The following article is reprinted from the Quarter 3, 2006 issue of On the Edge, the Interactive Data Fixed Income Analytics bimonthly newsletter.

Performance Attribution for Municipal Portfolios

Teri Geske
Senior Vice President, Product Development

---

Performance attribution is becoming increasingly important to managers of municipal bond portfolios. There are a number of challenges to attribution in the municipal arena: municipal securities are categorized by multiple characteristics (such as State, Purpose, Insured status, etc.), benchmark indices typically contain tens of thousands of securities, portfolios may include both taxable and tax-exempt securities, etc. The Returns-based Performance Attribution system in BondEdge offers an intuitive, user-friendly approach to municipal portfolio attribution that allows managers to demonstrate the impact of portfolio allocation decisions on relative returns, according to the dimensions that mirror the investment management process. This article uses a hypothetical portfolio to illustrate the approach.

In this analysis, the total return of a portfolio and benchmark are segregated among user-selected categories, such as State, Primary Purpose (G.O., Revenue, Pre-refunded, Insured or ETM), Maturity or Duration buckets, etc., and performance differences are then attributed to the portfolio manager's decision to under- or overweight those categories, and to the manager's selection of under- or outperforming securities relative to the benchmark within each category. The returns for the portfolio and the benchmark are segregated bond-by-bond into the chosen categories; therefore, the analysis requires the use of constituent-level information for the benchmark as well as the portfolio. This analysis is particularly useful for municipal portfolio managers serving high net worth individuals with balanced accounts, as it matches the type of attribution typically done for equity portfolios. It also circumvents certain challenges encountered when applying a Factors-based style of performance attribution to the municipal market, such as “what is the benchmark yield curve?” and “what sector definitions are used to determine sector spread movements?”

The following example examines a hypothetical municipal portfolio benchmarked against the Merrill Lynch Muni 1-10 Year Index. It is a multi-state portfolio, and the manager's primary decision is the allocation between G.O., Revenue and Pre-refunded bonds. Within those broad categories, the manager determines the desired term structure exposure using duration ranges as shown below. The analysis shows that April was a cruel month for this portfolio manager, as his (or her) portfolio underperformed the benchmark by almost 30 bps. The question is now, was this due to over- or under-weightings among the various types of munis, or was it due to the manager's unfortunate bond selections within the categories?

At the top of the report, under the columns labeled “Analytics: Weighting and Selection”, we can see that the manager's underperformance was due almost entirely to poor selection – in other words, the manager did not under-weight an outperforming primary category, or overweight a primary category that underperformed the overall index in any meaningful way. For example, the manager allocated 51.78% of the portfolio to Revenue bonds, compared to the index's allocation of 50.67%. The return for Revenue bonds in the index was 0.151, compared to the overall index return of 0.148, so Revenue bonds slightly outperformed (by less than 2 bps) and the manager slightly over-weighted the category. This favorable weighting decision, however, was entirely overwhelmed by the poor performance of the bonds in the Revenue category, particularly in the 5.0 – 9.9 year duration bucket (these duration buckets are user-defined in the Attribution system). The portfolio's holdings in this category had a weighted average return of almost −18 bps, while the average return for the same category in the index was +6 bps, and, the manager chose to invest almost half of the portfolio's Revenue bonds in this duration bucket.

At this secondary level of analysis, (the duration allocation), the Weighting Effect is based on whether the portfolio was over- or under-weighted in the secondary category (the particular duration bucket) and the index's return for the secondary category compared to the return for the index within the primary category. So, the portfolio was substantially over-weighted in the 5.0 – 9.9 duration range (42.8% vs. 19.8% for the index) and the index return for bonds in the 5.0 – 9.9 was only 6 bps, versus 15.1 bps for Revenue bonds overall. Thus, the Weighting Effect for this category is negative. The Selection Effect is then determined by whether the portfolio's return within a duration bucket is greater or less than the benchmark's return within the same bucket.

Interpreting the results with a second layer in the analysis requires us to recognize that when attributing the return difference within the primary category (e.g., primary type or purpose) to the secondary categories below it (e.g., duration buckets), the decision to be in the primary category had already been made. Given that decision, the Selection Effect at the primary category level is both a function of the manager's ability to select among the categories at the second level (i.e., the weightings across the duration buckets), and his ability to select good or bad bonds within each of those categories. Since the analysis at the secondary level is essentially a further breakdown of the Selection Effect for the primary level, the values shown in the combined Weighting and Selection Effect columns at the secondary level sum to the total Selection Effect for the primary category.

The Returns-based Attribution system allows you to drill down into any cell to see the bonds within the selected category, for the portfolio or for the benchmark. So, we can easily see the individual returns that comprised this category. The following is a subset of the bonds, those with a total return worse than 0.50% for the month:

BondEdge also offers a “Factors-based” approach to attribution that quantifies the sources of return and return differences versus a benchmark according to the key factors of income, term structure movements, spread changes, currency moves and issuer-specific performance. Both the Factors-based and the Returns-based attribution analyses can offer important insights into the sources of return differences. Most core and core-plus managers rely heavily on the Factors-based approach for internal assessment of portfolio structuring decisions because it quantifies returns in terms that mirror the portfolio management process, but may also use the Returns-based approach in client reports and for presentations to the consulting community. The Returns-based attribution system is particularly well-suited to the needs of municipal bond portfolio managers. It offers flexibility in presenting performance according to the criteria by which the portfolio is managed, portfolios can by examined according to multiple dimensions, such as allocations by State, then by Bond Type (purpose) within each State, or vice versa, and the results are intuitive and easily explainable to clients and other stakeholders.

If you would like more information about subscribing to the Returns-based Performance Attribution module in BondEdge, please contact your BondEdge Representative.