Working Papers
Expected Return, Realized Return, and Machine Learning with Jens Kvaerner and Marc Stam
Machine learning models struggle to estimate expected returns due to stock returns’ low signal-to-noise ratio (SNR). Direct estimates of expected returns from the implied cost of capital or indirect estimates via option prices have higher SNR but are only available for a short period and a particular sample. We develop a methodology that circumvents the low SNR in realized stock returns and is applicable to all stocks. We uncover a stable relationship between characteristics and stock returns across all US stocks that hold out-of-sample and across geographical regions.
The Economic Value of Eliminating Diseases with Daniel Kárpati, Jens Kvaerner and luc Renneboog
We develop a framework to quantify the welfare gains of reducing health risks. The framework integrates causal effect estimates of health shocks on medical expenses, mortality, disability, labor market participation, earnings, and the need for nursing home care into a life cycle model. Economic benefits reflect both individuals' willingness-to-pay to reduce a particular health risk and net effects on government finances. We apply our framework to Dutch administrative data on medical diagnoses of 6.9 million people and 334 distinct medical diagnoses. Our estimates show that curing cancer or cardio-vascular diseases would result in economic benefits equivalent to 9.5% and 9.1% of the GDP. The corresponding estimates for preventive measures such as eradicating smoking or preventing overweight and obesity are 7.7% and 5.6%.
Evolutionary Arbitrage with Jens Kvaerner, Åvald Sommervoll, Dag Einar Sommervoll, Niek Stevens
The prices of exchange-traded funds (ETFs) can deviate significantly from their net asset values (NAVs). Exploiting such inefficiencies is often too costly because it involves taking positions in hundreds of underlying illiquid securities. We develop a method that identifies a liquid mimicking portfolio that tracks the NAV using only ETFs. Our method combines a genetic algorithm with non-negative least squares. We apply it to the fixed income ETF market. Our long-short strategy generates a Sharpe ratio of 4-5, incurs little transaction cost, and does well under all market conditions.