Recipient Organization
PURDUE UNIVERSITY
(N/A)
WEST LAFAYETTE,IN 47907
Performing Department
Ag Economics
Non Technical Summary
This proposal is a Research Project for the Economics, Markets, and Trade (EMT) program within the Agricultural Economics and Rural Communities priority area under the AFRI Foundational Program. This proposal responds to the EMT stated priority of "development of theories, methods and applications of economics...in the areas of...agricultural market structure and performance; agricultural policy design and impacts; technology development and adoption." This project aims at developing theory and methods that will permit ex-ante and ex-post economic evaluation of policies in the agricultural and food sectors; especially those that can alter the spectrum of product variety accessible to consumers.Reliable measurement of welfare implications of food policies requires unbiased estimation of structural parameters. Unbiased estimation of structural parameters in strategic settings (increasingly common in food markets due to product differentiation) hinges upon correct specification of the mode of competition (i.e. players' strategies and the resulting equilibrium concept). Yet, little theoretical guidance and no empirical testing is available to understand and predict how and when specific modes of competition emerge in equilibrium or to guide econometric estimation. We propose to address this important gap through the following steps. First, we will generate theoretical predictions regarding modes of competition by integrating fragmented strands of the theory of oligopoly intermediation. Second, we will conduct experiments based on the theory. Since experiments allow us to control the game form and mode of competition, specify the economic primitives, and observe the strategies used by the players, we will have a much deeper knowledge of the data generating process than what is typically available to the econometrician using observational data. Third, we will conduct structural econometric analysis of the experimental data. We will do this by treating experimental data as if it were observational (i.e. we will assume structural parameters are unobservable to the researcher) and imposing commonly invoked assumptions in the structural econometrics literature about the nature of the game played, the equilibrium concept, and the underlying economic primitives. We can then compare the results to the results that are generated under the correct assumptions about the data generating process to evaluate biases arising from miss-specification. We can also identify which assumptions are robust to misspecification. Finally, we will identify the link between estimation bias and observable primitives, strategies, and mode of competition. We expect that these tasks, in combination, will result in guiding principles that inform structural econometric modeling underpinning welfare analysis of modern food policies.
Animal Health Component
70%
Research Effort Categories
Basic
30%
Applied
70%
Developmental
(N/A)
Goals / Objectives
This project aims at developing theory and methods that will enhance ex-ante and ex-post economic evaluation of policies in the agricultural and food sectors. Reliable measurement of welfare implications of food policies requires unbiased estimation of structural parameters (technology and preferences). As food marketing channels have evolved away from commoditized goods and towards sophisticated product differentiation, firms in agricultural and food markets operate in increasingly strategic environments (Myers, Sexton and Tomek 2010). Unbiased estimation of structural parameters in strategic settings hinges upon correct specification of the mode of competition (i.e. players' strategies and the resulting equilibrium concept). Yet, little theoretical guidance and no empirical testing is available to understand and predict how and when specific modes of competition emerge in equilibrium or to guide econometric estimation.We propose to address this important gap through the following steps. First, we will generate theoretical predictions regarding modes of competition by integrating fragmented strands of the theory of oligopoly intermediation. Second, we will conduct experiments based on the theory. Since experiments allow us to control the game form and mode of competition, specify the economic primitives, and observe the strategies used by the players, we will have a much deeper knowledge of the data generating process than what is typically available to the econometrician using observational data. Third, we will conduct structural econometric analysis of the experimental data. We will do this by treating experimental data as if it were observational (i.e. we will assume structural parameters are unobservable to the researcher) and imposing commonly invoked assumptions in the structural econometrics literature about the nature of the game played, the equilibrium concept, and the underlying economic primitives. We can then compare the results to the results that are