Diversification: Does it Really Matter in a Loan Portfolio? (Q3-10)
Sep 30th, 2010
The following was excerpted from Third Eye Capital Management Inc’s Q3 2010 Investor Letter.
One of the most frequent questions we get asked by prospective investors is whether our advised funds are sufficiently diversified in loan number, size, and industry/sector composition. Their concern is that the default of a proportionally large borrower (or a larger number of borrowers with correlated businesses) may lead to investment losses. The notion of diversification was first presented by Harry Markowitz in 1952, which finance textbooks call “Modern Portfolio Theory” or “MPT.” Markowitz measured the risk of securities in relation to their return and constructed portfolios consisting of securities that give the maximum amount of return with the lowest possible risk (based on historical data). The measure of risk was defined as volatility, which is the movement of the security’s value around the mean. Hence, if one can measure volatility, the correlations among the securities can also be calculated, making portfolio diversification manageable. The theory being that the more diversified a portfolio, the lower the total volatility and therefore the total risk of that portfolio.
MPT applied to a portfolio of private loans should have the same principles. The main objective for a lender diversifying a portfolio is minimizing exposure to any single borrower and reducing the risk of multiple borrowers defaulting in a specific industry or geographic region simultaneously. The risk of a sudden decline in an industry or the economy of a certain region cannot be ignored, as the recent financial crisis proved that shocks can arise without giving enough time for lenders to hedge or neutralize these positions. However, the quantifiability of correlations of borrowers or industries is more complex than for stocks, due to the lack of reliable and consistent data. Sure some lenders, such as the major Canadian banks, do use quantitative products such as CreditMetricsTM or Credit RiskTM, to make portfolio risk a function of allocating across various loan sectors using loan history data. However, the loan history data in these products contain survivorship bias, so represents above-average lending performance by lenders that have been active in a given sector for some time and successful enough to survive and be willing to share their history. A lender using such data to diversify its portfolio may in fact be understating total portfolio risk.
Most lenders, even those that make use of quantitative models, manage diversification intuitively by lending to businesses that have previously exhibited performance that is independent of business cycles or into industries that have little or no affect on each other. But demand for credit in a particular industry is not always larger then what is available to be loaned, and some lenders may be forced to give credit to companies that just happen to choose them. A fact that makes MPT inapplicable to lending is that it is impractical for a lender to exclude creditworthy companies based on correlations with existing borrowers. Of course borrowers that can meet our strict investment criteria are scarce (otherwise we would lend to any business), and our total loan outstandings would reduce considerably if we focused on diversification as a primary imperative. Investors would surely be upset if we failed to seize opportunities that met our objectives.
Usually, a lender evaluates each investment individually and focuses on the repayment potential of the borrower, or credit-risk, not the correlation of that investment with others in the loan portfolio. Loan portfolio diversification is often believed to naturally occur through growth in the number of loans, even though MPT shows that portfolio composition not size is the better contributor to lowering total risk.
Studies have found that contrary to what MPT says, there may be diseconomies in attempting to diversify a loan portfolio. The credit risk of a lender’s portfolio is in large part endogenous; for example, it is greatly influenced by the intensity and efficacy of a lender’s monitoring. Increased diversification increases the costs, and therefore the disincentives, of monitoring. A lender’s monitoring effectiveness is lower when the number of loans in the portfolio grows and in newly entered sectors where learning costs are present. Evidence presented by researchers at the NYU Stern School suggests that, in contrast to the recommendations of MPT and even regulatory dictums, diversification of loan assets is not guaranteed to produce superior return performance or greater safety for lenders.
We do not blame investors for advocating loan diversification. Diversification is easy to observe while monitoring ability and effort are not. Because most firms are unique, lenders must have a flexible credit evaluating process in order to capture the individuality of each loan. These differences are the size of the loan and the idiosyncratic company risks. This makes it difficult to apply diversification to a loan portfolio in the same way as a portfolio of bonds or stocks. We attempt to minimize credit specific downside risks by rigorous credit-evaluation processes and protect our portfolios from economic shocks through regular and persistent monitoring and by constantly calibrating loan availability to underlying collateral. Loan monitoring improves returns not by increasing best-case outcomes but by reducing the frequency and severity of worst-case outcomes; after all, a loan cannot earn more than its stated principal and interest, but effective monitoring can stop a troubled loan from deteriorating too far.
Building organizational knowledge for proper monitoring takes time, effort, and resources that a lot of lenders are reluctant to take. In fact, diversification (or the semblance of it) could just be the lazy lender’s excuse for improper or non-existent monitoring.