Modern Macro, Measurement, and Modeling
Paper Session
Sunday, Jan. 3, 2027 8:00 AM - 10:00 AM (EST)
- Chair: Aleksandr Gevorkyan, St. John's University
Biased Applications of Economic Models: How Common Is This?
Abstract
Bias in the use of economic models has been widely studied and debated. Modeling bias can be found across all three branches of the modeling process: (1) accurate specification of the model, (2) logical application of the model, and (3) objective testing of the model. The first and the third branches have been most often studied. Arguments about the appropriateness of model specifications and empirical tests of model conclusions abound in the economics literature. But there are fewer detailed studies of how objectively and meticulously models, whatever their inherent other biases, are actually used to generate conclusions. This paper looks at a sample of articles that use a variety of economic models to examine how accurately each particular model’s logical framework is respected. While the research is still underway, initial analyses suggest that inaccurate or incomplete applications of models are common. Several examples from international trade theory show that authors often exploit their models partially, thus failing to take full advantage of the scope of the fundamental model. We find that important conclusions are missed. In macroeconomics, the fundamental framework of popular growth models is seldom fully exploited. This paper also reports on models of immigration, technological change, and factor markets. In all these cases, the practical issues related to the models examined are politically and economically important. We thus conclude with a section on the political economy of modeling, and we ask whether the uncovered application biases were intentional or merely cultural.Fisher Dynamics across the Income Distribution
Abstract
It is well-known that household debt has risen relative to GDP since the 1980s. During this period, debt-to-income ratios have also risen across the income distribution, dominated by increases in mortgage debt. In this paper, we examine how these similarities mask differences in the drivers of rising household leverage across the income distribution. We use the Survey of Consumer Finances for 1989-2022 to show how ‘Fisher Dynamics’ – namely, new borrowing, interest rates, income growth, and inflation rates – differentially affect the evolution of household debt by quantiles of the income distribution. We show that there are systematic differences in these factors by income quantile: while new borrowing drives rising debt at higher incomes, slower income growth and higher borrowing costs, in addition to new borrowing, drive debt-to-income ratios at lower quantiles of the income distribution.Solidarity Externalities and Suboptimal Competitive Wages
Abstract
Standard general equilibrium models treat wages as market-clearing prices, abstracting from labor's intrinsic dignity value and the solidarity externalities social economics holds central to economic life. This paper develops a tractable general equilibrium model in which workers derive utility from consumption and from the dignity embedded in their wage, and in which solidarity preferences generate positive externalities across workers. We establish four theorems. Competitive equilibrium is Pareto-inefficient when solidarity externalities are positive: the planner's wage schedule strictly dominates competitive wages for all productivity types. A solidarity premium — the gap between socially optimal and competitive wages — is strictly positive and increasing in the solidarity parameter. The socially optimal schedule compresses the wage distribution, providing a welfare-theoretic foundation for concern about inequality. Under private information, an optimal wage floor set at the planner's wage for the least-productive worker raises social welfare when solidarity externalities exceed a lower bound derived from labor supply parameters, a condition satisfied under all empirically plausible calibrations. Structural parameters are calibrated using the RAND Health Insurance Experiment (N ≈ 20,000 person-years), which identifies consumption utility curvature through health demand responses to price variation, and a longitudinal wage panel (Vella and Verbeek 1998), which identifies effort cost curvature through within-person variation; both available via the statsmodels Python library. Under parameterization consistent with experimental estimates of social preferences, results imply a solidarity premium of 11.4% of the competitive median wage, a 22% compression of the wage ratio between the highest and lowest productivity types, and a welfare gain equivalent to a 6.3% uniform consumption increase from implementing the optimal wage floor. These findings formalize the social economics critique of competitive wage-setting and establish rigorous microfoundations for minimum wage and living wage policy.Neural Networks, Natural Bubbles: Trust in AI and Aggregate Financial Risk
Abstract
This paper studies how heterogeneous trust in AI-generated signals influences financial decision-making and affects aggregate market stability. As artificial intelligence increasingly provides forecasts and recommendations in financial environments, individuals differ in the extent to which they rely on algorithmic signals relative to their own beliefs and information. The paper develops a parsimonious framework in which agents combine private priors with a common AI signal through an individual trust parameter that determines the weight assigned to algorithmic advice. Within this framework, higher average trust in AI increases the sensitivity of aggregate outcomes to the AI signal and generates a divergence between privately optimal reliance on AI and socially optimal market stability. Even when the AI signal is accurate, heterogeneous reliance can amplify fluctuations in aggregate outcomes because individuals respond differently to the same information. To examine these mechanisms empirically, the study implements a case-based survey experiment in which participants make sequential investment decisions. Respondents first choose an investment allocation based on their own judgment, then observe a common AI-generated forecast, and finally revise or maintain their decision. This design allows direct measurement of prior beliefs, trust in AI, and updating behavior following the algorithmic signal. The results document substantial heterogeneity in AI reliance across individuals, with more confident or experienced participants tending to rely more on their private beliefs and less confident participants placing greater weight on AI recommendations. The analysis shows that heterogeneous trust can generate aggregate inefficiencies through dispersion in actions despite common signals, selection effects in aggregate beliefs arising from correlations between trust and priors, and dynamic polarization in AI reliance as individuals learn from outcomes over time. These findings highlight important implications for AI governance, human–AI decision design, and portfolio management in increasingly algorithm-driven financial environments.JEL Classifications
- E1 - General Aggregative Models
- B5 - Current Heterodox Approaches