Research

Distinguishing between Collusion and Competition Using BLP’s Random Coefficients Logit Model (Job Market Paper)

Abstract: There is a long tradition of economists trying to determine the extent to which firms collude in an industry. Their approaches to answering this question include both behavioral and structural methods. In this paper, I build on behavioral demand literature to examine firm conduct in the frozen juice industry, which meets many criteria that facilitate cartel formation: the frozen juice industry has few firms, there is little product differentiation, and demand has been stable over time. This paper uses a random coefficients logit model to estimate demand and infer the marginal costs for all frozen juice products under both competition and monopoly conduct. I compare these estimates to the observed marginal costs. Using a least squares method, I find that the observed costs are closer to the costs estimated under the competition assumption than to the costs estimated using the monopoly or collusion assumption. This implies that the frozen juice industry experiences low collusion.

Compiled sales data of 272,754 frozen juices from Dominick’s database, the product attribute information of each product on manufacturer websites, and the store-specific demographics dataset from U.S. Census Data (1990)
Learned Python, operated it to write BLP’s code, tested Prof. Nevo’s result within one month, and ran my essay’s code for another month to arrive at the result
Described how to use private label products and product characteristics to generate reasonable approximations for differentiated products’ costs
Reformed the supply function and used the observed marginal cost data and the least square method to find the weight of competitiveness, which is 0.74 in the frozen juice industry

Demand for Differentiated Products: PSB’s Distance Metric Approach (Dissertation Essay)

Abstract: Economists usually use the discrete choice model to investigate industries with differentiated products. However, the discrete choice model may not work well in industries where consumers buy more than one product. In this paper, I use PSB’s distance metric model to estimate all frozen juices’ marginal costs assuming no collusion. I find that the distance metric model is not better than the random coefficient model when we don’t have enough product characteristics. By comparing the elasticities be- tween weekly frequency and quarterly frequency, I conclude that with enough data, we don’t need to consider the stockpiling problem while using the distance metric model, and consumers are more price-sensitive for uncommon goods and large-size products.

Built more than 100 thousand matrices with Python to run the regression
Improved PSB’s distance metric approach by adding average price variables and developed a consumer behavior theory to explain the negative cross-price elasticity
Compared BLP’s model and PSB’s approach and evaluated the limitation of PSB’s distance metric method
Determined that consumers are more price-sensitive for uncommon goods and large-size products

Estimating the Efficiency of the Production Line (Dissertation Essay )

Abstract: Understanding the efficiency of the production line is very important for both manufacturers and competition authorities. However, it is hard for economists to gauge differences in firms’ efficiency of the production line using publicly available data. This paper develops a new technique, combining the discrete choice demand estimation and a new cost function utilizing the firms’ entire product line and product prices, to estimate differences in firms’ efficiency of production and then applies this technique to analyze the production efficiency in the U.S. automobile industry.

Collected, cleaned, and analyzed data covering 120 million automobile sales in the U.S. market
Evaluated flaws of the standard discrete choice approaches and improved the discrete choice demand estimation
Implemented a new cost function and wrote Python code to estimate the differences in firms’ efficiency of production
Demonstrated the value of this new technique and applied it to conduct a counterfactual analysis

Page updated

Google Sites

Report abuse