MAN VS. MACHINE: QUANTITATIVE AND DISCRETIONARY EQUITY MANAGEMENT (the paper can be downloaded here)
I use a machine learning technique to classify the universe of active equity mutual funds into “quantitative”, who mostly rely of computer-driven models and fixed rules, or "discretionary", who mostly rely on human judgement. I propose an equilibrium model in which quantitative funds have greater information processing capacity but less adaptive strategies. The model predicts that quantitative funds hold more stocks and display pro-cyclical performance, but their trades are vulnerable to “overcrowding”. Discretionary funds alternate between stock picking in expansions and market timing in recessions, display counter-cyclical performance and focus on stocks for which less overall information is available. My empirical evidence supports these predictions.
WORK IN PROGRESS
MACHINE LEARNING AND THE CHANGING ECONOMICS OF KNOWLEDGE PRODUCTION (with Laura Veldkamp)
LEARNING FROM SOFT INFORMATION: CAPITAL ALLOCATION IN THE MUTUAL FUND INDUSTRY (with Anton Lines)
THE IMPACT OF MUTUAL FUND REGULATION ON ALLOCATIVE EFFICIENCY (with Anton Lines)
MAN VS. MACHINE:
QUANTITATIVE AND DISCRETIONARY EQUITY MANAGEMENT (paper download) (online appendix)
I use machine learning to categorize US active equity mutual funds as quantitative (reliant on computer models and fixed-rules) or discretionary (reliant on human judgment). I then formulate hypotheses of how their holdings and performance might differ, based on the conjecture that quantitative funds might have more learning capacity but less flexibility to adapt to changing market conditions than discretionary funds. Consistent with those hypotheses, I find that quantitative funds hold more stocks, specialize in stock picking, and engage in more overcrowded trades. Discretionary funds hold lesser-known stocks, switch between picking and timing and outperform quantitative funds in recessions.
THE CHANGING ECONOMICS OF KNOWLEDGE PRODUCTION
with Laura Veldkamp (paper download)
R&R at the Review of Financial Studies
Big data technologies change the way in which data and human labor combine to create knowledge. Is this a modest technological advance or a data revolution? Using hiring and wage data, we show how to estimate firms' data stocks and the shape of their knowledge production functions. Knowing how much production functions have changed informs us about the likely long-run changes in output, in factor shares, and in the distribution of income, due to the new, big data technologies. Using data from the investment management industry, our structural estimates predict that the labor share of income in knowledge work may fall by 5%. The change associated with big data technologies is similar in magnitude to estimates of the change brought on by the industrial revolution.
DO MUTUAL FUNDS KEEP THEIR PROMISES?
with Anton Lines (paper download)
This paper was previously circulated under the title: "Text-Based Mutual Fund Peer Groups?"
R&R Journal of Financial Economics
Mutual fund prospectuses contain a wealth of qualitative information about fund strategies, yet a systematic analysis of this content is missing from the literature. We use machine learning to group together funds with similar strategy descriptions, and ask whether they act in accordance with the text. Despite weak legal recourse for investors, we find that mutual funds largely do keep their promises. We document a market-based disciplinary mechanism: when funds diverge from their group's core strategy, investors withdraw capital. Funds respond to these punitive outflows by reducing their divergence from the peer group average at a faster rate.
LEARNING FROM PROSPECTUSES
with Andrea Buffa, Apoorva Javadekar and Anton Lines. (paper download)
We study qualitative information disclosure by mutual funds when investors learn from these disclosures in addition to past performance. We show theoretically that fund managers with specialized strategies optimally choose to disclose detailed strategy descriptions, while managers with standardized strategies provide generic descriptions. Generic descriptions lead to errors in benchmarking by investors and thus higher volatility in capital flows. While all fund managers dislike such volatility, those with above-average factor exposures also benefit from benchmarking errors as investors incorrectly ascribe factor returns to managerial skill. The model generates a number of predictions that we are able to test empirically using a
comprehensive dataset of fund prospectuses. Consistent with the model's predictions, funds with standardized strategies include more boilerplate in their descriptions, grow larger and have lower flow-performance sensitivity, despite having greater flow volatility.
MARKET TIMING IN BAYESIAN PORTFOLIO OPTIMIZATION
I propose a portfolio allocation model that combines a data-based approach with macroeconomic considerations of the business cycle. It accounts for the two key features of business cycles, namely co-movement among macroeconomic variables and asymmetric development of the cycles. The joint treatment of these characteristics improves the ability of the model to time market turns, consequently enhancing portfolio gains. The estimation technique developed allows to simultaneously address the issues of parameter uncertainty, mispricing uncertainty and the uncertainty relative to structural instability within a Bayesian portfolio optimization problem.
WORK IN PROGRESS