Conditioning information in mutual fund performance evaluation: Portuguese evidence

Abstract

We estimate and compare the performance of Portuguese-based mutual funds that invest in the domestic market and in the European market using unconditional and conditional models of performance evaluation. Besides applying both partial and full conditional models, we use European information variables, instead of the most common local ones, and consider stochastically detrended conditional variables in order to avoid spurious regressions. The results suggest that mutual fund managers are not able to outperform the market, presenting negative or neutral performance. The incorporation of conditioning information in performance evaluation models is supported by our findings, as it improves the explanatory power of the models and there is evidence of both time-varying betas and alphas related to the public information variables. It is also shown that the number of lags to be used in the stochastic detrending procedure is a critical choice, as it will impact the significance of the conditioning information. In addition, we observe a distance effect, since managers who invest locally seem to outperform those who invest in the European market. However, after controlling for public information, this effect is slightly reduced. Furthermore, the results suggest that survivorship bias has a small impact on performance estimates.

Keywords:

• mutual funds
• conditional performance evaluation
• spurious regressions
• survivorship bias
• distance effect

JEL Codes :

• G11
• G12

Notes

Although a longer evaluation period would be desirable, it would substantially reduce the number of funds. On the one hand, the Portuguese mutual fund industry is a very recent one when compared with the major markets (US and UK). On the other, the availability of data regarding the public information variables represented an additional constraint on the length of the sample period.

The average size of these funds is presented in Appendix 2.

In December 2004, the five largest equity fund management companies represented, in terms of net asset values, around 91% of the market.

The composition of this sample was checked with the Portuguese Unit Trust Association (APFIPP) and the Portuguese Securities Market Commission (CMVM).

Although not reported in the paper, the average monthly returns for individual funds are negative for all surviving funds and for 15 of the non-surviving funds. Using the Jarque–Bera statistic, only one fund (ESAE) rejects the null hypothesis of a normal distribution of returns (at a 5% significance level).

Appendix 4 presents some summary statistics for market returns (both negative, on average, and normally distributed) and for the risk-free rate, over the sample period.

Even though it would be preferable to use a German government bond rate as the short-term rate, we could not obtain that data due to the non-existence of a liquid Treasury bill market.

Using lags of up to six periods and a 5% MacKinnon critical value, the hypothesis of a unit root is rejected for all variables.

This result would have been different without the stochastic detrending of the information variables, since we found that without this procedure the sign of the coefficient would be positive, as expected.

This effect is not merely a consequence of the indexes used as benchmarks. In fact, additional tests showed that even if we had used the same benchmark for the two fund categories, National funds would still perform better than European Union funds.

Similar evidence is obtained by comparing the mean raw returns of surviving and non-surviving funds.