Publications

We use a rigorous three-stage many-analysts design to assess how different researcher decisions—specifically data cleaning, research design, and the interpretation of a policy question—affect the variation in estimated treatment effects. A total of 146 research teams each completed the same causal inference task three times each: first with few constraints, then using a shared research design, and finally with pre-cleaned data in addition to a specified design. We find that even when analyzing the same data, teams reach different conclusions. In the first stage, the interquartile range (IQR) of the reported policy effect was 3.1 percentage points, with substantial outliers. Surprisingly, the second stage, which restricted research design choices, exhibited slightly higher IQR (4.0 percentage points), largely attributable to imperfect adherence to the prescribed protocol. By contrast, the final stage, featuring standardized data cleaning, narrowed variation in estimated effects, achieving an IQR of 2.4 percentage points. Reported sample sizes also displayed significant convergence under more restrictive conditions, with the IQR dropping from 295,187 in the first stage to 29,144 in the second, and effectively zero by the third. Our findings underscore the critical importance of data cleaning in shaping applied microeconomic results and highlight avenues for future replication efforts.

Large Language Models (LLMs) such as ChatGPT are transforming how scientists conduct and validate research, offering promise as tools to improve scientific reproducibility. However, computational reproducibility and error detection remain expensive and labor-intensive. We experimentally test how collaboration between researchers and LLM assistants influences the reproduction of quantitative social science findings across different levels of AI autonomy. We randomly assigned 288 researchers to 103 teams working under three conditions: human-only, AI-assisted (using ChatGPT as a collaborative tool), or AI-led (ChatGPT operating with minimal human oversight). Teams reproduced published results from leading social science journals, detected coding errors, and proposed robustness checks. Human-only and AI-assisted teams achieved comparable reproduction rates (94% vs. 91%) and performed similarly on most outcomes, except human-only teams identified significantly more major coding errors. Both substantially outperformed AI-led teams, which achieved only a 37% reproduction rate, detected fewer errors across all categories, proposed weaker robustness checks, and required more time. This autonomous approach, however, likely represents only a lower bound of AI capabilities. Despite rapid model advances, expert human judgment currently remains indispensable for reliable empirical verification. While AI assistance did not degrade most outcomes, it provided no measurable advantages and was associated with reduced detection of major errors. However, the 37% autonomous reproduction rate indicates that AI could provide value in settings where scale or cost constraints preclude human review of papers, even though general-purpose LLMs offer no immediate advantages for human-supervised verification.

Colocation services offered by stock exchanges enable market participants to achieve execution costs for large orders that are substantially lower and less sensitive to transacting against high-frequency traders. However, these benefits manifest only for orders executed on the colocated brokers' own behalf, whereas customers' order execution costs are substantially higher. Analyses of individual order executions indicate that customer orders originating from colocated brokers are less actively monitored and achieve inferior execution quality. This suggests that brokers do not make effective use of their technology, possibly due to agency frictions or poor algorithm selection and parameter choice by customers.

Science aspires to be cumulative. Reproducibility efforts strengthen science by testing the reliability of published findings, promoting self-correction, and informing policy-making. Computational reproductions, whereby independent researchers reproduce the results of published studies, are an essential diagnostic tool. Such efforts should have greater visibility. However, little social science reproduction and robustness has been conducted at scale. Here we reproduced original analyses and conducted robustness checks of 110 articles that were published in leading economics and political science journals with mandatory data and code sharing policies. We found that more than 85% of published claims were computationally reproducible. In robustness checks, our reanalyses showed that 72% of statistically significant estimates remain significant and in the same direction, and the median reproduced effect size is nearly the same as the originally published effect size (that is, 99% of the published effect size). Additionally, 6 independent research teams examined 12 pre-specified hypotheses about determinants of robustness. Research teams with more experience found lower levels of robustness, and robustness did not correlate with author characteristics or data availability.

The same dataset can be analysed in different justifiable ways to answer the same research question, potentially challenging the robustness of empirical science. In this crowd initiative, we investigated the degree to which research findings in the social and behavioural sciences are contingent on analysts' choices. We examined a stratified random sample of 100 studies published between 2009 and 2018, in which, for one claim per study, at least five reanalysts independently reanalysed the original data. The statistical appropriateness of the reanalyses was assessed in peer evaluations, and the robustness indicators were inspected along a range of research characteristics and study designs. We found that 34% of the independent reanalyses yielded the same result (within a tolerance region of ±0.05 Cohen's d) as the original report; with a four times broader tolerance region, this indicator increased to 57%. Of the reanalyses conducted, 74% reached the same conclusion as the original investigation, 24% yielded no effects or inconclusive results and 2% reported the opposite effect. This exploratory study indicates that the common single-path analyses in social and behavioural research should not be simply assumed to be robust to alternative analyses. Therefore, we recommend the development and use of practices to explore and communicate this neglected source of uncertainty.

In statistics, samples are drawn from a population in a data-generating process (DGP). Standard errors measure the uncertainty in estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence-generating process (EGP). We claim that EGP variation across researchers adds uncertainty—nonstandard errors (NSEs). We study NSEs by letting 164 teams test the same hypotheses on the same data. NSEs turn out to be sizable, but smaller for more reproducible or higher rated research. Adding peer-review stages reduces NSEs. We further find that this type of uncertainty is underestimated by participants.

Prior research has established that the presence of designated market makers (DMMs) in an electronic open limit order book increases liquidity. We analyze whether the presence of additional DMMs results in a further improvement in liquidity. Using data from Deutsche Börse's Xetra system we find that increases in the number of DMMs significantly improve liquidity, and vice versa for decreases in the number of DMMs. Our results are confirmed when we use an instrumental variables approach to overcome potential endogeneity issues.

