seminars

Getting the Picture

Standard economic theory assumes that people have perfect information processing ability: they can derive all logical consequences of information immediately. In reality, however, people struggle to work out the consequences of information. In this paper, we offer a theory of the reasoning process—how people go about “connecting the dots.” In our theory, an agent may have all of the information needed to draw a conclusion yet they still fail to see it. Our key assumption is that agents have limited “working memory.” This constrains the number of pieces of information—“dots”— an agent can think about and connect at once. We explore how agents analyze pieces of information and work their way to a big picture—or, “narrative”—and how narratives, in turn, shape the agent’s view of the parts. We show that limited working memory makes sense of why people struggle with tradeoffs, suffer from choice overload, are influenced by defaults, engage in satisficing, selectively attend to attributes, are subject to primacy effects, and may vacillate between interpretations. We apply the framework to political persuasion, showing how supplying a simple narrative can lock in an interpretation of ambiguous evidence.

How to Train Your AI When Humans Matter

AI predicts; its prediction is used by humans in decisions. The value of AI therefore depends on how humans extend, verify, and act on that prediction. We model AI as part of a composite experiment where agents can verify predictions at cost, delegate to AI, or avoid the model altogether. We derive the optimal coverage-onditional accuracy tradeoff for training, showing that maximizing unconditional accuracy is generally suboptimal. The optimum is discontinuous where users switch between autonomous and verified AI regimes. Heterogeneous users disagree on the ideal model. Because the optimal model maps directly from the economic environment – downside risk, verification cost, adversarial pressure, task complementarities – knowledge of the economic environment can inform about the nature of optimal training ex-ante.

Wage Expectations and Job Search

In a field experiment with 9,000 Danish job seekers, we study how unemployed workers’ wage expectations affect job search and re-employment. In our survey, we generate exogenous variation in respondents’ wage expectations by informing a random half of them about re-employment wages of comparable workers. The intervention increases job-finding as measured in administrative data for both initially optimistic and initially pessimistic respondents, but through different channels: initial optimists lower their reservation wages and intensify search, while pessimists raise reservation wages and redirect applications toward local vacancies. Consistent with spatial search frictions, narrowing the geographic scope accelerates job finding among pessimists.

Mapping AI into Production: A Field Experiment on Firm Performance

AI can deliver productivity gains on individual tasks, yet evidence on whether these gains aggregate to firm performance remains limited. We study a central friction in AI adoption, which we call the mapping problem: discovering where and how AI creates value within a firm’s production process. Across 515 startups from around the world, we run a field experiment in which treated firms receive information about how other firms have reorganized production around AI, exposing founders to use cases across a broader set of firm functions and prompting them to search beyond the familiar applications they would otherwise default to. We find that treated firms discover more AI use cases, a 44% increase, concentrated in product development and strategy. These changes result in economically meaningful performance gains. Treated firms complete 12% more tasks, are 18% more likely to acquire paying customers, and generate 1.9x higher revenue. Revenue and investment gains are largest at the 90th percentile and above, consistent with AI unlocking the potential of especially promising ventures rather than modestly improving marginal ideas. Despite faster growth, treated firms do not scale inputs proportionally. Their demand for external capital investment falls by just over $220,000, a 39.5% decrease, relative to the control group, while their demand for labor remains unchanged. These results provide causal evidence that AI improves firm performance and productivity even at its current capabilities, and that discovering where and how to deploy AI is a key bottleneck in realizing the gains from this technology.

A Relational Theory Of Power Alternation

We study the spontaneous emergence of power alternation from the need for cooperation by developing a model in which two parties repeatedly cooperate and negotiate the position of power, defined as the control right of a productive regime. Unless the party in power, “incumbent,” compromises, the party out of power, “opposition,” would withdraw from cooperation. Central to our analysis are two impediments: the incumbent’s hold-up problem and information asymmetry. We establish a recursive structure of the model, taking into account the endogenous roles—incumbent or opposition—the two parties play. We find that alternation of power is necessary to sustain cooperation in the long run, while within-period compromise is essential for efficiency. In efficient self-enforcing agreements, incumbents always compromise minimally with oppositions insofar as to have cooperation sustained, while two norms endogenously emerge, prescribing an implicit bargaining protocol and the persistence of power. We characterize the implied history-dependent dynamics of political compromise and power alternation and illustrate the results in historical contexts.

