Back to search

University of British Columbia

Essays of instrument-free causal inference, machine learning and marketing applications

Abstract

dc:description

A prominent challenge when drawing causal inference using observational data is the ubiquitous presence of endogenous regressors. This dissertation investigates causal inference and endogeneity correction, both in methodology development and empirical analysis. The first essay (in Chapter 2) develops a new instrument-free method using copula to address the endogeneity problem. The classical econometric method to handle regressor endogeneity requires instrumental variables that must satisfy the stringent condition of exclusion restriction. We use the statistical tool copula to directly model the dependence among the regressors and the error term, and abstract information from existing regressors as a generated regressor added to the outcome regression. Our proposed 2sCOPE method extends the existing copula method to a more general setting by allowing (nearly) normal endogenous regressors and correlated exogenous regressors, and is straightforward to use and broadly applicable. We theoretically prove the consistency and efficiency of 2sCOPE, and demonstrate the performance of 2sCOPE via simulation studies and an empirical application. The second essay (in Chapter 3) further studies the causal inference and endogeneity correction methods for high-dimensional data. The more and more common high-dimensional data in the current big data era make the classical causal inference methods suffer finite-sample bias, inefficiency, or even fail to work when the dimension is larger than the sample size. In this essay, we extend the 2sCOPE method developed in Chapter 2 to the high-dimensional setting, and propose a lasso-based 2sCOPE method. We demonstrate the performance of the proposed method via simulation studies and an empirical application. The third essay (in Chapter 4) empirically studies vertical differentiation in two-sided markets, where network size plays the central role. Vertical differentiation is a common strategy in one-sided markets, but whether it is profitable for two-sided platforms is hard to say because of the network effect. In this essay, I take advantage of a unique data set from a leading ride-hailing platform, and develop a structural simultaneous demand and supply model to quantify network externalities. The result shows that besides the product intrinsic value, network value is crucial in determining the degree of product differentiation in two-sided markets.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy - PhD
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Business Administration - Marketing
Grantor dc:publisher
University of British Columbia
Year dc:date
2022

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Yang, Fan
Contributors dc:contributor
  • Qian, Yi

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/82240

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Yang, Fan. Essays of instrument-free causal inference, machine learning and marketing applications. doctoral thesis, University of British Columbia, 2022. http://hdl.handle.net/2429/82240