Are Returns of Intermediate Bond Funds Persistent?

by Michael Edesess, 2/11/14

In the search for skillful managers, the most valued characteristic is persistence – the ability of a manager to achieve superior returns consistently over time. Finding such managers is critical for fixed-income allocations, since the theoretical basis for indexing is weaker than it is for equities. Our study found, however, that persistence is elusive among a large sample of taxable bond funds.

A preliminary analysis of intermediate-term bond fund returns from 2004 to 2013 shows how difficult it can be to confidently assign causes to investment performance. From different quantitative analytical perspectives, the differences in intermediate-term bond fund performance could be due to persistent managerial skill, underrating of bonds by ratings agencies, cyclical changes in sentiments or data and statistical noise.

Advisor Perspectives obtained year-by-year investment return and standard deviation data from Morningstar Inc. for the years 2004-13 for 280 U.S. intermediate-term bond mutual funds. The aim was to investigate whether manager performance was persistent from year to year over this period.

From a preliminary study of the data, I was only able to come to one of the four tentative conclusions listed above. Further study would be required to choose from those conclusions, and even then it may be very difficult.

First look: apparent persistence of managerial performance

I initially limited the study to those funds for which there were 10 full years of data. This reduced the number of funds under consideration to 200. I subsequently reduced that number to those funds for which Morningstar could provide an average Standard & Poors quality rating for at least one of the 10 years. The reason for this will be explained shortly.

For those 200 funds for which there were 10 years of returns and standard deviations, I calculated annual Sharpe ratios by subtracting each year’s 3-month Treasury-bill return from each fund’s rate of return for that year and dividing by the corresponding year’s standard deviation, as measured by Morningstar.

A standard statistical test for a trend – a positive correlation of one year’s data with that of the next year – is to calculate the serial correlation coefficient, also known as the autocorrelation.

When the serial correlation for raw annual rates of return was calculated for each the 200 funds, it averaged approximately zero. However, the serial correlation for excess returns – in excess of the 3-month T-bill return – averaged an almost-significant 0.19. The serial correlations for the Sharpe ratio, moreover, averaged 0.28. This serial correlation was positive for more than 90% of the funds.

This suggests a persistent skill factor enabling outperformers to outperform year after year (in a statistical sense — that is, more often than not), while underperformers, more often than not, underperform year after year.

However, I was worried that there may be some additional factor causing consistently superior performance for some types of funds and consistently inferior performance for other types of funds. For example, it is well known that funds with higher risk will tend, more often than not, to outperform funds with lower risk over time. The Sharpe ratio is supposed to correct for that effect by dividing by a surrogate for risk: standard deviation.

It is also well known that standard deviation can be an imperfect measure of risk. Hence, I searched for another factor that may provide either a consistent boost or a consistent depressant to performance.