Investment scores generated by ChatGPT using conference call transcripts are strongly predictive of companies’ future capital expenditures for up to nine quarters, controlling for other predictors. The scores also predict intangible investment, R&D, and total investment, and are negatively correlated with future stock returns.

Using GPT to summarize corporate disclosures results in summaries that are shorter and more informative than the originals. “Bloated” disclosures – or, those with more uninformative content – are associated with adverse capital market outcomes.

Little support exists for a strong causal relation between corporate governance and firm outcomes, which are rather driven by such characteristics as profitability, liquidity, and growth, among others.

By applying machine-learning algorithms to human behavior data, this paper offers a novel approach to hypothesis generation for scientific research; the authors apply this framework to reveal that judges rely on facial characteristics of defendants when making incarceration decisions.

This novel research finds little evidence that smart thermostats have a statistically or economically significant effect on energy use, thus challenging assumptions about the effective scaling of new user-based technologies.

Online advertisers benefit from data that are shared across applications; loss of such data would increase costs per incremental customer for the median advertiser by 37% and mainly impact advertisers in consumer-packaged goods, e-commerce, and retail.

Bitcoin’s new form of decentralized trust suffers from a pick-your-poison conundrum with two possible outcomes: Either this trust is extremely expensive relative to its economic usefulness, or it is vulnerable to collapse.