Skip to main content
Yale

Cowles Foundation for Research in Economics

Fostering the development and application of rigorous logical, mathematical, and statistical methods of analysis

Cowles Foundation Discussion Papers

New Cowles Foundation Discussion Papers

Discussion Paper
Abstract

I develop a growth model with multiproduct firms where technological obsolescence and idiosyncratic volatility jointly shape product and firm turnover. Quantifying the model using data on all U.S. private nonfarm employer businesses, I find both are important, with strong technological obsolescence needed to match firm exit by age. Rationalizing the observed sustained entry of new firms at a stable average size then requires substantial improvements across successive generations of entrants. The intertemporal knowledge spillovers needed to endogenously sustain these improvements generate an 8 to 28 percentage point wedge between social and private rates of return to innovation under laissez-faire.

Discussion Paper
Abstract

Young adults in advanced economies have grown more pessimistic about their economic prospects in recent years. At the same time, gambling-like financial activities have proliferated. We link these trends through a Friedman-Savage (1948) motive: when life goals such as homeownership or marriage seem out of reach, individuals turn to financial gambles. We test this theory using an original survey combined with millions of linked records from Korea’s leading credit bureau, telecom provider, and credit card issuer. Consistent with the theory, individuals who place greater importance on hard-to-reach goals take more risk, and the form of risk they take varies systematically with their distance from those goals. Within individuals, cryptocurrency app usage rises during unemployment and falls after winning the right to purchase price-regulated housing through a lottery. A calibration shows how hard-to-reach goals lead households to act more aggressively on their beliefs, amplifying the consequences of overoptimism.

Discussion Paper
Abstract

Bilateral bargaining under incomplete information—where a buyer and seller each hold a private reservation price—provides a controlled testbed demanding individual rationality, strategic surplus maximization, and coordination to realize gains from trade. We develop a structured bargaining environment in which LLMs negotiate via tool calls within an event-driven simulator, separating binding offers from natural-language messages to enable automated evaluation at scale; the environment serves both as a benchmark for frontier models and as a training environment for open-weight models. In a round-robin benchmark among five frontier models (15,000 negotiations), effective strategies implement price discrimination through sequential offers: aggressive anchoring, calibrated concession, and temporal patience are associated with both the highest surplus share and the highest deal rate, while accommodating strategies that concede quickly yield the lowest surplus capture and deal completion. Strategically competent models scale their behavior proportionally to item value, maintaining consistent performance across price tiers; weaker models succeed only when wide zones of possible agreement compensate for suboptimal strategies. In a pilot training study on Qwen3 (8B, 14B) using supervised fine-tuning (SFT) followed by Group Relative Policy Optimization (GRPO) under a specific composite reward, we observe a tension between the two stages in our setup: SFT approximately doubles surplus share but reduces deal rates, while GRPO recovers deal rates at the cost of surplus gains. The mechanism is that GRPO converges to an accept-heavy policy—accepting the opponent’s standing offer is the lowest-variance way to guarantee closure under the group-normalized advantage—trading per-deal surplus for high deal rate; we interpret the training experiments as a diagnostic study that identifies an open reward-design question rather than a proposal for a generally-applicable recipe. SFT also compresses surplus variation across price tiers, and this compression generalizes to opponents unseen during training, suggesting that behavioral cloning instills proportional strategies rather than memorized price points.

Discussion Paper
Abstract

We study the impact of generative AI on labor market signaling using the introduction of an AI-powered cover letter writing tool to a subset of workers on a large online labor platform. Our data track both access to the tool and usage at the application level. Difference-in-differences estimates show that access to the tool increased textual tailoring between cover letters and job posts. After the tool’s introduction, the correlation between cover letter tailoring and callbacks fell by 52%. The decline is immediate and larger for experienced employers. In response, employers shifted toward alternative signals, namely workers’ work histories.

Discussion Paper
Abstract

We derive an aggregate state-space representation of state-dependent pricing models with random menu costs and strategic complementarities. The representation, derived from first principles, replaces the cross-sectional distribution of price gaps with a finite set of its moments (the first three), while preserving the key economic forces of state-dependent pricing–movements in the frequency of price adjustment and selection. Along the impulse response, inflation is an odd (cubic) polynomial in the average price gap. Embedded in a general equilibrium setting, the model can be solved by perturbation, filtered, and estimated with standard DSGE tools, and it matches well the global solution across shock sizes. Applied to the 2021–22 inflation surge in the United States and the euro area, it shows that much of the variation that linear models would attribute to cost-push shocks is in fact endogenous amplification through the frequency of price changes and selection.

Discussion Paper
Abstract

Limited attention forces organizations to decide not only how much to discuss, but also which issues merit discussion. We study strategic communication about a multidimensional decision when a receiver can respond only along a few endogenously chosen issues. Players agree on the ideal action but prioritize different errors. In equilibrium, communicated and omitted issues must be statistically unrelated and separable according to the sender’s preferences. Thus, the sender’s priorities determine the agenda; the receiver’s do not. Synchronized priorities raise the receiver’s best equilibrium payoff but lower his worst, so he may prefer a less synchronized sender.

Discussion Paper
Abstract

A long-standing challenge is how to solve and characterize equilibrium in environments with persistent, dispersed, and endogenous information, known as the “forecasting the forecasts of others” problem (Townsend, 1983). In this paper, we develop a new method that overcomes this challenge. We show that equilibrium laws of motion are algebraic functions: they satisfy explicit polynomial equations, although they need not be rational functions. Our method yields a finite-dimensional equilibrium characterization even when its dynamics in the time domain do not admit a finite-order ARMA representation. We apply the method to derive comparative statics, characterize price dynamics in asset pricing and industry equilibrium models, and construct multiple equilibria with endogenous information aggregation.

Discussion Paper
Abstract

We characterize equilibria in a canonical class of economies with endogenous production networks, showing that log-prices equal a min-max over networks of a primitive affine function of log-productivities. This formula captures how resilience and fragility interact: the "min" reflects resilience from substituting suppliers, while the "max" reflects fragility from complementarity-induced rigidity. Applying these results, we generalize Hulten's theorem, finding that local Domar weights are generally insufficient to describe propagation of even small shocks. Moreover, we characterize how microeconomic volatility affects GDP, finding that smooth changes in Domar weights overstate its adverse effects by omitting option value from switching suppliers.

cowles-foundation-1954

History

In 1932, Alfred Cowles founded the Cowles Commission for Research in Economics in Colorado Springs. The Commission moved to Chicago in 1939, and finally to the Yale Department of Economics in 1954, where it was renamed the Cowles Foundation for Research in Economics.

Our Research Programs

Algorithms, Data, and Market Design

Econometrics

Economic Theory

Industrial Organization

Journal Publications