Real estate investment by Bank Holding Companies and their risk and return: nonparametric and GARCH procedures

Abstract

We investigate the association between real estate investment by US Bank Holding Companies (BHCs) and their return, risk and risk-adjusted returns. Three portfolios are formed of BHCs according to whether they do or do not invest in real estate, strictness of the regulation on real estate investment and the ratio of real estate investment to assets. Wilcoxon tests of differences in portfolio returns, risk, risk-adjusted returns and value at risk between each pair of portfolios are conducted to determine how engagement in real estate, stricter regulation and increased real estate investment affect BHC performance. These effects are also investigated within a GARCH framework. Wilcoxon tests indicate that real estate investment or operating under lenient rules lower return and risk-adjusted returns and raise risk. Within GARCH, increases in real estate investment are associated with lower returns and greater systematic risk for BHCs with higher real estate shares in assets. These results indicate that benefits from real estate investment by banks are outweighed by greater variability of real estate prices and BHCs’ lack of expertise in the field. BHCs in the sample invested no more than 4.54% of their assets in real estate, leaving open the possibility that a higher threshold exists, beyond which performance improvements would be manifested.

Keywords:

• Bank Holding Companies
• product diversification
• real estate
• risk

JEL Classification:

• G21
• G28

Notes

¹ The Dodd–Frank Wall Street Reform and Consumer Protection Act of 2010 requires the Federal Reserve to restrict proprietary trading of securities, commodities and derivatives by BHCs and forbids BHCs from investing in hedge funds and private equity firms. It does not, however, reimpose restrictions on interstate branching or ownership of investment banks and insurance companies that were relaxed in 1999. See H.R. 4173–111th Congress: Dodd–Frank Wall Street Reform and Consumer Protection Act of 2010. Available at: http://frwebgate.access.gpo.gov/cgi-bin/getdoc.cgi?dbname=111_cong_bills&docid=f:h4173enr.txt.pdf

² Recent World Bank surveys of national differences in bank regulations, including restrictions on real estate activity, have been led by James Barth, Gerard Caprio, and Ross Levine (http://econ.worldbank.org).

³ See the following links for information on the cited laws and regulations: US Code Title 12 Section 29 (National Bank Act section on real estate ownership by national banks): vlex.com/vid/sec-power-hold-real-property-19225851. Federal Reserve Regulations H and Y: www.federalreserve.gov/bankinforeg/reglisting.htm.

⁴ González (2005) and Ramirez (2002) find that bank product diversification raises charter value.

⁵ Thrift service corporations are subsidiaries of thrift holding companies that are authorized to engage in any activity regulators deem reasonably related to the business of thrifts (Williams, 1988).

⁶ The authors chose the dates to reduce the effects on returns of the high initial costs of investing in nontraditional assets that were incurred in the years following the passage of these two acts in 1980 and 1982, respectively.

⁷ The Z-score measures the probability that a BHC will fail or alternatively the number of SDs below mean that ROA must fall in order to bankrupt the firm. See Lown et al. (2000) for more details.

⁸ This index includes all of the stocks traded on the New York Stock Exchange (NYSE), the American Stock Exchange (AMEX) and the NASDAQ Stock Market (NASDAQ). Each stock in the portfolio is given equal weight, and dividends are included in the returns. The quarterly return is calculated from CRSP monthly return data.

⁹ To this end, we regress the interaction term on a constant, MARKET, and RE and use the residuals in the Model. For a more detailed explanation of orthogonalized interaction terms, see Aiken and West (1991).

¹⁰ We also tried estimating other models from the GARCH family of models, including EGARCH and GARCH-in-Mean (GARCH-M), but the relatively limited number of observations turned out to be an obstacle against achievement of parameter convergence. We used a variety of software towards this end, including RATS, SAS and Stata without success. However, the Ljung–Box test statistics of the residuals and squared residuals for the models we fitted are found to be statistically insignificant, indicating that the models fit the data well (Table 6).

¹¹ An alternative distribution, the t-distribution, allows for fat tails too but has a peak rather than a cusp at the origin.

¹² We also estimated additional models (1) to include interest rate and exchange rate as additional regressors, (2) to employ alternative volatility specifications and (3) to distinguish the stock market behavior during the recent crisis by introducing dummy variables. The algorithms for calculating the MLE of these models frequently failed to converge, possibly because the number of model parameters is high relative to the number of observations (82). We also included the portfolio mean BHC size as a regressor in the volatility equation because volatility may be sensitive to the BHC size, but we then omitted it since it was insignificant. Results are available on request.

¹³ Chi-square tests of the restriction θ₁ + θ₂ = 1 were performed for each model estimate for each portfolio. In each case, the restriction was rejected at the 1% level of significance.