Publications

In 2015, Yara International faced simultaneous financial, legal, and reputational pressures as industry conditions deteriorated. Newly appointed CEO Svein Tore Holsether concluded that restoring performance required more than cost-cutting and reputation repair, and instead moved to future-proof the firm by embedding sustainability into Yara’s purpose, strategy, and operating model. This case examines the rationale, execution, and 2025 outcomes of Yara’s purpose-led transformation amid evolving market and societal constraints.

The design of algorithms for political redistricting generally takes one of two approaches: optimize an objec- tive such as compactness or, drawing on fair division, construct a protocol whose outcomes guarantee partisan fairness. We aim to have the best of both worlds by optimizing an objective subject to a binary fairness constraint. As the fairness constraint we adopt the geometric target, which requires the number of seats won by each party to be at least the average (rounded down) of its outcomes under its worst and best possible partitions of the state.

We document an understudied form of discrimination based on femininity expressed in headshots. To do so, we use a Generative Adversarial Network to create realistic headshots of people and then manipulate their femininity independently of other attributes like pose, general facial expression, background, hairstyle, and clothes. Then, we devise an experimental design that allows us to identify the separate and combined causal impact of femininity and gender identity (proxied by gender pronouns) on real-life outcomes. In a field experiment within a naturalistic advertising environment, we find that prospective customers of an education company discriminated against femininity at various stages of the purchase funnel, to a largely similar extent for men, women, and non-binary people. The findings inform managers and policymakers of an important form of discrimination that would otherwise be underestimated by disregarding headshots. Methodologically, we introduce a novel framework for testing specific hypotheses pertaining to causal effects of treatments which are derived from unstructured data. The approach involves using natural text to readily identify hypothesis-relevant features from a generative AI model (in a particularly disentangled and interpretable representation space) and then creating realistic stimuli that vary controllably in those features.

Freemium upgrade is a primary monetization lever for mHealth apps, yet whether paying for premium features actually sustains user engagement remains an open question. We study this question using large-scale data from a fitness-tracking mobile app and a staggered difference-in-differences design. Premium adoption generates an immediate increase in food and exercise tracking, caloric budget adherence, and exercise calories, but these effects attenuate within several weeks and do not translate into sustained weight loss. Sensitivity analysis and matched-sample designs confirm that the engagement responses are not solely driven by time-varying selection. Heterogeneous engagement lifts by pre-upgrade exposure are more consistent with hedonic decline than with sunk-cost effects or motivational mean reversion: users with limited prior exposure to the free version exhibit substantially larger post-upgrade engagement lifts than those who have already interacted extensively with it. We further demonstrate that failing to account for endogeneity and selection in premium adoption leads to a substantial overstatement of its effects on both engagement and health outcomes.

The Securities and Exchange Commission (SEC) reviews firms’ financial reports and issues comment letters to ensure compliance with applicable disclosure and accounting requirements. We explore the nature, determinants, and consequences of SEC comment letters that refer to information disclosed in voluntary earnings conference calls. Using hand-collected data, we document that the SEC primarily references these voluntary disclosures to illustrate insufficiencies and, less commonly, inconsistencies in mandatory filings across a wide range of topics. These letters are more likely to be issued when filing reviews are more complex, SEC staff are less resource constrained, and for firms with more institutional investors and analysts. Conference call-related comments tend to occur during higher-quality review processes and require greater remediation costs than other comments. The SEC’s use of call disclosures also leads to more pronounced changes in firms’ subsequent mandatory filings, particularly when the firm indicates agreement with SEC comments. However, we observe a mixed effect on the overall information environment, consistent with possible unintended consequences for the quality of firms’ voluntary disclosures.

This paper studies how local competition affects pricing and product assortment decisions. We leverage a unique circumstance in the context of the legalized cannabis industry in Washington State, where retail licenses were allocated via a lottery, which generates quasi-random variation in the number and proximity of competitors. This natural experiment helps address the endogeneity issues commonly encountered in causal analyses of competitive effects. The analyses yield three key findings: first, additional nearby competitors reduce markups, with nearly all of the effect concentrated in the first two competitors. Second, retailers facing more nearby competition differentiate their assortments. Third, this differentiation helps mitigate the intensity of price competition among retailers. In a partial equilibrium simulation, we evaluate a 10% increase in the number of licensed retailers and estimate that it would reduce markups and generate approximately $7.5 million in annual consumer savings. These findings shed light on the nature of localized competition and the implications of entry restrictions in regulated retail markets.