generated under the correct assumptions about the data generating process to evaluate biases arising from miss-specification. We can also identify which assumptions are robust to misspecification. Finally, we will identify the link between estimation bias and observable primitives, strategies, and mode of competition. We expect that these tasks, in combination, will result in guiding principles that inform structural econometric modeling underpinning welfare analysis of modern food policies.The principles developed in this project will be general enough to inform welfare analysis of any type of policy. However they may be particularly relevant for policies that have the potential to alter the spectrum of product varieties available to consumers or their relative prices, a defining feature of many modern food policies. The most prominent examples of such policies include: 1) restrictions on marketing of foods containing certain ingredients such as sugar or transfat (Wilson, Stolarz-Fantino, and Fantino, 2013) or potential contaminants (Henson and Caswell, 1999); 2) labeling and certification of product attributes including, but not limited to, organic production (Lohr, 1998), geographical origin such as country-of-origin (e.g. Plastina, Giannakas, and Pick, 2012) and "buy local" (Lim and Hu, 2016), genetically modified foods (Park, 2014), fair trade (Dragusanu et al., 2014), animal welfare (Kehlbacher and Balcombe, 2012), and genetic animal traits (Thompson et al., 2016); and 3) policies taxing/subsidizing products based on specific attributes (e.g. the so called "fat taxes"). These, and more broadly policies that seek to empower consumers and make them more aware of the health and environmental effects of various food choices (a movement known as "food democracy" and discussed by Norwood, 2015 and McFadden and Stefanou, 2016, among others) share common patterns, namely they have the capacity to affect properties of the variety spectrum.Policies, and especially those causing alterations in product variety and/or their relative profitability, may trigger complex behavioral responses by firms. Recent theoretical advances in oligopoly intermediation (Hamilton et al. 2015) suggest that changes in the space of product characteristics have the potential to affect the mode of competition itself. But the ways in which policies affect firm behavior and market equilibrium in this context remain obscure in current debates; as do the economic implications of such responses throughout the food marketing channels. Such dearth of information is partly explained by gaps in the theory of competition when the mode of competition and product variety/quality are both under the control of the firm. In particular, models that characterize the choice of the mode of competition keep the choice of variety exogenous (e.g. Hamilton et al., 2015), and models that examine the choice of variety hold the mode of competition exogenous (e.g. Richards and Hamilton, 2015; Gentzkow et al., 2014).But while theoretical predictions are a necessary starting point for our analysis, solution concepts from game theory are notorious for imposing (sometimes excessively) strong assumptions regarding players' cognitive abilities, preferences, and conjectures regarding other player's behavior. Violations of such assumptions may introduce errors in theoretical predictions. Therefore empirical testing and subsequent refinement of theoretical solution concepts can greatly enhance the predictive performance of the proposed models. Unfortunately, such empirical refinements are often not possible with observational data as simultaneous identification of structural parameters and the mode of competition, is typically infeasible. Therefore we propose to use cutting edge experimental economics tools to empirically test and refine theoretical developments. Multiple structural econometric models, differing by the mode of competition assumed, will then be fitted to data generated by the experiments, and compared based on their ability to minimize bias in estimation of structural parameters.Our long term objective is to advance our understanding of the economic impacts of current and prospective food policies with a particular emphasis on those that have the potential to alter product variety, in addition to prices, in food markets. Attainment of our long term goal rests on five key supporting objectives: 1) extend the model of oligopolistic competition to consider circumstances in which the firms can choose both product variety/quality and the mode of competition; 2) use the extended model to generate testable predictions of firms' responses to policy, as a function of observable primitives; 3) conduct lab experiments to test the validity of theoretical predictions and identify behavioral anomalies; 4) characterize biases resulting from assumptions on the mode of competition, commonly held in structural estimation of oligopoly games; and 5) use identified patterns in biases to develop protocols that can alleviate them, using information on observable primitives.