Internal Monitoring and Collective Reputation

We study how internal monitoring and imperfect public signals together shape the value and sustainability of a collective reputation in organizations that operate across multiple markets. The organization comprises a global player—such as a franchisor—whose actions affect all markets, and market-specific local players—such as franchisees—whose behavior affects other markets only via reputational spillovers. In a repeated game with both imperfect public monitoring (noisy public signals of effort) and partial information (perfect monitoring of local effort by the global player known to the local player), we characterize the conditions under which it is beneficial for the global player to acquire and utilize internal monitoring technology. When internal monitoring is feasible, the global player can employ a contagion strategy, triggering organization-wide disciplinary actions in response to detected shirking by any local player. This mechanism fully eliminates local players’ incentives to free-ride on the collective reputation. However, sustaining such a strategy requires placing the global player under countervailing incentives to ensure the credibility of the contagion strategy, imposing additional costs on the organization. We show how the trade-off between these positive and negative effects determines whether internal monitoring, alongside imperfect public signals, enhances the durability of collective reputation.

Mini-Workshop on AI in HRM & Professional Services

The Impact of Cartels on Productivity

We study the impact of cartels on productivity using a unique plant-level dataset from the Japanese ready-mix concrete industry, where cartels can be legally permitted. Our annual panel covers 1993–2004 with information on inputs, outputs, cartel membership, and the timing of cartel formation and collapse. After estimating plant-level productivity, we implement a difference-in-differences analysis around these events. Results show that cartel collapse increases both plant-level and market-level productivity, while cartel formation has no effects. A decomposition indicates that improvements are driven by heightened competition following cartel collapse rather than by reallocation across plants.

Embracing the Enemy

Two agents repeatedly compete for the power to set policy. A principal partially influences the power allocation. All three players may disagree on policy, but one agent (the “friend”) aligns more closely with the principal than the other (the “enemy”). The principal’s optimal contract aims to exclude the enemy initially. However, once the enemy gains power, the principal embraces him, trading power for policy moderation. Moreover, the principal leverages the enemy’s moderation to move the friend’s policies toward her bliss point. If her commitment is strong enough, a principal offers more embrace to the enemy when her friend is close.

Nothing but the Truth? Private Information and Reporting on Corporate Social Responsibility

Firms routinely make unverifiable claims about their commitment to corporate social responsibility (CSR), leaving markets uncertain about which statements reflect genuine engagement and which amount to greenwashing. We develop a model in which firms first make unverifiable CSR claims, then may privately learn verifiable performance signals – such as greenhouse gas (GHG) emissions – that they can choose to disclose. Anticipating how markets will interpret future (non)disclosure disciplines firms’ initial reporting: when private signals are more likely to become available, the expected cost of being exposed as insincere outweighs the benefit of exaggeration. The model thus predicts that increases in the availability of private, engagement-correlated signals reduce the extent of greenwashing in reported CSR engagement. Exploiting a distinctive feature of the UK Companies Act of 2013, which “mandated” GHG disclosure only when firms deemed it “practical,” we interpret the reform as an exogenous increase in signal availability rather than strict mandatory transparency. Using difference-in-differences analyses comparing UK and other European firms, we find that environmental CSR reporting rose significantly less in the UK after the Act, consistent with the model’s predictions. Additional tests confirm that the disciplining effect is strongest among firms least likely to engage in CSR. Our findings show how even privately held information can curb greenwashing and foster more truthful CSR communication.

