Natacha Spitzer
Ph.D. candidate in Economics, University of Pennsylvania
On the 2026–27 academic job market
I am a Ph.D. candidate in Economics at the University of Pennsylvania, working in microeconomic theory. I am interested in how information shapes incentives within strategic relationships, in particular when it is revealed over time. My job market paper studies contracting between a principal and an agent when neither knows how productive their relationship will be. My other work concerns monitoring and reputation, and repeated games on networks.
Job market paper
Contracting with Unknown Complementarity: Ratchets and Replacement Risk
How do a firm and a worker contract while learning how productive their relationship is? A worker's effort simultaneously generates value for the firm and reveals information about the fit between the worker and his assignment. But the prospect of future wage adjustments and replacement distorts his incentives to experiment. I study these forces in a two-period principal–agent model with moral hazard, limited commitment, and limited liability. I first establish existence of a pure-strategy equilibrium and characterize the optimal spot contract: when there is a chance that effort is counterproductive, the agent may be rewarded following both sufficiently strong and sufficiently poor performance, as both are evidence of effort. After the first period, the principal decides whether to replace the agent and, conditional on retention, offers a new wage contract. A forward-looking agent therefore has an incentive to influence her belief, and because he captures limited-liability rents in each period, replacement is costly to him. This incentive decomposes into an extensive retention force and an intensive wage force, which acts as a ratchet when good news raises the bar for pay. Relative to a benchmark without belief manipulation, the retention force dominates when the parties are initially pessimistic about the fit, leading to over-experimentation, while the ratchet dominates when they are initially optimistic and effort is not too costly, leading to under-experimentation. Dynamic incentives thus raise experimentation and learning when the agent's job is at risk and lower it once it is secure.
Working papers
Optimal Monitoring of AI Use with Reputation Concerns
How should managers monitor workers' use of AI when workers care about how they are perceived? A worker of privately known ability chooses between costly effort and an AI tool to complete a task, while an accuracy-seeking manager decides how often to monitor the worker at a cost. Using AI affects the worker's reputation through two channels. Its effect on the probability of success in the task depends on his ability, and, when he is monitored, his choice of technology itself becomes a signal of that ability. The optimal policy is simple: either the manager does not monitor and everyone uses the tool, or she monitors just enough to sustain separation, in which high-ability workers exert effort and low-ability workers use the tool. Mandatory disclosure is never optimal, and optimal monitoring is non-monotonic in AI quality. Relative to the efficient benchmark, the optimal policy can induce both over- and under-adoption. The policy has the same shape whether effort and the tool are substitutes or complements. When monitoring can depend on the outcome, the optimal audit targets success when able workers are rare and failure when they are common.
On the Irrelevance of Networks in Repeated Games
When does the behavior that can be sustained in markets and societies depend on how their members are related? I study this question in environments with many bilateral repeated games, in which each player plays an identical game simultaneously with each of her opponents. Using a multimarket Cournot counterexample with symmetric firms and markets, I show that whether collusion can be sustained in all markets depends both on which firms interact and on what each firm observes about other markets. Observing other markets lets firms coordinate punishments harsher than any single market can sustain. I then develop a general framework and show that network structure is irrelevant whenever each player's minmax payoff is a static Nash equilibrium payoff, a condition I call Optimality of Nash Reversion. Under it, an action profile is sustainable between all pairs of opponents if and only if it is sustainable in the corresponding two-player game.
Teaching
At Penn I have taught Game Theory (Summer 2025) and Intermediate Microeconomics (Summer 2023) as instructor, and have been a teaching assistant for Contract Theory, Industrial Organization, Mathematical Economics, Intermediate Microeconomics, Introduction to Microeconomics, Microeconomics for Managers (Wharton Executive MBA), and Behavioral Economics. I have earned Penn's CETLI Teaching Certificate.
Contact
- Address
- Department of Economics, University of Pennsylvania
133 South 36th Street, Philadelphia, PA 19104 - References
- George Mailath (advisor), gmailath@econ.upenn.edu
Kevin He, hekevin@econ.upenn.edu
Rakesh Vohra, rvohra@sas.upenn.edu
Kent Smetters (teaching), smetters@wharton.upenn.edu