Project Methods
Since our overall goal is to enhance our understanding of biases in welfare analysis of policy due to miss-specification of the mode of competition, we will illustrate our approach by conducting welfare analysis of a policy intervention. In particular we will consider a policy that has an effect common to most modern policies in the food sector; that of altering the spectrum of varieties available to consumers.In line with our first supporting objective, we will extend the model by Hamilton et al. (2015) to endogenize the choice of variety and, hence, rivalry. The model extension has been detailed in the submitted proposal.The extended model will be used to generate testable predictions regarding the effect of policies. We illustrate our approach by considering policies that affect the variety spectrum, and that were recently subject of intense popular debate. It should be noted, however, that we choose to examine policies affecting the variety spectrum because they provide the richest setting to evaluate biases from miss-specification of the mode of competition. But our framework is suitable for other, simpler, policies that change the relative profitability of varieties, rather than the variety spectrum itself. We propose to develop a simple model that can generate clean comparative statics characterizing the effect of policies on behavior and equilibrium.Lab experiments allow the researcher strict control over the structure of the game and its perturbations. Inspired by Binmore et al. (2002) and subsequent literature, we propose to conduct experiments (Supporting Objective 3) of sequential games with complete information to test theoretical predictions generated in Supporting Objective 2, based on the model developed in Supporting Objective 1. The specific procedures for conducting experiments will follow standard experimental economics protocol (e.g. Friedman and Sunder 1994). We will design an artificial economy that simulates, as close as possible, the theoretical model. For that purpose, we will build on co-PI Wu's experience using a software called Z-tree. Z-tree is a widely used program that allows human subjects to conduct transactions through computer terminals. Human subjects will be recruited using protocols approved by the Purdue Internal Review Board (IRB). Subjects are explained that will make simple transactions within the rules established by the researcher. Subjects will be paid for such transactions so as to reduce to hypothetical nature of the experiment (Friedman and Sunder 1994).Transactions are made under different sets of rules. Each set of rules is called a "treatment". Subjects are randomly assigned to treatments, and within treatments randomly assigned to roles (farmer, intermediary, or consumer). Consumers are randomly assigned willingness to pay values for each product variety, and they can then choose what variety (if any) to purchase. Intermediaries are randomly assigned production costs for each variety, and they are informed of functions describing diversion ratios and relative rivalries. Intermediaries then choose what variety to offer, what mode of competition to follow, and input and output prices. Finally, farmers are randomly assigned production cost parameters for input varieties, and they can then decide what farm product to produce.Separate treatments are created to evaluate the impact of a generic shock, designed to capture key features of modern food policies. As an example, consider a policy that expands the variety spectrum through certification and labeling of a product attribute. We can create a treatment corresponding to the original variety spectrum and a separate treatment where transactions are based on a variety spectrum modified by policy. The latter treatment is created to examine behavioral responses and equilibrium displacement triggered by such policy. A key contribution of this proposal is the recognition and formalization of the idea that firms may respond to such policies not only by adjusting their choice of variety, but also by changing the mode of competition in a manner consistent with the new equilibrium in the variety space. Therefore we will create separate treatments to tease out implications of this conceptual addition on policy analysis. Subjects will be forced to make decisions under three different sets of rules governing the the mode of competition. First, we exogenously impose a PTS mode of competition. Second, we impose a PTO mode of competition. Finally, we allow for endogenous choice of the mode of competition. The first two situations, in effect, eliminate the second stage of the game described above. The combination of treatments is presented in the following table:Comparisons between theoretical and experimental results by treatment reveal behavioral anomalies; i.e. empirical deviations from theoretical predictions. The model developed in Supporting Objective 1 will provide a theoretical baseline in which markets achieve an equilibrium under a given variety spectrum. Completion of Supporting Objective 2 will generate a theoretical policy-induced counterfactual equilibrium, which when contrasted with the theoretical baseline without policy, will reveal predictions regarding policy effects. We will also generate an experimental baseline without policy and an experimental counterfactual with policy. Three pair-wise comparisons can be made under each treatment:Theoretical prediction under baseline vs. experimental results under baseline,Theoretical prediction under policy counterfactual vs. experimental results under policy counterfactual,Theoretical prediction of policy effect (difference between theoretical baseline and theoretical policy counterfactual) vs. experimental policy effect (difference between experimental baseline and experimental policy counterfactual).Comparison of results across treatments reveals errors in prediction resulting from ignoring the endogenous nature of the mode of competition. For instance, pair-wise comparison 1 under treatment 1 and pairwise comparison 3 under treatment 3, may indicate that theoretical predictions are robust in the lab. Comparisons can be made then between treatments 1 and 3 to calculate a total prediction error and decompose it into inappropriate theoretical premises, and empirical deviations from theoretical predictions.Finally, we conduct econometric analysis of experimental data to estimate biases introduced by commonly held assumptions on the mode of competition. Our econometric estimations will assume structural parameters are unobservable to the researcher and, essentially, interpret experimental data as researchers would interpret observational data. To characterize potential biases in the extant literature, we employ an econometric approach widely used in the empirical literature on games of intermediation with differentiated products (e.g. Bajari et al. 2010). The empirical workhorse of this literature has been surveyed by Berry and Tamer (2006), Bajari, Hong, and Nekipelov (2013), and De Paula (2013).