Countervailing Platform Power: Spotify and the Major Record Labels

While large digital platforms have attracted concern for possible exercise of power against consumers and small suppliers, they may have different effects when downstream from concentrated suppliers. Digital platforms can carry many suppliers’ products, test the products’ consumer appeal, and choose which products to promote, potentially shifting power from the suppliers to the platforms.  We study these mechanisms in the recorded music industry, where a few platforms have replaced fragmented radio stations and retailers.   We study Spotify’s use of playlists, utilities for testing and promoting music to consumers, using data covering 2017-2020. First, Spotify used their expanded playlist capacity to test – and discover – proportionately more independent songs to promote on their playlists. Second, at least relative to major–label playlists, Spotify-operated playlists promoted new independent songs more than was indicated by their subsequent success.  Third, placement on Spotify new-music playlists has a large causal impact on streams. The independent-label share of new-music promotion rose from 38 percent in late 2017 to 55 percent in early 2020, which helps to explain the reported decline in the share of Spotify royalty payments to major-label suppliers over the same period.

Charismatic Leadership: An Antidote to the Pitfalls of Incentives?

Past research has established that charismatic leadership tactics can be a powerful motivator. In some settings, the increase in work output induced by a charismatic speech is comparable in size to the positive effect of high-powered financial incentives. But what about settings in which incentives backfire? In a between-subject laboratory experiment, we set up a real-effort work environment in which participants can execute a task in two ways: they can either “work hard” so that each produced unit creates a sizable benefit for the principal, or they can “take shortcuts”, which takes much less effort but also substantially reduces the benefit of a produced unit for the principal. When compensation is a fixed wage and the motivation speech is “standard”, we observe that participants mostly focus on the socially optimal, hard version of the task, but the general effort level is not particularly high. Exposing participants to financial incentives motivates participants to raise the overall effort level substantially, but the revenue created for the principal decreases drastically. This counterproductive effect of performance pay is caused by the workers’ decision to concentrate almost exclusively on the inefficient, easy version of the task when incentivized. Combining the fixed wage with a charismatic motivation speech, in contrast, increases both the overall effort level and the revenue for the principal. The positive effect on the effort level is smaller than the one of incentives, but the charismatic speech induces workers to focus on the difficult version of the task. A combination of incentives and a charismatic speech leads to similar outcomes as using incentives alone. These results establish novel insights: On one hand, we show that charisma can be an effective motivation tool even in situations where incentives fail. On the other hand, however, charisma does not shield participants from the corrupting effects of incentives when the two tools are combined in our setting.

Social Anxiety and Evaluative Interviews

Managers and the Cultural Transmission of Gender Norms

This paper examines the influence of managers from countries with different gender norms on workplace culture and gender disparities within organizations. Using data from a multinational firm operating in over 100 countries, we exploit cross-country manager rotations to estimate the impact of male managers’ gender attitudes on gender pay gaps within a team. Managers from countries with one standard deviation more progressive gender attitudes narrow the pay gap by 5 percentage points (18%), primarily by promoting women at higher rates. The effects last beyond the manager’s rotation and are concentrated in countries with more conservative gender attitudes. Managers with progressive views appear to influence the local office culture, as local managers who interact with but are not under the purview of the foreign manager begin to have smaller pay gaps in their teams. Our evidence points to individual managers as critical in shaping corporate culture.

The Turing Valley: How AI Capabilities Shape Labor Income

Do improvements in Artificial Intelligence (AI) benefit workers? We study how AI capabilities influence labor income in a competitive economy where production requires multidimensional knowledge, and firms organize production by matching humans and AI-powered machines in hierarchies designed to use knowledge efficiently. We show that advancements in AI in dimensions where machines underperform humans decrease total labor income, while advancements in dimensions where machines outperform humans increase it. Hence, if AI initially underperforms humans in all dimensions and improves gradually, total labor income initially declines before rising. We also characterize the AI that maximizes labor income. When humans are sufficiently weak in all knowledge dimensions, labor income is maximized when AI is as good as possible in all dimensions. Otherwise, labor income is maximized when AI simultaneously performs as poorly as possible in the dimensions where humans are relatively strong and as well as possible in the dimensions where humans are relatively weak. Our results suggest that choosing the direction of AI development can create significant divisions between the interests of labor and capital.