By employing a dynamic model with two limit order books, we show that fragmentation is associated with reduced competition among liquidity suppliers and lower picking-off risk of limit orders. Due to these countervailing channels, the impact of fragmentation on liquidity and welfare differs with asset volatility: When volatility is high (low), liquidity and aggregate welfare in a fragmented market are higher (lower) than in a single market. However, fragmentation always shifts welfare away from agents with exogenous trading motives and toward intermediaries. We empirically corroborate our model’s predictions about liquidity. Our model reconciles the mixed results in the empirical literature.

We analyze the determinants of the trading volumes of different trading mechanisms in equity markets using an extensive panel data set from European markets comprising public limit order books, call auctions, dark pools, internalization platforms, and the over-the-counter market. Market shares, resulting from investors' order routing decisions, are driven by the degree of immediacy and anonymity offered by the venues, their ability to offer off-tick executions, as well as the informational environment and conditions in the market. Findings for small and large trades are distinctly different, likely because traders jointly choose trade size and venue type.

Many equity markets combine continuous trading and call auctions. Oftentimes designated market makers (DMMs) supply additional liquidity. Whereas prior research has focused on their role in continuous trading, we provide a detailed analysis of their activity in call auctions. Using data from Germany's Xetra system, we find that DMMs are most active when they can provide the greatest benefits to the market, i.e., in relatively illiquid stocks and at times of elevated volatility. Their trades stabilize prices and they trade profitably.

Corporate insider trades predict idiosyncratic return skewness. CEO purchases are followed by an increase and CEO sales by a decrease in idiosyncratic skewness. The evidence suggests that this effect is driven by personal preferences rather than behavioral biases such as overconfidence. Our findings are consistent with the interpretation that CEOs, who are generally considered to be underdiversified, optimize their holdings by taking their preference for positive return skewness into account. We observe particularly robust results for CEO sales, which substantiates the less common notion that insider sales can be informative for investors.

This paper presents the most extensive analysis of liquidity in the German equity market so far. We examine the evolution of liquidity over time, the determinants of liquidity, and commonality across liquidity measures and countries. We make use of a new publicly available dataset, the Market Microstructure Database Xetra (MMDB-Xetra). We find that liquidity has generally increased over time, and that in times of crisis liquidity is lower and the volatility of liquidity is significantly higher. Commonality in liquidity is highest during the financial crisis. We also find significant commonality between liquidity in the US and the German equity markets.

This paper investigates whether corporate insiders trade when asymmetric information is high, using data on U.S. corporate insider transactions between 1986 and 2012. We generalize the literature focusing on insider trading around the announcement of different categories of corporate events. The key innovation of this paper is our asymmetric information proxy relivol, which measures deviations of idiosyncratic volatility from a firm's normal level. Our findings suggest that relivol positively predicts insider purchases, indicating that insiders buy shares when their informational advantage is high. However, insiders appear to sell less when relivol is high, which is consistent with existing evidence on sales being driven by alternative, non-information-related trading motives such as liquidity or diversification needs. Furthermore, we find that profits are significantly higher when insiders buy during periods of high relivol but not when they sell shares.

Technological advances and regulatory initiatives have led to the emergence of a competitive, but fragmented, equity trading landscape in several markets around the world. While these changes have coincided with benefits like reduced transaction costs, advancements in trading technology, and access to a diverse array of execution venues, regulators and market participants have also raised concerns about the welfare implications of innovations like dark pools as well as the resulting increase in execution complexity. Exchanges are often viewed as natural monopolies due to the presence of network externalities and economies of scale. However, heterogeneity in traders' preferences means that no single venue can serve the interests of all investors. Fragmentation of the marketplace can be seen as a direct outcome of this heterogeneity. In this paper, we review the theoretical and empirical literature examining the economic arguments and motivations underlying market fragmentation, the resulting implications for liquidity and price efficiency, and the role for public policy. Beyond the concerns for equity markets, the lessons from this literature are relevant for other asset classes experiencing an increase in competition between trading venues.

In early 2014, the European Council, the European Parliament, and the European Commission reached an agreement on MiFID II (Markets in Financial Instruments Directive) and on MiFIR (Markets in Financial Instruments Regulation). Because the new regulatory framework will significantly influence and likely reduce European over-the-counter (OTC) equities trading, our article investigates the current state of OTC trading in Europe. Therefore, we investigate OTC transaction sizes for Euro STOXX 50 constituents between January 2008 and June 2013, discuss whether and how they might have caused market impact, and give insights on how this affects the European trading landscape.

This paper reconsiders the effect of investor sentiment on stock prices. Our main contribution is that, in addition to the intermediate term return predictability, we also analyze the immediate price reaction to the publication of survey-based investor sentiment indicators. We find that the sign of the immediate market response is the same as that of the predictability at intermediate time horizons. This is consistent with underreaction to cash flow news or with investor sentiment being related to mispricing. It is inconsistent with the alternative explanations of a rational response to cash flow news or sentiment indicators providing information about future expected returns.

We apply the conditional approach of Pettengill, Sundaram, and Mathur (1995) to the predominant model in asset pricing, the Fama-French three-factor model. We find that all three risk factors cross-sectionally drive asset returns. Further, we extend the test developed by Freeman and Guermat (2006) to multi-factor models and test if risk premia are priced within the conditional approach. Our test leads to qualitatively identical results as the Fama-MacBeth (1973) test, thus confirming its validity. Additionally, our use of portfolios sorted by various characteristics accentuates how critical the choice of test portfolios is to empirical asset pricing.