1. An empirical evaluation of fixed income fund performance : new evidence across alternative methods
- Author
-
Littlejohn, Elizabeth Grace, Fletcher, Jonathan, and Marshall, Andrew P.
- Abstract
With a wealth of research directed at fund performance evaluation, that which is specific to bond funds is relatively minimal. This thesis aims to shed further light on this topic using a new sample of UK data. To most effectively judge the merits of active management the specification of a reliable benchmark model is paramount. The first empirical chapter involves testing a range of single and multi-factor models to determine which of these can be considered the most suitable for evaluation of the funds in question. The Gibbons, Ross, and Shanken (1989) test of mean-variance efficiency is first applied as an absolute test, with further alpha test statistics used for a relative evaluation of the candidate models. Overall, the results show that a five-factor model (named Maturity 5) that includes adjustment for both term and credit is the most reliable. This is primarily indicated by the lowest absolute alphas and corresponding goodness of fit statistics, i.e., low standard errors and high (adj)R2. A multi-period analysis is conducted to determine the extent to which this varies over time. The results are consistent, with the same model performing the best in each case. The second empirical chapter builds upon these findings and uses Maturity 5 to evaluate the sample of UK bond funds. The results over the whole period from January 1999 until July 2016 are consistent with much of the academic literature; the funds are found to underperform on a risk-adjusted basis by approximately a magnitude of costs. However, during the recent subsample from September 2009 the performance is neutral. The use of dummy variables and Wald tests indicates that the Government, Corporate, and Diversified funds perform significantly better here. Having identified during testing that minimal bias is likely to be induced by the Maturity 5 model, it can be inferred that the performance is not just a result of exposure to passive portfolios (as represented by the test assets). Instead, there appears to be some ability beyond this being employed by the active funds. The measure of Treynor and Mazuy (1966) has been used to identify if market timing makes a positive contribution to performance. Evidence is minimal when the Barclays Sterling Aggregate Index is used as a proxy, however, 25% of Corporate bond funds exhibit positive ability relative to the category-specific benchmark as assigned by Morningstar. The third and final empirical chapter seeks to shed some light as to whether the active managers are lucky with respect to alpha generation or indeed exhibit true outperformance. A bootstrap procedure is first applied to the individual funds to do so. The method used here is known as entire-cases resampling (Fama and French 2010), whereby the time-ordering is maintained across the sample. This differs from the approach of KTWW (2006), which is used prevalently throughout the fund performance literature. To date, the entire-cases method has not yet been applied to a sample of bond funds. The initial results in this chapter support superior performance in the low-rate environment; this being evident across all funds from the 97h percentile, and Corporate from the 95th. To add further robustness, two variations of the false discovery rate method have been used to adjust for luck. The "classical" approach of BSW (2010) also finds positive alpha, isolated to the post-crisis period. Lower expenses characterise these funds, along with a higher number of observations per fund, and the average alpha is approx. 1.68% p.a. Recent literature has proposed many refinements to the methods used to address multiple hypothesis testing issues. The Ferson and Chen (2019) approach expands upon the entire cases method already used in this chapter, allowing for not only a lucky distribution to be considered, but also simulates those defined as both "good" and "bad", incorporating power and confusion parameters. The results here are the most positive yet. At 10% 10% level, the proportion of outperforming funds is 11% and 33% across the whole and recent periods respectively. This is again driven primarily by the Corporate category of funds.
- Published
- 2021
- Full Text
- View/download PDF