This paper estimates the cross-sectional distribution of Epstein-Zin preference parameters in a Large administrative panel of Swedish households. We consider a life-cycle model of saving and portfolio choice that incorporates risky labor income, safe and risky financial assets inside and outside retirement accounts, and real estate. We study middle-aged stockowning households grouped by education, industry of employment, and birth cohort as well as by their accumulated wealth and risky portfolio shares. We find some heterogeneity in risk aversion (a standard deviation of 0.47 around a mean of 5.24 and median of 5.30) and considerable heterogeneity in the time preference rate (standard deviation 6.0% around a mean of 6.2% and median of 4.1%) and elasticity of intertemporal substitution (standard deviation 0.96 around a mean of 0.99 and median of 0.42). Risk aversion and the EIS are almost cross-sectionally uncorrelated, in contrast with the strong negative correlation that we would find if households had power utility with heterogeneous risk aversion. The TPR is weakly negatively correlated with both the other parameters. We estimate lower risk aversion for households with riskier labor income and higher levels of education, and a higher TPR and lower EIS for households who enter our sample with low initial wealth.
Humans and artificial intelligence (AI) algorithms increasingly interact on unstructured managerial tasks. We propose that tailoring this human-AI interaction to align with individuals' cognitive preferences is essential for enhancing performance. This hypothesis is examined through a field experiment in a multinational pharmaceutical firm. In the experiment, we manipulated four contextual parameters of human-AI interaction-work procedures, decision-making authority, training, and incentives-to align with sales experts' cognitive styles, categorized as either adaptors or innovators. Our results show that tailored interaction significantly improves sales performance, whereas untailored interaction results in negative treatment effects compared with both the tailored and control conditions. Qualitative evidence suggests that this negative outcome arises from role conflicts and ambiguities in untailored interaction. Exploring the mechanisms underlying these outcomes further, a mediation analysis of AI login data reveals that human-AI interaction tailoring leads sales experts to adjust their AI utilization, which contributes to the observed performance outcomes. These findings support a human-centered approach to AI that prioritizes individuals' information-processing needs and tailors their interaction with AI accordingly.
Agricultural extension programs often train a subset of farmers and rely on social networks for knowledge dissemination. We evaluate this approach through a two-stage experiment of an agronomy training program among Rwandan coffee farmers. The first stage randomized trainee concentration at the village level; the second randomly selected participants within villages. The program increased knowledge and self-reported adoption of the taught practices. It also reshaped farmers’ social networks by creating new social ties primarily among co-trainees. At first glance, the intervention appeared modestly effective, with treated farmers exhibiting higher yields than control applicants within the same village, although this difference is not statistically significant. However, knowledge did not diffuse, and control farmers with more treatment friends at baseline reduced audited adoption and input use. Villages with high trainee concentrations showed suggestive evidence of negative spillovers, consistent with reallocation of labor and other shared inputs toward treated farmers, as formalized in a simple model in which social ties shape access to inputs. Relative declines in the yields of control farmers in these villages account for treatment-control differences, raising concerns both about this dissemination strategy and estimates that fail to consider potential negative spillovers.
Despite significant efforts by humanitarian actors, initiatives, and donors to improve coordination among humanitarian organizations during disaster response, the challenge of insufficient coordination persists. Drawing on practical considerations, we develop a stylized non-cooperative game-theoretical model to examine the coordination dynamics between large international and small local humanitarian organizations in the aftermath of a disaster, comparing both pooled and partitioned coordination models. Our findings reveal that bureaucratic delays commonly associated with coordination efforts not only deter collaboration but can also result in coordination levels that are detrimental to overall relief system performance. This analysis underscores the importance of rethinking coordination structures to better reflect the specific context of relief efforts. In alignment with calls for increased localization, we demonstrate that an efficiently designed partitioned coordination model outperforms a pooled model, which often marginalizes smaller local actors. This partitioned approach proves particularly effective in emergency settings, bringing actors’ decisions closer to the optimal outcome for the entire system.
Background and objective: Assessing embryo quality objectively and with greater precision remains a key challenge in assisted reproductive technologies (ART). RNA molecules excreted by the embryo are detectable in embryo spent media (ESM) and blastocoel fluid (BF). Their potential for predicting pregnancy outcome is however unknown. The aim of the study was to determine whether embryo excreted RNA can serve as biomarkers to predict pregnancy outcome in ART. Material and methods: the primary outcome of this multicenter study was to develop a prediction model for implantation outcome based on mRNA molecules in ESM and BF. We included 255 couples undergoing ART with single embryo transfers between 2018 and 2019 in four IVF clinics in Sweden. ESM and BF were collected from individually cultured embryos. RNA extraction was performed using our novel protocol developed for ESM and BF. To identify the mRNA signature, each sample underwent mRNA sequencing (Smart-seq3), followed by differential gene expression analysis. A machine learning (ML) approach was applied with the mRNA genes in BF as input data to create a prediction model for pregnancy results measured as a positive or negative urine human chorionic gonadotropin (hCG) test result 14 days after embryo transfer. Results: After pre-processing and quality controls, we included ESM from 105 cleavage stage and 230 blastocyst stage embryos and BF from 75 blastocyst stage embryos in the analysis. In total we identified 345 and 1132 mRNA genes in ESM from cleavage and blastocyst embryos respectively and 1157 mRNA genes in BF. There was a significant difference in RNA abundancy of 108 mRNA genes (p-value<0.05; fold change +2/-2) between hCG positive and negative embryos in BF but not in ESM. Consequently, the mRNA data from BF was selected as input data to train different ML models to predict pregnancy result after embryo transfer. The most optimal ML model combined L2 Regularized Logistic Regression with differentially abundant genes as feature and principal component analysis for preprocessing. The ML model predicted pregnancy result with a sensitivity of 83% and a specificity of 59% (AUC of 76% and accuracy of 69%). Testing of the model in an independent subset of 9 samples, not included in the training dataset, demonstrated a sensitivity of 60%, specificity of 75%, and accuracy of 67% in prediction of pregnancy result. Conclusion: For the first time, we demonstrate that RNA molecules can be comprehensively mapped in individual ESM and BF samples using a highly sensitive RNA isolation and sequencing protocol. Moreover, mRNA in BF can be used to predict pregnancy outcome of individual embryos in ART using a ML approach. Our novel prediction model has the potential to enhance embryo selection, thereby improving the success rate of ART treatment. In the future, larger clinical trials are needed to validate the robustness of the method and compare it against standard care. ### Competing Interest Statement The authors have declared no competing interest.
