Hubert Dichtl

This study presents a systematic comparison of portfolio insurance strategies. We implement a bootstrap-based hypothesis test to assess statistical significance of the differences in a variety of downside-oriented risk and performance measures for pairs of portfolio insurance strategies. Our comparison of different strategies considers the following distinguishing characteristics: static versus dynamic protection; initial wealth versus cumulated wealth protection; model-based versus model-free protection; and strong floor compliance versus probabilistic floor compliance. Our results indicate that the classical portfolio insurance strategies synthetic put and constant proportion portfolio insurance (CPPI) provide superior downside protection compared to a simple stop-loss trading rule and also exhibit a higher risk-adjusted performance in many cases (dependent on the applied performance measure). Analyzing recently developed strategies, neither the TIPP strategy (as an ‘improved’ CPPI strategy) nor the dynamic VaR-strategy provides significant improvements over the more traditional portfolio insurance strategies.

Embracing the concept of factor investing, we design a flexible framework for building out different factor completion strategies for traditional multi-asset allocations. Our notion of factor completion comprises a maximally diversified reference portfolio anchored in a multi-asset multi-factor risk model that acknowledges market factors such as equity, duration, and commodity, as well as style factors such as carry, value, momentum, and quality. The specific nature of a given factor completion strategy varies with investor preferences and constraints. We tailor a select set of factor completion strategies that include factor-based tail hedging, constrained factor completion, and a fully diversified multi-asset multi-factor proposition. Our framework is able to organically exploit tactical asset allocation signals while not sacrificing the notion of maximum diversification. To illustrate, we additionally embed the common trend style that permeates many asset classes, and we also include the notion of style factor momentum.

The old and simple investment strategy “Sell in May and Go Away” (also referred to as the “Halloween effect”) enjoys an unbroken popularity. Recent studies suggest that the Halloween effect even strengthened rather than weakened since its first publication by Bouman and Jacobsen (2002). We implement regression models as well as Hansen’s (2005) “Superior Predictive Ability” test to analyze whether stock markets are really so inefficient. In line with the predictions of market efficiency, our results reject the hypothesis that a trading strategy based on the Halloween effect significantly outperforms.

We analyze the performance of a comprehensive set of equity premium forecasting strategies. All strategies were found to outperform the mean in previous academic publications. However, using a multiple testing framework to account for data snooping, our findings support Welch and Goyal (2008) in that almost all equity premium forecasts fail to beat the mean out-of-sample. Only few forecasting strategies that are based on Ferreira and Santa-Clara’s (2011) sum-of-the-parts approach generate robust and statistically significant economic gains relative to the historical mean even after controlling for data snooping and accounting for transaction costs.

The performance potential of forecasting-based tactical asset allocation strategies is difficult to assess. The fundamental law of active management suggests that the value added through active investment decisions depends on the forecasting quality and the number of independent forecasts. Although easy to use, the law depends on several specific assumptions that are not fulfilled in practice. Therefore, it is not clear ex ante whether the actual performance of tactical asset allocation is close to what the fundamental law predicts. Using a simulation approach, we quantify the entire distribution of information ratios, active returns and tracking errors under realistic conditions (for example, with transaction costs and tactical bounds rather than a simple mean-variance optimisation). Our results reveal that the fundamental law systematically underestimates the required forecasting quality to reach a very good information ratio. While all other assumptions of the law seem innocuous, transaction costs are responsible for most of the wedge between the law's prediction and the performance of tactical asset allocation in a realistic setup. Our results are robust for stock and bond market data from different countries.

Dollar-cost averaging requires investing equal amounts of an investment sum step-by-step in regular time intervals. Previous studies that assume expected utility investors were unable to explain the popularity of dollar-cost averaging. Statman [1995] argues that dollar-cost averaging is consistent with the positive framework of behavioral finance. We assume a prospect theory investor who implements a strategic asset allocation plan and has the choice to shift the portfolio immediately (comparable to a lump sum) or on a step-by-step basis (dollar-cost averaging). Our simulation results support Statman's [1995] notion that dollar-cost averaging may not be rational but a perfectly normal behavior.

Factor‐based allocation embraces the idea of factors, as opposed to asset classes, as the ultimate building blocks of investment portfolios. We examine whether there is a superior way of combining factors in a portfolio and provide a comparison of factor‐based allocation strategies within a multiple testing framework. Factor‐based allocation is profitable beyond exploiting genuine risk premia, even when applying multiple testing corrections. Investment portfolios can be efficiently diversified using factor‐based allocation strategies, as demonstrated by robust economic performance over various economic scenarios. The naïve equally weighted factor portfolio, albeit simple and cost‐efficient, cannot be outperformed by more sophisticated allocation strategies.

This paper uses a comprehensive set of predictor variables from the five largest Eurozone countries to compare the performance of simple univariate and machine learning-based multivariate models in forecasting stock market crashes. In terms of statistical predictive performance, a support vector machine-based crash prediction model outperforms a random classifier and is superior to the average univariate benchmark as well as a multivariate logistic regression model. Incorporating nonlinear and interactive effects is both imperative and foundation for the outperformance of support vector machines. Their ability to forecast stock market crashes out-of-sample translates into substantial value-added to active investors. From a policy perspective, the use of machine learning-based crash prediction models can help activate macroprudential tools in time.