Asset Management

The maximum drawdown (MDD), the maximum peak-to-trough loss associated with a series of returns, is a simple but highly important measure for investors with a downside risk budget. This paper compares the performance of three bootstrap simulation methods to estimate the entire distribution of MDDs from various global stock-bond allocations, quantifying the economic costs of biased estimates for three realistic decision-making scenarios. Compared to its benchmarks, the stationary bootstrap of Politis and Romano (1994) leads to the most precise estimates for the MDD which, in turn, helps avoid costly investment errors in portfolio construction and dynamic risk control strategies.

In this paper, we analyze the relationship between speculators in commodity futures markets and generic time series momentum (TSMOM) traders as well as the impact of this relationship on the subsequent TSMOM strategy performance. We find strong empirical evidence across all commodity markets that speculators in commodity markets tend to trade a TSMOM strategy, which confirms the results found by Boos and Grob (Journal of Financial Markets 64, 100774). On the basis of this result, we also ascertain whether the degree of such alignment has an impact on the performance of the TSMOM strategy. We find that there is weak, but statistically significant and robust evidence to suggest that the higher the degree of alignment between speculators and a generic TSMOM strategy, the lower the realized performance of trading TSMOM in these markets. Albeit we find little evidence that this can be exploited in a dynamic investment strategy, this negative relationship suggests that if a Commodity Trading Advisor (CTA) trades commodity futures markets which are less commonly traded by other CTAs, these markets may not only increase the internal diversification of their fund but these markets may also have a higher TSMOM Sharpe ratio by themselves. Consequently, our analysis provides valuable insights into improving the portfolio construction of CTAs.

Machine learning-based market beta estimators outperform established benchmark models both statistically and economically. Analyzing the predictability of time-varying market betas of U.S. stocks, we document that machine learning-based estimators produce the lowest forecast and hedging errors. They also help create better market-neutral anomaly strategies and minimum variance portfolios. Among the various techniques, random foests perform best overall. Model complexity is highly time-varying. Historical stock market betas, turnover, and size are the most important predictors. Compared to linear regressions, allowing for nonlinearity and interactions significantly improves predictive performance.

This paper tests the predictive performance of machine learning methods in estimating the illiquidity of U.S. corporate bonds. Machine learning techniques outperform the historical illiquidity-based approach, the most commonly applied benchmark in practice, from both a statistical and an economic perspective. Tree-based models and neural networks outperform linear regressions, which incorporate the same set of covariates. Gradient boosted regression trees perform particularly well. Historical illiquidity is the most important single predictor variable, but several fundametal and return- as well as risk-based covariates also possess predictive power. Capturing interactions and nonlinear effects among these predictors further enhances predictive performance.

We study the role of social trust in the equity allocation decisions of global investors using a large sample of institutionally managed portfolios of 8,088 investors from 33 countries over the 2000 through 2017 period. The negative relationship between social trust and foreign bias suggests that institutional investors from high-social trust countries are less prone to underinvesting in foreign equity. Our results provide credence to an information-based explanation, indicating that social trust reduces foreign bias by compensating for the lack of information about foreign stock markets. Moreover, the effect of social trust on foreign bias is stronger if host-country institutions are weak, while it vanishes when the host country is characterized by strong institutions. The informal institution of social trust compensates for the lack of well-functioning formal country-level institutions in international portfolio decisions. Finally, the allocation effect resulting from social trust is different from “blind” trust. The portfolios of high-trust investors exhibit higher cross-country diversification and an enhanced portfolio risk-return trade-off.

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.

This paper evaluates the predictive performance of machine learning methods in forecasting European stock returns. Compared to a linear benchmark model, interactions and nonlinear effects help improve the predictive performance. But machine learning models must be adequately trained and tuned to overcome the high dimensionality problem and to avoid overfitting. Across all machine learning methods, the most important predictors are based on price trends and fundamental signals from valuation ratios. However, the models exhibit substantial variation in statistical predictive performance that translate into pronounced differences in economic profitability. The return and risk measures of long-only trading strategies indicate that machine learning models produce sizeable gains relative to our benchmark. Neural networks perform best, also after accounting for transaction costs. A classification-based portfolio formation, utilizing a support vector machine that avoids estimating stock-level expected returns, performs even better than the neural network architecture.

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.