Abstract: This paper studies the dynamic evolution of human behavior and collaborative value—along both efficiency and fairness dimensions—in algorithm-assisted credit approval. Using transaction-level data from a leading Chinese auto finance leasing firm, we analyze loan officers’ override (rescue) decisions for applications initially rejected by a machine-learning credit scoring system. We document three main findings. First, rescued applications shift over time from being broadly distributed across low score ranges to clustering near the algorithmic cutoff, indicating increasingly margin-focused human intervention. Second, human overrides reduce reliance on coarse group-level information, thereby mitigating statistical discrimination, without sacrificing default performance or profitability, with collaborative value shifting from fairness gains early on to risk reduction over time. Third, mechanism analyses point to a career-concern channel: following algorithm adoption, loan officers initially exert greater effort to signal competence, as reflected in richer approval texts and longer processing times; as career concerns attenuate over time, incentives weaken, leading to partial free-riding on the algorithm and reduced human input. Together, these findings highlight the dynamic nature of human–machine collaboration and the importance of organizational incentives in sustaining its value.
Abstract: We find that consumption spreads through social networks via a “visibility bias” channel, consistent with the model of Han et al. (2023). Using county-level Facebook data and exogenous fracking-induced income shocks, we find that a 1% increase in a closely connected county’s consumption raises local spending by 0.35% in the following year. The effect is stronger for more socially connected households and for socially visible goods. Lacking a corresponding income boost, households respond by buying cheaper goods. This peer-induced spending strains household finances and increases local delinquency rates, underscoring how biased social observation can undermine financial stability
Discussant: Michaela Pagel, Washington University-St. Louis