Labor as Capital: AI and the Ownership of Expertise

Workplace surveillance generates data that can train AI systems to replicate worker expertise. Using a large online survey experiment of U.S. full-time workers, we show that workers adjust their knowledge contributions when made aware of this dynamic: they rationally withhold expertise due to career concerns. We formalize this behavior in a model of knowledge supply under surveillance-enabled AI and use it to evaluate alternative policies. Individual data ownership— workers’ preferred policy—eliminates knowledge withholding but creates negative externalities: one worker’s data strengthens the firm’s bargaining position against others, potentially making all workers worse off. In contrast, collective data ownership achieves the first-best outcome, promoting knowledge sharing while allowing workers to benefit from AI-driven productivity gains. These findings highlight the importance of labor agreements in shaping AI adoption in labor markets.

Norms at Work: Masculinity, Well-being and Performance in Academia

Workplaces across many industries are characterized by what is stereotypically called “masculine” norms: i.e. highly competitive and aggressive norms, often portrayed as necessary to increase performance. Using rich survey and archival data from faculty and staff in business schools, we develop a novel way to measure these hyper-competitive norms and show that they are negatively correlated with employee well-being, both increasing turnover intentions and reducing workplace well-being. We then examine why these norms persist despite their negative consequences and find that the associated lower well-being is not offset by higher performance – neither in terms of research quantity nor impact. Finally, we show that no organizational subgroup thrives in hyper-competitive environments. While neither men nor women benefit from such norms, even “superstar” performers in the top performance deciles experience negative implications.

Real-Time Monitoring and Relational Contracts in Usage-Based Insurance

The rise of the Internet of Things (IoT) and big data technologies enables insurance companies to monitor policyholders’ behavior in real time, leading to innovative usage-based insurance (UBI). This paper studies the optimal UBI contract that employs both a traditional objective signal (e.g., official accident report) and a novel subjective signal (e.g., driver safety score) about the insured’s behavior in the presence of moral hazard. We show that under limited liability, the subjective signal may not be used even when enforceable if it is relatively imprecise. Moreover, the objective and subjective signals can serve as either complements or substitutes, depending on their precisions. While a more precise subjective signal always enhances insurance market efficiency, the welfare implication of the objective signal can be non-monotonic. In particular, when a more precise objective signal leads to a highly efficient traditional insurance contract, it may reduce the efficiency of the UBI contract or even make the subjective signal infeasible to use. Our paper thus explains the conditions under which UBI programs can emerge and highlights key factors for the success of UBI programs. In addition, we show that UBI market regulation can mitigate distortions in UBI contract design and investment in monitoring technologies.

Neglect

This paper studies how problems are, or are not, solved in collaborations. We develop a dynamic model of neglect, which we define as the failure to solve problems even when doing so would benefit all members of a collaboration. Neglect arises because solving a problem requires revealing it, which has the unintended consequence of making others less optimistic about the future of the relationship. In equilibrium, neglect arises when the party who learns of a problem derives more value from the collaboration than others. We characterize equilibrium dynamics, show when and why neglect arises, study how communication and exit convey information, and consider how collaborations can be structured to encourage the revelation of problems as they emerge.

Narrative Entanglement: The Case of Climate Policy

Political economy models often assume that voter beliefs are consistent with available information. Recent work emphasizes instead the role played by narratives, subjective causal models that may be incorrectly specified. In this paper, we study the role of political narratives in the context of climate policy. We develop a theory of narrative entanglement, where policy dimensions—initially distinct—become strategically intertwined through narratives created by politicians to sway support. Shocks in one dimension can thus influence unrelated policy areas. We test this theory in the context of EU climate policy before versus after Russia’s invasion of Ukraine, which affected the economic costs of climate policy but not its ability to address climate change. Using a large language model to analyze speeches in the EU Parliament, we find that narratives are strongly entangled: Members of the European Parliament that emphasize the need to address climate change also emphasize economic benefits, while those denying climate change stress economic costs. After the energy price shock associated with the invasion of Ukraine by Russia, narratives shift not only in the economic dimension but also in the climate dimension, with speeches becoming less likely to imply that climate policy is necessary to combat climate change. This pattern holds at the individual politician level, with politicians from right-wing parties showing a more pronounced narrative change than those from the left.

Mini-Workshop on Organizational Economics