This paper develops a formal identification framework for structural learning models, commonly estimated in economics. We show that learning models can be formally identified under mild conditions. In particular, we derive the key conditions for identification. These conditions can plausibly be satisfied with a few periods of data. We later demonstrate how these conditions can be operationalized in practice, with a series of numerical examples. Finally, we highlight that some of these derived conditions are qualitatively consistent with heuristic identification arguments used in past research.

Firms increasingly buy forecasts from the same analytics platforms. Pooling data sharpens each forecast, but routing decisions through one layer replaces idiosyncratic error, which averages out, with shared error, which does not. Finitely many firms screen candidate markets, each choosing proprietary research or a shared platform. Adoption raises a firm's own precision, which it internalizes, and correlates evaluation errors, which it does not. When a good market's value is realized if some firm recognizes it, this correlation lowers the value of independent evaluation, so private adoption exceeds the screening optimum whenever the shared error is present. Whether adoption is excessive on net turns on the precision adoption buys. Provider multiplicity diversifies the exposure only in proportion to providers' structural independence.

This paper formally establishes the identification properties for counterfactuals in discrete choice models where individuals face uncertainty. Using a flexible nonparametric discrete choice model, we demonstrate that counterfactuals for utility with affine transformations in general are not identified. In particular, we show that counterfactuals involving utility rescaling transformations are not identified, while counterfactuals involving additive constants and type changes are identified. Moreover, for a forward-looking agent, the counterfactuals involving changes in the model dynamics that impact the learning process are also identified. Taken together, this research offers practical guidance about what types of counterfactuals are and are not nonparametrically identified.

How Online Grocery Shopping is Changing What We Buy

This research documents the potential information signal value of organizational form decisions (i.e., franchise versus company-owned) made by retail chains. As expansion via company-owned stores requires more upfront investment, on the part of the retailer, we hypothesize that franchisee-based expansion signals the retailer's unwillingness to put "skin in the game." Using data about retail establishment entry, we confirm via descriptive analysis that a large proportion of franchisee stores, relative to company-owned stores, is associated with dampened entry of new establishments.

This paper presents a novel decomposition approach for measuring deterrence motives in dynamic oligopoly games. Our approach yields a formalized, scale-free, and interpretable measure of deterrence motives that informs researchers about the proportion for which deterrence motives account of all entry motives. In addition, the decomposition leads to a set of conditions for counterfactual analysis where hypothetical scenarios with deterrence motives eliminated can be explored. We illustrate the use of our measure and counterfactual by conducting an empirical case study about the dynamics of coffee chain stores in Toronto, Canada. The inferred deterrence motives suggest that a noticeable proportion of entry motives can be attributed to deterrence; it can be as high as 43% for the increasingly dominant coffee chain, Starbucks, in certain types of markets. Finally, counterfactual analysis confirms that deterrence motives are indeed associated with Starbucks’ aggressive presence as the number of its outlets and its market share are markedly lower once these motives are eliminated. This paper was accepted by Matthew Shum, marketing. Supplemental Material: The data files and online appendix are available at https://doi.org/10.1287/mnsc.2023.4864 .

Although local market conditions are normally considered to be the key drivers of forward-looking and strategic retail chain expansion, broader macroeconomic conditions such as business cycles might also be of relevance. This paper provides a new partially observable stochastic games framework for modeling and estimating dynamic investment games that accommodates for business cycles. In particular, firms in our model make strategic investment decisions based on beliefs they have about the business cycle. These beliefs evolve over time as new macroeconomic data becomes available. We develop identification results, as well as a likelihood-based estimation methodology. With this new framework, we provide an empirical case study about fast food industry dynamics. Analysis of an estimated model reveals that some of the retail chains are particularly resilient to recessions. Furthermore, we establish a connection between heightened volatility in the partially observable macroeconomic conditions and dampened retail investment. Finally, we show that this relationship between volatility and investment also has the potential to help preserve the market power of already dominant firms.

This paper investigates the interaction between tactics commonly used in platform competition (e.g., digital marketplaces), namely exclusive dealing and loyalty programs. Exclusive dealing has the potential to leverage brand sellers to attract brand-loyal users, while loyalty programs can foster allegiance directly towards the platform. Our findings suggest these tactics predominantly act as complements in platform competition. Moreover, we show that while exclusive dealing can enhance profitability, it can also reduce consumer surplus. Finally, we demonstrate the existence of a pure Nash equilibrium whereby competing platforms choose symmetric tactics for exclusive dealing contracts (i.e., both or neither use exclusivity).

