Research Seminars Series

The Streetlight Effect in Data-Driven Exploration

We consider settings such as innovation-oriented R&D where agents must explore across different projects with varying but uncertain payoffs. How does providing partial data on project payoffs affect individual performance and social welfare? While data can typically reduce uncertainty and improve welfare, we present a simple theoretical framework where data provision can decrease group and individual payoffs. In particular, we predict that when data shines a light on sufficiently attractive (but not optimal) projects, it can crowd-out exploration activity, lowering individual and group payoffs as compared to the case where no data is provided. We test our theory in an online lab experiment where we show that data provision on the true value of one project can hurt individual payoffs by 12% and reduce the group's likelihood of discovering the optimal outcome by 48%. Our results provide a theoretical and empirical foundation outlining the conditions under which the streetlight effect emerges, where data leads agents to look under the lamppost rather than engage in individually and socially beneficial exploration.

An Experiment on Gender Representation in Majoritarian Bargaining

Women are underrepresented in business, academic, and political decision-making bodies across the world. To investigate the causal effect of gender representation on multilateral negotiations, we experimentally manipulate the composition of triads in a majoritarian, divide-the-dollar game. We document a robust gender gap in earnings, driven largely by the exclusion of women from alliances rather than differential shares within alliances. Experiments with different subject pools show that distinct bargaining dynamics can underlie the same inequitable outcomes: While gender-biased outcomes can be caused by outright discrimination, they can also be driven by more complex dynamics related to differences in bargaining strategies. We identify two fundamental gender differences in bargaining dynamics. First, men are more likely to make opening offers and enjoy a payoff advantage for doing so, yet women that propose first do not and may even suffer backlash. Second, mixed-gender alliances are less stable when the excluded party is male rather than female. These findings show that there is no “one-size-fits-all” solution to the gender gap we uncovered and highlight the importance of studying bargaining dynamics in detail.

Does Portfolio Disclosure Make Money Smarter?

We provide causal evidence that mandatory portfolio disclosure helps investors evaluate and select hedge fund managers. Using a staggered difference-in-differences analysis, we demonstrate that investor capital flows better predict fund performance among funds that publicly disclose their portfolio holdings. Additional cross-sectional analyses suggest that this gain in selection ability varies with the informational value of disclosure. Furthermore, examining investor-level allocations, we find that institutional investors earn higher returns on their allocations to disclosing funds. Overall, these results help contribute to the cost-benefit analysis of mandatory portfolio disclosure.

Smoothed Semicovariance Estimation for Portfolio Selection

Minimizing the semivariance of a portfolio is analytically intractableand numerically challenging due to the endogeneity of the semicovariance matrix. In this paper, we introduce a smoothed estimator fort he portfolio semivariance. The extent of smoothing is determined by a single tuning constant, which allows our method to span an entire set of optimal portfolios with limit cases represented by the minimum semivariance and the minimum variance portfolios. The methodology is implemented through an iteratively reweighted algorithm, which is computationally efficient for optimization problems with many assets. Our numerical studies confirm the theoretical convergence of the smoothed semivariance estimator to the true semivariance. The resulting minimum smoothed semivariance portfolio performs well in- and out-of-sample compared to other popular selection rules.

On the validity of elicited risk attitudes

Increasing evidence points to the fact that the behavioral measures we use to elicit risk attitudes fail us. In recent years evidence has accumulated that our measures 1. correlate poorly with self-reported risk attitudes, real-world risk behaviors, and among themselves; 2. introduce distinct measurement errors and behavioral biases; 3. are not robust to sit-resit exercises.

Why do Risk Elicitation Tasks (RET) show so little predictive validity? Despite the large number of studies comparing risk elicitation procedures, the extent to which current RETs are able to capture self-reported or out-of-the-lab behavior is still partly unknown. Since each study can cover only a limited subset of tasks and self-reported or real-world behaviors, the map of the cross-correlations across tasks and behavior is far from being complete.

