We analyze the computational reproducibility of more than 1,000 empirical answers to six research questions in finance provided by 168 international research teams. Running the original researchers’ code on the same raw data regenerates exactly the same results only 52\% of the time. Reproducibility is higher for researchers with better coding skills and for those exerting more effort. It is lower for more technical research questions, more complex code, and for results lying in the tails of the results distribution. Neither researcher seniority, nor peer-review ratings appear to be related to the level of reproducibility. Moreover, researchers exhibit strong overconfidence when assessing the reproducibility of their own research. We provide guidelines for finance researchers and discuss several implementable reproducibility policies for academic journals.

The integration of algorithmic trading and reinforcement learning, known as AI-powered trading, has significantly impacted capital markets. This study utilizes a model of imperfect competition among informed speculators with asymmetric information to explore the implications of AI-powered trading strategies on speculators' market power, information rents, price informativeness, market liquidity, and mispricing. Our results demonstrate that informed AI speculators, even though they are ``unaware'' of collusion, can autonomously learn to employ collusive trading strategies. These collusive strategies allow them to achieve supra-competitive trading profits by strategically under-reacting to information, even without any form of agreement or communication, let alone interactions that might violate traditional antitrust regulations. Algorithmic collusion emerges from two distinct mechanisms. The first mechanism is through the adoption of price-trigger strategies (``artificial intelligence''), while the second stems from homogenized learning biases (``artificial stupidity''). The former mechanism is evident only in scenarios with limited price efficiency and noise trading risk. In contrast, the latter persists even under conditions of high price efficiency or large noise trading risk. As a result, in a market with prevalent AI-powered trading, both price informativeness and market liquidity can suffer, reflecting the influence of both artificial intelligence and stupidity.

Machines dominate the trading process in modern markets, and it may be tempting to conclude that human behavior no longer affects liquidity generation and consumption. Our research challenges this view. Liquidity costs follow strong behavioral patterns, with costs highest in early winter and lowest in late spring, a difference driven by changes in risk aversion and impatience that correlate with seasonal changes in daylight exposure. As informed traders become more impatient from late summer to early winter, they generate more adverse selection, while liquidity providers demand greater compensation as they become more risk-averse. Together these patterns drive liquidity costs up and down annually.