Predictive algorithms inform consequential decisions in settings with selective labels: outcomes are observed only for units selected by past decision makers. This creates an identification problem under unobserved confounding — when selected and unselected units differ in unobserved ways that affect outcomes. We propose a framework for robust design and evaluation of predictive algorithms that bounds how much outcomes may differ between selected and unselected units with the same observed characteristics. These bounds formalize common empirical strategies including proxy outcomes and instrumental variables. Our estimators work across bounding strategies and performance measures such as conditional likelihoods, mean squared error, and true/false positive rates. Using administrative data from a large Australian financial institution, we show that varying confounding assumptions substantially affects credit risk predictions and fairness evaluations across income groups.

We investigate whether judges’ political ideology affects corporate tax behaviors. We find that firms engaging in less aggressive tax planning when Circuit Court judges are more liberal. Cross-sectionally, the deterrent effect of liberal judge ideology is more pronounced for firms that engage in judiciary-sensitive tax strategies, face higher enforcement risk from the Internal Revenue Service (IRS), or have larger reputational costs from tax disputes. Our findings further suggest that liberal judge ideology reduces firms’ R&D investments and market value by constraining tax planning. Overall, our evidence highlights the importance of judge ideology for firm behavior in the context of corporate tax planning.

We introduce a simple measure of informed trading risk, , the residual to liquidity quote‐improvement‐to‐deterioration ratio times . When facing with increased informed trading risk, liquidity providers compete less to provide liquidity, reducing their undercutting activity. Reductions in undercutting leave footprints in trade and quote data that are captured by . Unlike prior measures, is easy to construct, can be computed intraday, and is orthogonal to liquidity. The measure outperforms prominent existing alternatives in reflecting the extent of information asymmetry before earnings announcements, predicting unscheduled press releases, and identifying informed trading spillovers around them.

We derive tight pricing kernel restrictions from options with same‐day expiration (“0DTEs”). These restrictions concern the volatility of small and frequent asset price moves that the equity and options markets must agree on in a frictionless economy. Their violation leads to pseudo‐arbitrage opportunities, characterized by nontrivial reward‐to‐risk ratios over arbitrarily short horizons and achieved by a combined position in 0DTEs and the underlying asset. Empirically, we find no evidence of feasible pseudo‐arbitrage opportunities, as transaction costs, estimation risk, and short‐term volatility risk prevent investors from taking advantage of small and infrequent disagreements about volatility between equity and options markets.

We consider the dynamics of analyst inattention by investigating how analysts allocate their attention when a firm in their portfolio experiences CEO turnover. Our analysis shows that analysts tend to divert their attention toward firms that experience such events, resulting in less attention and a corresponding reduction in forecasting accuracy for nonevent firms. Furthermore, this reduction in accuracy varies with factors related to the costs and benefits of rationally allocating attention to firms that have experienced CEO turnover. Collectively, our analysis responds to the call for research on rational inattention among analysts and illustrates the specific intraportfolio events that alter attention allocation and information.