🌐ENGLISH

We study how private information affects household responses to environmental risk. Using data from residential air quality monitors, we exploit the timing of monitor installation and high-frequency fine particulate matter (PM2.5) readings to identify responses to new information about indoor pollution risk. We find that indoor PM2.5 concentrations decline by 2.5 ug/m3 over the 12 weeks following installation, conditional on contemporaneous outdoor pollution, with effects significantly larger among households with high initial indoor pollution. The indoor–outdoor pollution gradient declines over time, indicating that households become increasingly effective at mitigating exposure when marginal health damages are highest. Using machine learning techniques to infer cooking activity and air purifier adoption, we show that households respond primarily through durable defensive investments rather than reductions in pollution-generating behavior, with back-of-the-envelope calculations implying positive net benefits. Our results suggest that personalized risk information increases the salience of indoor pollution as a controllable risk for households, in contrast to spatially coarse public information that frames pollution primarily as an outdoor threat requiring avoidance.

We test the real consequences of climate reporting harmonization. Leveraging CDP’s 2018 alignment of its questionnaire with the TCFD framework that made climate disclosures more comparable across firms and jurisdictions, we employ a difference-in-differences design comparing firms that adopted the TCFD-aligned CDP framework with contemporaneous non- adopters. Firms adopting the TCFD framework exhibit persistent and notable declines in Scope 1 and upstream Scope 3 emissions, consistent with a firm’s engagement in direct abatement in the post-adoption period rather than in indirect outsourcing in the same period. These effects are stronger where industry reporting converges and where cross-border competition heightens the value of comparability. We further provide evidence in support of an investor-driven mechanism: cross-border climate fund ownership increases from the pre- to the post-adoption period while climate-related shareholder proposals decline, consistent with improved ability of investors to allocate capital toward credible firms and reduced needs for direct monitoring via shareholder activism. Overall, the evidence suggests that a globally comparable climate disclosure framework enhances information processing and induces meaningful real reductions in corporate carbon emissions.

Recent cross-cultural work documents that for some traits and economic preferences, gender differences increase with economic development and gender equality. Two main explanations have been given in the literature for this paradoxical finding: one, that increased availability of material resources allows individuals to express more freely their gender-specific interests, and another, that gender stereotypes are amplified by culture in postmaterialist societies. In this paper we study this paradoxical phenomenon and disentangle its explanations for the trait of competitiveness. Using a sample of over 500,000 students representative of 77 countries, we find evidence of a process of cultural transmission that leads to a decrease in the competitiveness of girls as societies become more developed and gender equal while keeping the competitiveness of boys high and stable across.

This paper constructs a new long-run measure of geopolitical risk for China based on Chinese-language state media. Using the full text of People’s Daily articles from 1946 to 2022, we develop the People’s Daily Geopolitical Risk Index (PDGPR) and study its economic effects. VAR estimates show that geopolitical risk shocks are followed by short-run declines in output growth and inflation. Using firm-level data, we further find that higher PDGPR is associated with lower investment. While PDGPR shares substantial information content with global GPR measures, it provides a useful lens for tracing how international geopolitical shocks are transmitted into China’s domestic information environment and economic behavior.

We examine the impact of legal adult prostitution on the prevalence of trafficked child sex workers and children engaging in prostitution out of economic necessity. Using a nationwide survey of child sex workers, historically determined locations of legal brothels, and counterfactual simulations, we find that the presence of legal brothels increases the prevalence of trafficked children by 18.8 percent nationwide, primarily because they are camouflaged as legal prostitutes. Rigorous screening procedure within brothels mitigates this increase. However, the inability to exclude non-trafficked children who misreport their age as adults raises their prevalence beyond that of actual brothels, revealing policy tradeoffs.

We investigate possible motives for investment in `status goods', and in particular the role of instrumental motives. In a laboratory experiment, participants can accumulate a payoff irrelevant virtual good that may serve as a signal in partner selection for risky joint projects. We vary the observability of the status good distribution, the role of ability/effort with respect to the success probabilities of joint projects, and information about the nature of these success probabilities. We find that participants do accumulate the status good when the distribution of status goods is observable, unless the success probabilities of joint projects are independent from their ability/effort and they are explicitly told that this is the case. That is, we see accumulation of the status good even when the success of joint projects is not dependent on the participants' ability/effort. Accumulation of status goods increases the likelihood of being selected as a partner, but does not improve returns, and reduces overall earnings. These findings provide causal evidence of inefficient status-seeking driven by signaling motives.

This paper uses the Beijing sample of the China Household Finance Survey (CHFS) to evaluate large language models as tools for simulating economic behavior, using household consumption as a real decision setting. We focus on how survey information granularity and historical information affect simulation quality. Using DeepSeek V4 Pro, we construct household-level LLM simulation tasks and compare different input formats, including structured household summaries, fine-grained questionnaire responses, and their combination. We also benchmark the LLM results against traditional methods, including OLS, Random Forest, XGBoost, Ridge, and Elastic Net. We find that fine-grained questionnaire information improves not only household-level consumption fit, but also the fit to the overall distribution, tail behavior, and subgroup heterogeneity in the real data. Historical information mainly helps when current information is limited. Compared with traditional prediction methods, LLMs do not show a consistent advantage in point prediction, but perform better in distributional fit and heterogeneity preservation.

Fertility decline has become a global policy challenge. In this paper, we examine the impact of high-speed mobile internet access on fertility in a developing-country setting. For identification, we use a dynamic difference-in-differences design that exploits the staggered, regulator-driven rollout of 3G and 4G infrastructure across Brazilian municipalities. The results show that mobile internet access increases fertility. Furthermore, while digital connectivity delays family formation, it reshapes local labor markets toward jobs that can be performed from home. This effect is particularly large for women and suggests that the increase in fertility reflects women entering flexible formal jobs that are more compatible with having children. Policies that expand digital connectivity may therefore help slow fertility decline by easing the female employment–fertility trade-off.

We estimate the local incidence of U.S. data center openings using nationwide housing transaction and household demographic data via a stacked difference-in-differences design. Hyperscale data centers reduce nearby house prices by 6.8 percent, with monotonic effects fading beyond 14 kilometers. Price impacts increase with facility power capacity and are larger for the first data center in a county. Large non-hyperscale facilities induce little capitalization on average, though negative effects emerge in locations with greater residential exposure. Additional evidence is consistent with local utility costs, environmental burdens, social salience, and generic local undesirable land use contributing to the capitalization effects. Our findings highlight the importance of siting, scale, and residential exposure in evaluating place-based incentives for digital infrastructure build-up.