nelson (1998), Morley et al. (2003), Morley (2007), Sinclair (2009), among others.

9. The estimated cointegrating relationship takes the form: $\mathrm{L}\mathrm{R}\mathrm{E}\mathrm{I}{\mathrm{T}}_t=0.44+1.14\ast LPCR{E}_t,$ where L represents natural log. The estimated cointegrating residual is stationary at all significance levels.

10. For estimation details, see Chapters 5, 6 of Kim and Nelson (2000).

11. The first few years of the index was sponsored by Moody’s.

12. Note that the cycles have VAR representation, so both PCRE and REIT cycles have 2 lags.

13. To examine if this result holds for a simple Hodrick-Prescott (HP) trend-cycle decomposition, we calculated the correlation between HP cycles of both the series during the financial crisis and for the whole sample. The results clearly show that the degree of correlation between cycles during the financial crisis is much higher than the normal times.

14. See for example, Hale (2012).

Additional information

Notes on contributors

N. Kundan Kishor

N Kundan Kishor is currently a professor of economics at University of Wisconsin-Milwaukee. His primary research interest is in empirical macroeconomics and forecasting. His research work has appeared in multiple journals that include Journal of Business and Economic Statistics, Journal of Money, Credit and Banking, Journal of Economic Dynamics and Control, Oxford Bulletin of Economics and Statistics among others.  He received PhD in Economics from University of Washington in 2005. Previously, he received MA in Economics from Delhi School of Economics in 2000.