Luckily, the data to create such a map already exists. Experimental economists have been routinely eliciting risk attitudes for over three decades. In this presentation, I describe what we know so far about the psychometric validity of elicited risk attitudes, present a comprehensive, large dataset of the external validity of RETs, and describe two experimental designs (no data yet) aimed at tackling two of the main problems identified in the previous analyses -- measurement error and risk perception and modeling.

The ongoing effort, code and paper can be found at https://github.com/paolocrosetto/METARET and all data can be explored interactively at https://paolocrosetto.shinyapps.io/METARET_APP/

Informational Rents and the Excessive Entry Theorem: The Case of Hidden Action

Entry in a homogeneous Cournot-oligopoly is excessive if there is business stealing. This prediction assumes that production costs reduce profits and welfare equally. However, this need not be the case. If there is asymmetric information, suppliers or employees can utilize their superior knowledge to extract informational rents. Rent payments reduce profits and deter entry, but affect neither the optimal number of firms nor welfare directly. Therefore, entry becomes insufficient if informational rents are large enough. In the context of a moral hazard model, we show that insufficient entry occurs if entry costs are sufficiently high. Such costs lower the number of firms and, thereby, raise informational rents.

Global house prices since 1950

Stock-Oil Comovements Through Fear, Uncertainty, and Expectations: Evidence From Conditional Comoments

This paper investigates the dependencies and comovements between the S&P 500 and WTI by means of time-varying conditional comoments from April 1983 to December 2021 at the daily level. The conditional comoments mark a new pattern between the two markets’ dependencies since the 2008 global financial crisis (GFC). We employ three macroeconomic sentiment measures, including VIX (representing fear), economic policy uncertainty (representing uncertainty), and expected business condition index (representing expectation), to investigate the underlying mechanism for the new emerging stock-oil co-movements, using the time-varying parameter vector autoregression (TVP-VAR) to investigate the time-varying impulse responses, and the nonlinear autoregressive distributed lag (NARDL) model to analyze the asymmetries in the short- and long-run effects of the sentiment indices for the pre- and post-GFC periods. The conditional comoments of both markets change direction since the GFC, with crude oil showing a stronger dependence on the S&P 500’s return, skewness, and tail events than its counterpart. Overall, the time-varying impulse responses show heightened short-lived responses to the three sentiment indices after the 2008 GFC, with an asymmetric response from the conditional comoments of WTI (negative) and S&P 500 (positive) to the positive shocks in the fear index during the post-GFC period. The NARDL regression results prove that the explanatory power of the three sentiment indices increase largely after the GFC, pointing to the strong short-run asymmetries, and the ever-increasing effect of VIX . Further investigations reveal that oil-specific fear index (OVX) has weaker effect on stock-oil comoments than VIX.

Why Do People Commit Crime: Evidence from Inmates' Survey.

This project aims to document inmates' knowledge, perceptions, and preferences and examine how they differ from the general population and how they change over time. The project will also test selected theories explaining the causes of criminal behavior in a unified framework. The analysis will be based on survey, experimental, and administrative data collected from inmates in Czech prisons. Data collection includes two waves of surveys with around 500 inmates and 220 students and one wave of surveys with 1000 subjects representing the general population. Surveying inmates twice is key to measuring how their knowledge, perceptions, and preferences evolve during their incarceration.

Writing letters pays off: How to improve tax compliance quickly and efficiently?

Merged data of changes in real estate ownerships from the Cadastre Portal of the Slovak Republic and databases of taxplayers from the Financial Administration of the Slovak Republic revealed many sold properties over last 5 years, for which a capital gains tax was not paid. To improve the tax compliance, reminder letters were sent to potential evaders. In addition, reminder letters were also sent to former owners of properties that were sold in the current reporting period. On top of a baseline letter, we focused on behavioural effects of deterrence and morale and an explanatory effect of a graphical leaflet. The letters significantly increased the number of taxpayers paying the tax but there was no significant difference between the baseline letter and its behavioural modifications and the leaflet even backfired.