Abstract: We show that machine learning methods, in particular extreme trees and neural networks (NNs), provide strong statistical evidence in favour of bond return predictability. NN forecasts based on macroeconomic and yield information translate into economic gains that are larger than those obtained using yields alone. Interestingly, the nature of unspanned factors changes along the yield curve: stock and labor market related variables are more relevant for short-term maturities, whereas output and income variables matter more for longer maturities. Finally, NN forecasts correlate with proxies for time-varying risk aversion and uncertainty, lending support to models featuring both of these channels.
Abstract: I empirically show that changes in investor holdings exhibit a low-dimensional factor structure that is economically interpretable. Using an extended version of Instrumented Principal Components Analysis (IPCA), I model changes in a large sample of sector-level investor holdings and recover latent factors and sector-specific loadings on the factors. I find that the recovered factors reflect the state of the macroeconomy and financial constraints of investors. Investor loadings on the factors reveal partially pro-cyclical trading behavior of the banking sector and of mutual funds, while hedge funds and pension funds act partially counter-cyclically. In addition, I document that the set of characteristics relevant for explaining changes in holdings is likely wider than implied by common risk factor models. Finally, using the decomposition of holdings changes implied by IPCA, I demonstrate asset pricing effects consistent with institutional price pressures from banks and mutual funds, as well as market-timing ability of investment advisors that is unrelated to common asset characteristics.
Abstract: Due to their short lifespans and migrating moneyness, options are notoriously difficult to study with the factor models commonly used to analyze the risk-return tradeoff in other asset classes. In- strumented principal components analysis (IPCA) solves this problem by tracking contracts in terms of their pricing-relevant characteristics. We recover the latent common risk factors in option returns and the time-varying loadings of individual options on these factors. Five latent factors explain more than 90% of the variation in a panel of monthly S&P 500 option returns from 1996 to 2017. The factors we estimate are interpretable as jump, volatility, and term structure spread risks.
"What matters when? Time-varying sparsity in expected returns" joint with Daniele Bianchi (Queen Mary, Univ. of London) and Andrea Tamoni (Rutgers)
Abstract: We provide a measure of sparsity for expected returns within the context of classical factor models. Our measure is inversely related to the percentage of active predictors. Empirically, sparsity varies over time and displays an apparent countercyclical behavior. Proxies for financial conditions and for liquidity supply are key determinants of the variability in sparsity. Deteriorating financial conditions and illiquid times are associated with an increase in the number of characteristics that are useful to predict anomaly returns (i.e., the forecasting model becomes more dense). Looking at specific categories of characteristics, we find that only trading frictions is robustly present throughout the sample. A substantial amount of the time-variation in sparsity is attributable to the value, profitability, and investment categories. A strategy that exploits the dynamics of sparsity to time factors delivers substantial economic gain out-of-sample relative to both a random walk and a model based on preselected, well-know characeristics like size, momentum and book-to-market.
"Predictability of Order Imbalance, Market Quality and Equity Cost of Capital" joint with Daniele Bianchi (Queen Mary, Univ. of London) and Roman Kozhan (Warwick)
Abstract: We study the effect of the predictability of order imbalance on market quality. We measure the degree of predictability by using the predictive likelihood from a dynamic linear model where the dependent variable is the day-ahead order imbalance. Empirically, we show that increasing order imbalance predictability corresponds to significantly higher market liquidity and efficiency. This positive relationship is economically significant: a long-short portfolio based on past predictability generates significant risk-adjusted returns. Predictability of order imbalance measures a cost of asymmetric information that is not captured by traditional measures of adverse selection. The risk factor that is associated with asymmetric information is priced in the cross-section of stock returns, controlling for a variety of conventional sources of systematic risk. These results suggest the existence of a tight link between market microstructure features affecting order imbalance predictability and both market quality and the cost of capital of firms.