Dirk Nitzsche

False Discoveries in UK Mutual Fund Performance

We use a multiple hypothesis testing framework to estimate the false discovery rate (FDR) amongst UK equity mutual funds. Using all funds, we find a relatively high FDR for the best funds of 32.8% (at a 5% significance level), which implies that only around 3.7% of all funds truly outperform their benchmarks. For the worst funds the FDR is relatively small at 7.6% which results in 22% of funds which truly underperform their benchmarks. For different investment styles, this pattern of very few genuine winner funds is repeated for all companies, small companies and equity income funds. Forming portfolios of funds recursively for which the FDR is controlled at a ‘acceptable’ value, produces no performance persistence for positive alpha funds and weak evidence of persistence for negative alpha funds.

Explaining movements in UK stock prices

Abstract We examine movements in aggregate UK stock prices by decomposing the variance of unexpected real stock returns into components due to revisions in expectations of future dividends, discount rates, and the covariance between the two. The contribution of news about future discount rates is about four times that of news about future dividends, with no significant covariance between them. Our analysis of excess returns uncovers a positive covariance between news about dividends and news about real interest rates. Since these two elements have opposite effects on current stock prices, their combined effect is negligible. Persistence in expected returns, as well as predictability, are found to be important in explaining stock price movements.

Interest Rates in Germany and the UK: Cointegration and Error Correction Models

The authors test the expectations hypothesis (EH) of the term structure using U.K. and German weekly data on short dated instruments with maturities up to one year. For both data sets comprising k interest rates the authors find that the rank of the cointegrating space is (k - 1); but they can only accept that the cointegrating parameter estimates are of the form (-1, 1, 0,...) etc. when considering bilateral combinations of interest rates. When the authors test the joint null that the set of (k - 1) spreads forms a basis for the cointegration space, this is rejected. However, the point estimates of the cointegration parameters are close to unity and there is no diminution in outside sample forecasting performance of the ECM equations when the spread restrictions are imposed. On balance, one might conclude that the EH is not grossly at variance with the data. Copyright 1998 by Blackwell Publishers Ltd and The Victoria University of Manchester

Long rates, risk premia and the over-reaction hypothesis

Abstract We examine the possibility that the apparent failure of the expectations hypothesis EH on UK data at the long end of the maturity spectrum, may be due to the presence of a time varying (yet stationary) term premium. The presence of a time varying term premium (TVP) can result in the ‘over reaction hypothesis’ namely, that actual movements in the spread appear to be more volatile than expected changes in future short rates. We find, using UK data, that for very long maturities there is evidence of a time varying term premium which influences one-period excess holding period returns. The effect of this term premium on the coefficient of the spread in an equation which predicts the expected one-period change in long rates is sufficient to bias the results towards rejection of the EH. However, the impact of the average term premium on a weighted average of future short rates is relatively small, so that the spread is an unbiased predictor of future changes in short rates. Our analysis is conducted using a high quality UK data set for maturities of 2, 3, … 10 years and for 15, 20 and 25 years.

Market Segmentation and Stock Price Behaviour

We employ Campbell and Shiller’s (1989) VAR methodology to examine the relative performance of the CAPM and the consumption-CAPM. We find that although neither provides a complete description of stock price behaviour, the former clearly dominates the latter. We then consider the implications for sub-sectors of the market. According to the CAPM, sub-sector returns depend on the covariance of sub-sector returns with market returns. However, if analysts are more skilled at eliminating mis-pricing in sub-sectors of the market than in the market as a whole, the required return on a sectoral portfolio may depend only on the expected return variance within that sub-sector. Using quarterly UK data for five industry-based portfolios, we find little support for the covariance model at the sectoral level, whereas the own-variance model fares better. It seems therefore that moving from return variances to covariances produces little if any improvement in the performance of the CAPM.

Diversification Returns, Rebalancing Returns and Volatility Pumping

There is now a substantial literature on the effects of rebalancing on portfolio performance. It is widely argued in the theoretical literature that rebalanced strategies are inherently likely to generate greater terminal wealth than unrebalanced strategies, although empirical studies do not generally support this claim. We show that this claim is based on a misattribution between ‘rebalancing returns’ which are specific to the act of rebalancing, and ‘diversification returns’ which can be earned by both rebalanced and unrebalanced strategies. Confusion appears to have increased because in some situations these two distinct effects have the same magnitude. This issue has important implications for return attribution in diversified portfolios. Misleading claims about the benefits of rebalancing are likely to lead investors into strategies which involve insufficient diversification and excessive transactions costs.

Winners and Losers: German Equity Mutual Funds

We investigate the performance of winners and losers for German equity mutual funds (1990–2009), using empirical order statistics. When using gross returns and the Fama–French three-factor model, the number of statistically significant positive alpha funds is zero but increases markedly when market timing variables are added. However, when using a ‘total performance’ measure (which incorporates both alpha and the contribution of market timing), the number of statistically significant winner funds falls to zero. The latter is consistent with the bias in estimated alphas in the presence of market timing. We also find that many poorly performing funds are unskilled rather than unlucky.

Market and Style Timing: German Equity and Bond Funds

We apply parametric and non‐parametric estimates to test market and style timing ability of individual German equity and bond mutual funds using a sample of over 500 equity and 350 bond funds, over the period 1990–2009. For equity funds, both approaches indicate no successful market timers in the 1990–1999 or 2000–2009 periods, but in 2000–2009 the non‐parametric approach gives fewer unsuccessful market timers than the parametric approach. There is evidence of successful style timing using the parametric approach, and unsuccessful style timing, particularly in the 2000–2009 period. There is evidence of positive and negative bond timing in the 2000–2009 period.

Are Investors Better Off with Small Hedge Funds in Times of Crisis

With the benefit of a more comprehensive dataset than previous authors in this area, in this paper we revisit the relationship between hedge fund performance and size. Our results indicate that there is a strong, negative relationship between hedge fund performance and size. But, in addition, we also find that rather than dissipating during the two recent periods of financial crisis, other things equal, investors would have been better off with smaller hedge funds than with large ones during these crisis periods. Finally, we also document clear cross-sectional variation in this relationship by broad hedge fund strategy.