A seller wants to sell an item to n buyers. Buyer valuations are drawn i.i.d. from a distribution unknown to the seller; the seller only knows that the support is included in [a,b]. To be robust, the seller chooses a DSIC mechanism that optimizes the worst-case performance relative to the ideal expected revenue the seller could have collected with knowledge of buyers' valuations. Our analysis unifies the regret and the ratio objectives. 
For these objectives, we derive an optimal mechanism and the corresponding performance in quasi-closed form, as a function of the support information [a,b] and the number of buyers n. Our analysis reveals three regimes of support information and a new class of robust mechanisms. i.) When a/b is below a threshold, the optimal mechanism is a second-price auction (SPA) with random reserve, a focal class in earlier literature. ii.) When a/b is above another threshold, SPAs are strictly suboptimal, and an optimal mechanism belongs to a class of mechanisms we introduce, which we call pooling auctions (POOL); whenever the highest value is above a threshold, the mechanism still allocates to the highest bidder, but otherwise the mechanism allocates to a uniformly random buyer, i.e., pools low types. iii.) When a/b is between two thresholds, a randomization between SPA and POOL is optimal. 
We also characterize optimal mechanisms within nested central subclasses of mechanisms: standard mechanisms that only allocate to the highest bidder, SPA with random reserve, and SPA with no reserve. We show strict separations in terms of performance across classes, implying that deviating from standard mechanisms is necessary for robustness.

In most evaluation systems—such as those governing the allocation of prestigious awards—the evaluator’s primary task is to reward the highest quality candidates. However, these systems are imperfect; top performers may not be acknowledged and thus be underrecognized, and low performers may receive unwarranted recognition and thus be overrecognized. An important feature of many evaluation systems is that people alternate between being candidates and being evaluators. How does experiencing misrecognition as a candidate affect how people subsequently evaluate others? We develop novel theory that underrecognition and overrecognition lead people to reproduce those experiences when they are evaluators. Across three studies—a quasi-natural experiment and two preregistered, multistage experiments, we find that underrecognized evaluators are less likely to grant recognition to others—even to the highest-performing candidates. Conversely, overrecognized evaluators are more likely to grant rewards to others—even to the lowest-performing candidates. Whereas underrecognized evaluator behavior is driven by individuals’ perceptions that their experience was unfair, overrecognized evaluator behavior is driven by the informational cues people glean on how to evaluate others. Thus, in evaluation processes where people oscillate between being the evaluated and being the evaluator, we show how and why seemingly innocuous initial inefficiencies are reproduced in subsequent evaluations.

Should central banks’ inflation targets remain set in stone? We study a dynamic mechanism
design problem between a government (principal) and a central bank (agent). The central
bank has persistent private information about structural shocks. Firms learn the state from the
central bank’s reports and form inflation expectations. A dynamic inflation target implements the
full-information commitment allocation. The central bank is delegated the authority to adjust
the level and flexibility of its target as long as it does so one period in advance. All history
dependence of the mechanism is summarized by the current period’s target. We show that
a declining natural interest rate and a flattening Phillips curve imply opposite optimal target
adjustments. We leverage our framework to study longer-horizon time consistency problems
and speak to practical policy questions of inflation target design.

A strategic tension between consumers in a monopolistic market arises when many high-value consumers want to pool with a few lower-value consumers in order to obtain low prices from the seller. We study the interaction between consumers and the resulting market segmentation into consumer groups as the outcome of a cooperative game between the consumers. We introduce two new solution concepts, the weakened core and stability, which coincide with the core whenever it is nonempty. We show that these concepts are in fact equivalent and non-empty, and are characterized by efficiency and saturation. A segmentation is saturated if shifting consumers from a segment with a higher price to a segment with a lower price leads the seller to optimally increase the lower price. We show that stable segmentations that maximizes average consumer surplus (across all segmentations) always exist

This paper develops a theory-based, explainable deep learning convolutional neural network (CNN) classifier to predict the time-varying emotional response to music. We design novel CNN filters that leverage the frequency harmonics structure from acoustic physics known to impact the perception of musical features. Our theory-based model is more parsimonious, but it provides comparable predictive performance with atheoretical deep learning models while performing better than models using handcrafted features. Our model can be complemented with handcrafted features, but the performance improvement is marginal. Importantly, the harmonics-based structure placed on the CNN filters provides better explainability for how the model predicts emotional response (valence and arousal) because emotion is closely related to consonance—a perceptual feature defined by the alignment of harmonics. Finally, we illustrate the utility of our model with an application involving digital advertising. Motivated by YouTube’s midroll ads, we conduct a laboratory experiment in which we exogenously insert ads at different times within videos. We find that ads placed in emotionally similar contexts increase ad engagement (lower skip rates and higher brand recall rates). Ad insertion based on emotional similarity metrics predicted by our theory-based, explainable model produces comparable or better engagement relative to atheoretical models.

This article concerns itself with fees that Apple and Google might charge to business users in their respective mobile ecosystems. We lay out the economic analysis behind the goals of the DMA—contestability and fairness—as they apply to third-party app store access fees. We focus on the access fees for alternatives to the Apple App Store, as this has become contentious in the early enforcement of the DMA. Much of our analysis, however, also applies also to Google and/or any other designated gatekeeper.

For more than 150 years, the founding mission of the American Society for the Prevention of Cruelty to Animals (ASPCA) guided the organization while enabling it to address evolving challenges in animal welfare. Under Matt Bershadker, ASPCA’s president and CEO, the company faced mounting pressure to engage with a growing stream of societal matters far afield from its core purpose. The case explores Bershadker’s initiative to develop a clear, strategic framework for considering which societal issues to address. This effort would clarify the ASPCA’s approach to these issues by evaluating them against the organization’s strategy, history, policies, and key stakeholder relationships, ensuring consistency, transparency, and mission-driven decision-making.

