ESSAYS ON CONSIDERATION SET MODELS AND DISCRETE CHOICE MODELS
Open Access
- Author:
- Rho, Hyung Gyu
- Graduate Program:
- Economics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 01, 2022
- Committee Members:
- Patrik Guggenberger, Major Field Member
Joris Pinkse, Co-Chair & Dissertation Advisor
Yizao Liu, Outside Unit & Field Member
Sung Jae Jun, Co-Chair & Dissertation Advisor
Marc Henry, Professor in Charge/Director of Graduate Studies - Keywords:
- Consideration Set Model
Discrete Choice Model
Consumer Choice - Abstract:
- Chapter 1 presents two identification results in a consideration set model. The consideration set model relaxes the assumption that a consumer considers all products in a market. The model allows the consumer to endogenously choose products to consider buying. It is assumed that her attention determines this consideration. I discuss identification of model parameters while allowing arbitrary dependencies between preferences and attention. Specifically, I present two identification results. The first identification argument relies on an exclusion restriction and an identification at infinity concept. The exclusion restriction means that there is a variable that affects preferences but not attention. Further, I assume that this variable has unbounded support. The second identification argument relaxes the unbounded support assumption, but it requires two exclusion restrictions. I need an additional variable that influences attention but not preferences. Under the two exclusion restrictions, I show that I can identify the parameters of interest. Chapter 2 proposes a nonparametric test of the validity of an exclusion restriction in a consideration set model. It is known that point identification in the consideration set model can be achieved with two variables -- one variable only controls preferences, and the other only affects attention. Each variable works as an exclusion restriction as they are excluded from either attention or preferences. Conversely, partial identification can be obtained if only one of these variables is available. My test can be used to test the validity of one exclusion restriction when the other exclusion restriction is available. For example, the econometrician can test whether a candidate variable, in fact, only influences preferences when there is another variable that only shapes attention. The null hypothesis for my test is that the candidate variable does not affect attention. Consequently, under the null hypothesis, the candidate variable only influences preferences. To develop my test, I start by creating a function that depends on the consumer's attention under the null hypothesis. This function is called the test function. I show that if the test function does not vary with the candidate variable, then the candidate variable has no influence on attention. Consequently, the candidate variable only changes preferences. Then, I compute the difference between the test function and the integral of the test function with respect to some weighting function on the space of the candidate variable. This difference must be zero under the null hypothesis. My test is based on measuring the significance of the square of the difference. I then use the kernel method to derive test statistic. I show that the test statistic is consistent. Chapter 3 studies the welfare cost incurred by information asymmetry between a consumer and an E-commerce platform when the platform endogenously determines its ranking algorithm. In E-commerce, the platform usually has richer information about products than the consumer. The platform provides the consumer with a list of recommended products, called a ranking. The ranking aims to help the consumer, who has limited information, find the product that suits her needs. The platform can rank products according to the relevance score attached to each product based on the consumer's search query. However, the platform can also place more profitable products at the top of the ranking. This practice can happen because the consumer has limited product information, and her attention is drawn to the first few products in the ranking. To quantify the welfare costs of the information asymmetry in this circumstance, I develop a structural model where the platform decides and commits to its ranking algorithm before the consumer enters the market. After the platform's decision, the consumer enters the market and chooses a product. My result shows that the platform exploits the consumer's ignorance. To conduct a counterfactual analysis, I consider a scenario in which the consumer has complete knowledge. Once the consumer has full knowledge, the platform redesigns the ranking algorithm such that it is more beneficial to the consumer than it is under information asymmetry.. As a result, consumer welfare is improved by 2.24%. Surprisingly, the platform's revenue also increases by about 58%. Giving the consumer complete knowledge appears to balance the bargaining power between the consumer and the E-commerce platform.