We begin paper with a descriptive analysis of the growing role of modular innovation across several industries that uncovers regularities of US patent dynamics. Based on this, we develop a set of game-theoretic models for studying the impact of strategic modular innovation investments in technology on subsequent product market competition. We derive conditions for which technologies are shared with competitors via licensing, but only partially, given the modular nature of innovation. Finally, our analysis suggests that modularization might introduce new avenues for economic benefits via increased innovation activity and welfare.

We investigate the impact that uniform "one-size-fits-all" tax policies have on shaping the retail landscape. An entry model demonstrates that tax effects on retailer entry can amplify market concentration, such that retailers with preexisting brand power advantages are disproportionately responsive to tax policies that impact the sunk costs of entry. Using comprehensive data about all retail establishments in the United States from 1990 to 2014, we show that while they are more likely to open in markets with favorable state tax policies, these markets become more concentrated as entry is dominated by the largest chains. Therefore, tax policies might impact the retail industry competitiveness, despite the policies' uniform design.

This paper examines how differences in strategic sophistication shape firm entry dynamics. Using a cognitive hierarchy approach, I show that limited strategic reasoning can serve as a credible commitment mechanism, enabling unsophisticated firms to deter sophisticated rivals without costly financial commitments. This challenges the view that strategic sophistication always confers an advantage, demonstrating that cognitive constraints, rather than sunk costs, shape competitive expectations. Finally, I demonstrate that unsophisticated firms over-enter while sophisticated firms under-enter, revealing systematic distortions in entry probabilities.

This study offers an empirical investigation of inventory and sales dynamics in a large‐scale retail network setting. We infer the impact of product shortages on sales in neighboring outlets using unique data from a large fast fashion retailing chain and an Instrumented Difference‐in‐Differences (DDIV) methodology. Our analysis reveals that sales for a particular item at a focal store increases when that same item experiences stock‐outs in neighboring stores. Our empirical findings suggest that there is substitutability across stores, and that this substitutability is the strongest in the period when the stock‐out is observed for the first time, and decreases as time passes following the stock‐out. In order to assess the implications of considering the impact of stock‐outs on inventory allocation, we develop an optimization model that is calibrated using parameters estimated via our earlier DDIV analysis. The simulation analysis confirms that revenues markedly improve on average by 6% under low demand variance and by 14% under high demand variance when neighboring stock‐out information is taken into account for sales forecasting while optimizing initial inventory allocations. Finally, we conduct sensitivity analysis to evaluate how these potential revenue improvements vary with turnover, product price, and inventory.

This research explores the justification and implications of incorporating consumption variety into mobile-based food recommendation systems. Our study makes use of data from a popular mobile fitness app, in which we can observe large volumes of daily food logs of thousands of users. We first confirm that consumption variety is associated with lower overall calories consumed, higher vegetable consumption, and lower snack consumption. Motivated by these empirical patterns about variety and eating, we then seek out to design a novel multicriteria food recommendation system (FOODVAR) that can accommodate for variety in recommended foods. We then assess the impact of including this additional variety criterion in recommendation system performance, where we show that the incorporation of variety improves the algorithm’s evaluation metrics.

We study the impact of premium adoption on mHealth user engagement and its weight loss effectiveness. Our empirical analysis uses unique panel data from a popular mobile food and exercise tracking app. To estimate the causal effect of premium adoption we use a propensity score matching method. The analysis reveals that premium adoption is linked to elevated engagement levels for food and exercise calorie tracking, daily goal achievement, exercise calories, as well as weight loss. Furthermore, these effects on engagement levels from premium adoption dampen quickly over time, but the weight loss effectiveness of the premium version is relatively persistent. Taken together, the dampening patterns of engagement appear to be more consistent with behavioral mechanisms that relate to sunk costs, rather than hedonic decline, moral licensing, self-control, and goal pursuit. Finally, we demonstrate that failure to account for endogeneity and selection will lead to an overstatement of these effects on user engagement and weight loss outcome, and discuss managerial implications of our findings.