Investment style positioning of UK unit trusts

Abstract

We investigate the investment style positioning of UK equity unit trusts (mutual funds) over the 24-year period from 1987 to 2010 and assess if fund manager claims to follow a particular style strategy are evidenced in practice. Generally, UK unit trusts do not, in fact, consistently track declared styles but subject their funds to style switching or rotation. Nor do funds switch to become simple index trackers, as has widely been reported, but exhibit a mix of behaviour that we refer to as ‘market-momentum styling’. Our contribution is to offer a coherent, end-to-end picture of the evolution of investment styles over an economic cycle. In so doing we evidence that fund style positioning is subject to rotation and becomes subordinated to past portfolio performance or style momentum. Even this result is conditional as we go on to demonstrate that style investment is very likely to be driven by broader economic conditions, thereby creating market-momentum styling by default. This is arguably not a style at all and calls into question the intent behind fund ‘strategies’.

Keywords:

• unit trusts
• style investing
• momentum

JEL Classification:

• G11

Notes

1. It is theoretically inappropriate because it is logically inconsistent with superior information as it assumes homogenous expectations and, as a consequence, abnormal performance can only be observed due to the mean-variance inefficiency. CAPM is empirically unsuitable because of the varied and well-documented anomalies presented in its application. Additionally, the failing of CAPM in fund performance is that it demonstrates that the market portfolio is mean-variance efficient but cannot offer a prescription about which subsets of the economy are mean-variance efficient.

2. The method of calculation of abnormal returns requires the construction of a benchmark portfolio with near-identical style characteristics as that of the fund being examined. The method then proceeds in one of two directions. The first direction involves the calculation of returns of both actual and benchmarked portfolios over the sampling frequency period to produce a time series from which a regression abnormal performance may be detected. In this respect, the approach of calculating abnormal performance is identical to the factor method (Grinblatt and Titman 1989a). The factor loadings from the regression then determine portfolio weights from which abnormal returns are detected. The second direction matches each stock to a benchmark, based on style characteristics, as before. A return difference between the two is then calculated and, in the additional step that justifies the label ‘direct’, the actual portfolio weights are applied to the differences to form abnormal performance or benchmark-adjusted calculations of fund performance (Daniel et al. 1997).

3. This relates to the question of fund manager timing ability which can produce Jensen measures that are difficult to interpret. Consistent with nearly every researcher who has looked at this issue, we do not find timing ability of fund managers in our sample. In common with Grinblatt and Titman (1989a), we reject the impact of the sensitivity of the Jensen measure to timing issues as being empirically unimportant.

4. The only requirement in such instances is that the k-factors are locally mean-variance efficient (Grinblatt and Titman 1985).

5. We exclude unauthorised unit trusts due to the insufficient information to confirm their investment objectives. UK equity unit trusts have at least 80% of the fund invested in the UK equities. By restricting funds to those investing in UK equity, more accurate market benchmarks may be used (Cuthbertson, Nitzsche, and O'Sullivan 2008).

6. The FTSE 100 Index includes the largest 100 blue-chip companies, representing approximately 80% of the capitalisation of the UK market. The FTSE 250 Value Index comprises the mid-capitalised 250 value stocks in the market, while the FTSE 250 Growth Index contains the mid-capitalised 250 growth companies. The FTSE All-Small Index is a combination of the FTSE SmallCap Index and the FTSE Fledging Index. The FTSE SmallCap Index includes all the companies, representing the bottom 2% of the market capitalisation and the FTSE Fledging Index, which contains companies that are too small to be included in FTSE All-Share Index. Finally, the FTSE All-Share Index is an aggregation of the FTSE 100, the FTSE 250 and the FTSE All-Small Indices. Statistically, the index represents 98–99% of the UK market capitalisation. The FTSE 100 Value Index, the FTSE 100 Growth Index, the FTSE 250 Value Index, and the FTSE 250 Growth Index have been replaced by the FTSE Style Index in 2008.

7. See http://www.sec.gov/answers/mfclose.htm – refer to the 5th bullet point on this page.

8. Thus, profitable momentum might be ‘feasible’ in the face of transaction costs, we would argue, but the detail is impossible for any researcher to determine with accuracy for the following reason: no known measure of ‘soft dollar’ costs looks to be observable but which is a generally agreed component of transactions costs for most funds (see, e.g. Haslem 2006).

9. Employing the null that alpha is equal to zero in a 3F model is equivalent to saying that unsystematic risk captured by the error term is un-priced (Ferson and Harvey 1999). As we will show, this null is likely to be rejected in the 3F model but is likely to be accepted in the 4F model, thus validating our approach to employ momentum as a device explaining fund style rotation.

10. A total of 489 and 427 unit trusts have a significant momentum variable in Tables 2 and 3, respectively.

11. The basis of approach is adopted by major fund trackers, such as Morningstar and Lipper. Chan, Chen, and Lakonishok (2002) also analyse the style of mutual funds along similar dimensions.

12. We calculate but do not report style timing results using the Fama and French (1993) 3F model. Overall, we conclude that the trust managers do not possess the ability to time market, size or value factors. Our results are consistent with previous literature which is why we do not report them (see, e.g. Treynor and Mazuy 1966; Henriksson and Merton 1981; Chang and Lewellen 1984; Chan, Chen, and Lakonishok 2002 for US mutual funds; and Fletcher 1995; Byrne, Fletcher, and Ntozi-Obwale 2006; Cuthbertson, Nitzsche, and O'Sullivan 2008 for UK unit trusts).

13. Results from the Fama and French (1993) 3F model and value weighting are available on request. The results do not alter any of the conclusions to follow.

14. The t-statistic is the difference of the average between the past and future characteristic ranks. Its estimated standard error is the standard error of the difference between two ranks.

15. Based on a sample of the FTSE 350 stocks, Aarts and Lehnert (2005) investigate the profitability of style momentum strategies but find less profitable and more risky returns compared with regular momentum strategies, the results of which contrast with the evidence in the USA reported by Chen and De Bondt (2004).