A Comparative Analysis of the Performance of Hotels As Against Traditional Commercial Property in a Mixed-Real Estate Portfolio

Abstract

This article considers the role of hotels in a mixed-real estate portfolio. Using hitherto unexamined IPD data for leased hotels in the U.S. and the U.K. between 2001 and 2013, it investigates whether hotels provide diversification benefits and seeks to critically reassess the view that hotels are uniformly high-risk assets. The study finds, by comparison to more “traditional” property types, that hotels have received little attention from property researchers and tend to be overlooked in the asset allocation process. This might be a result of the commonly held perception that hotels are “alternative” and “high-risk.” Based on results from a portfolio optimization analysis, it was found that hotels in the U.S. are more volatile than traditional property types but do not contribute to the efficient frontier for the time period reviewed. By contrast, the empirical results for the U.K. indicated that hotels are much less risky than expected and contribute to the efficient frontier at lower-risk levels. This was confirmed by the de-smoothed results using the individual correlation coefficient for each property type. Consequently, it was concluded that hotels are an attractive real estate subsector offering credible diversification benefits. Furthermore, it is suggested that hotels are not necessarily deserving of their reputation as uniformly high-risk. This has important practical implications for institutional investors seeking to diversify their portfolios.

Notes

This study evolved from the author's MPhil dissertation in Real Estate Finance (submitted 2014) with the support of Arjun Singh, Professor in The School of Hospitality Business, Michigan State University; Gabriel Petersen, Managing Director in Real Estate, The Blackstone Group and Andrew Baum, Professor of Real Estate Investment in The Department of Land Economy, University of Cambridge.

¹ “In reconciling 2012 P&L data and our 2012 transaction data, we estimated overall value of hotel real estate in the U.S. to be approximately $350 billion” (Winkle, 2014).

² The survey was carried out with BLP in February 2014 and is based on data from over 400 online interviews. Respondents comprised hotel owner/operators, advisors, and investors.

³ Note that “no research on portfolio allocations to lodging real estate [had] been performed since Firstenberg et al. (1988)” (deRoos and Corgel, 1996, p. 33).

⁴ Otherwise known as mean-variance analysis (Fabozzi et al., 2002).

⁵ See Le Sourd (2007).

⁶t = 0 represents the lower bound of summation.

⁷ “Risk” and “volatility” (measured by standard deviation) are used interchangeably (Mangram, 2013).

⁸ MAR is typically set at the risk-free rate, target rate or zero. The author adopts zero in keeping with the literature (which assumes the goal is to avoid losses). For examples, see Lhabitant (2006) or Rollinger and Hoffman (2013).

⁹ Confirming the results of Wheaton and Rossoff (1998) that hotel demand moves closely in line with GDP.

¹⁰ See Black and Litterman (1992) for more on “corner solutions.”

¹¹ A review of available NCREIF data for the same period also confirms this result.

¹² Several modifications to the Sharpe ratio have been proposed in order to address the “negative excess return dilemma”; see, for example, the work of Israelsen (2003, 2005).

¹³ According to IPF (2007) this is the most robust de-smoothing technique.

¹⁴ Note that there is “no absolute way of determining whether a particular value of alpha is most appropriate for the de-smoothing process” (Byrne & Lee, 1995, para.42).

¹⁵ The necessary diagnostics were checked to ensure the relevant assumptions had been met. The only major issue found was multicollinearity, with particularly high VIFs for retail and industrial. To ensure robustness, retail was removed from the analysis. Consequently, all VIFs dropped below 10, which, for the purpose of this analysis, was deemed acceptable. The results were reaffirmed after a robust multiple regression analysis was run to correct for heteroskedasticity, ensuring that all assumptions had been satisfied.

Additional information

Notes on contributors

Nicholas Worsley

Nicholas Worsley is the 2014 recipient of the prestigious Mansford Award from the University of Cambridge for his MPhil dissertation on hotels in a mixed-real estate portfolio. He was also recently awarded the Best Graduate Student Research Article Award by the AHFME.