Diversification benefits from geographical dispersion in multifamily mortgage portfolios

RMA Journal, The, Sept, 2005 by Michaell Taylor, Randy Fuchs

Important advances in the estimation of quantitative commercial mortgage credit risk have led to a number of models, services, or tools that increase efficiency in asset risk estimation by exploiting the predictability of real estate returns. Portfolio analytics have not advanced as rapidly. Forward-looking portfolio analytics that measure diversification effects indicate that even modest geographic dispersion can reduce economic capital by about 40%.

One of the biggest challenges today in management of commercial real estate mortgage portfolios is the assessment of risk. Underdeveloped data and portfolio analytic models handicap portfolio managers in defining appropriate levels of economic capital, monitoring performance of mortgage portfolios, and executing strategies that can improve the risk profile of their book of loans.

Understanding and measuring the covariance of loans are fundamental to managing a well-diversified portfolio and efficiently capitalizing for low-probability events. Following Modern Portfolio Theory, where any two assets are less than perfectly and positively correlated, the portfolio manager can diminish the risk for a specific level of return or increase the expected return for a given level of risk. Only by measuring covariances can the portfolio manager be fully equipped to estimate or maximize the risk-adjusted return to the mortgage portfolio.

Institutions that fail to account for diversification in credit portfolios penalize themselves in establishing economic capital set-asides. In addition, improper quantification of portfolio risks may leave an institution with excessive exposure.

Historically based approaches to risk estimation are de rigueur in public securities analysis and are used frequently for commercial mortgages as well. Whether allocating capital across or within asset classes, historical volatility is used as a key risk measure. The standard approach is to measure volatility of past returns to derive a distribution. Future returns, in turn, are assumed to be drawn from this distribution. However, there are substantial differences between real estate and the other asset classes for which these models were initially developed. So analysts have developed models that exploit the predicability of real estate markets to create a forward-looking assessment of risk and asset scoring.

Although credit risk scoring methods for individual loans have improved, portfolio analytics for commercial mortgage portfolios have lagged behind. Sometimes, information from forward-leaning credit scoring models actually feeds into backward-looking portfolio analytic modules. There are consequences to using a forward loan-level model with a backward portfolio model. At best, it's inefficient and, at worst, it creates a biased view of portfolio diversification because of the variance between forward risk-and-return estimates and their backward-looking counterparts. The variance, though, is not constant from one metro to another.

Let's see what happens when we use a forward-looking, simulation-based, loan-level risk analysis approach to the analysis of portfolios of multifamily loans. (See Appendix A for a description of the methods employed.) After simulating portfolios of various geographic dispersion levels and concentrations to find the portfolio diversification effects of geography, (1) we isolate geographic dispersion from other factors, since many lenders often face strategic business decisions concerning their geographical reach and exposures, while other lenders specialize in originating loans of a single property type.

Geographic Diversification Benefits

To assess the impact of geographical dispersion on risk, we generate a random portfolio for a given number of apartment loan assets. All assets are granted equal weight and all have similar loan characteristics--that is, DSCR of 1.25 and LTV of 80%, 10-year maturity, and extended lockout provisions. For the sake of illustration, we begin with an assessment of the diversification of a 10-asset portfolio at the national level. Ten loans from 10 separate metros are combined into a portfolio drawn randomly from geographies across the country.

Though informative, this portfolio's degree of diversification is subject to the randomness of the metros selected for inclusion. Therefore, this process of random portfolio construction has been repeated 300 times with each portfolio consisting of loans from randomly selected geographies. In each case, the diversification ratio (DR) is measured. We define DR here as the ratio of unexpected loss for a diversified portfolio over the unexpected losses for the same portfolio given no credit for diversification--that is, ignoring potential benefits flowing from diversification. At the end of the process, we have a distribution of DRs for 10-asset portfolios complete with means, medians, and standard distributions of the range and expected dispersion of diversification effects. The distribution reflects the range of DRs for portfolios of generally healthy apartment loans distributed equally among 10 metros.