We study a retirement savings plan with a default contribution rate of 12% of income, which is much higher than previously studied defaults. Twenty-five percent of employees had not opted out of this default 12 months after hire; a literature review finds that the corresponding fraction in plans with lower defaults is approximately one-half. Because only contributions above 12% were matched by the employer, 12% was likely to be a suboptimal contribution rate for employees. Employees who remained at the 12% default contribution rate had average income that was approximately one-third lower than would be predicted from the relationship between salaries and contribution rates among employees who were not at 12%. Defaults may influence low-income employees more strongly in part because these employees face higher psychological barriers to active decision making.

Bayer AG is a German multinational pharmaceutical and life sciences company founded in 1863. It operates globally in over 90 countries with approximately 100,000 employees as of 2024. Bayer is structured into three main business segments: Pharmaceuticals, Consumer Health, and Crop Science, with significant global operations and an extensive patent portfolio. The company is a leader in agricultural products, prescription medicines, and over-the-counter health products. It also focuses on innovative solutions for healthcare and agricultural challenges.

The current dilemma for students to address are issues related to Bayer’s transition to a new operating model called Dynamic Shared Ownership (DSO). This model, introduced under the leadership of CEO Bill Anderson, aims to flatten the corporate hierarchy and create a nimbler, customer-centric organization. The transition involves removing the existing hierarchical structure, which previously consisted of 12 management layers, and replacing it with self-managed teams. This structural change was referred to internally as the "hardware."

The initial steps under DSO included substantial organizational redesign, involving layoffs and the establishment of self-managed teams. The Board of Management successfully halved the number of management layers to five or six and replaced thousands of middle managers with self-managed teams. A complementary aspect of DSO, labeled the "software," focused on fostering cultural changes to promote new mindsets, norms, and behaviors among Bayer’s nearly 100,000 employees.

Notwithstanding several early wins, the Board of Management is now grappling with implementing new talent management and personnel policies that align with the DSO model. Existing HR processes and systems, which are designed for a hierarchical organization, need to be reinvented. Questions about compensation, career pathing, and performance metrics must be addressed to ensure these new systems support the DSO framework effectively.

A critical element of word of mouth (WOM) or buzz marketing is to identify seeds, often central actors with high degree in the social network. Seed identification typically requires data on the relevant network structure, which is often unavailable. We examine the impact of WOM seeding strategies motivated by the friendship paradox, which can obtain more central nodes without knowing network structure. Higher degree nodes may be less effective as seeds if these nodes communicate less with neighbors or are less persuasive when they communicate; therefore, whether friendship paradox–motivated seeding strategies increase or reduce WOM and adoption remains an empirical question. We develop and estimate a model of WOM and adoption using data on microfinance adoption across village social networks in India. Counterfactuals show that the proposed strategies with limited seeds are about 13%–30% more effective in increasing adoption relative to random seeding. These strategies are also on average 5%–11% more effective than the firm’s leader seeding strategy. We also find these strategies are relatively more effective when we have fewer seeds.

We hypothesize that “low-contrast categories” (those lacking sharp differentiation from adjacent categories) catalyze the creation of groundbreaking inventions by influencing two key stages in the life of an invention: (1) idea-creation and (2) idea-positioning. During “idea-creation,” low-contrast categories increase the likelihood that descendant inventions will combine the focal invention with more (a) boundary-spanning, (b) novel, (c) original, and (d) atypical knowledge inputs. During “idea-positioning,” they allow greater leeway in articulating how descendant inventions depart from the focal invention’s lineage and chart new technological directions. We find robust support for our hypothesis using data from the United States Patent and Trademark Office’s classification system spanning nearly four decades. Further analyses demonstrate that the catalyzing effect of low-contrast categories has important material consequences: inventions classified in low-contrast categories spur descendant inventions that generate substantially higher economic value and exert more enduring technological impact than those in high-contrast categories. By introducing the concept of catalyzing categories, this study offers a novel theoretical perspective on the genesis of groundbreaking inventions and the role of categorical structures in the inventive process.

The 2020 revisions to the Federal Reserve’s monetary policy framework included a shift in the Fed’s policy focus to shortfalls (rather than deviations) from maximum employment and a commitment to “flexible average inflation targeting.” The new framework, and the associated guidance and asset purchases with which it was implemented, were tested by the surge in inflation in 2021 and 2022. We consider the lessons learned from this experience. We conclude that the changes to the framework were too focused on the experience following the financial crisis and hence were not robust in the face of unexpected changes in economic circumstances. We also argue that the Fed made mistakes with the calibration and communication of the tools used to implement the framework—the forward guidance on the policy rate and the asset purchase program. We recommend a broad framework that would be appropriate in a wide range of policy environments, with the specific policy approach to be taken in any given circumstance to be communicated through forward guidance and asset purchase announcements. We suggest ways in which the Fed could implement these tools with better calibration and communication, in order to avoid having its policy commitments exacerbate costly economic outcomes.