Marketing and the Analysis of Experiments and Quasi-experiments

Introduction

Marketing is about understanding consumer preferences and behavior, predicting future needs, and testing the effectiveness of different marketing activities. As such, it is a discipline that affects most industries, and with the advent of digital technologies and vast amounts of both aggregated and individual-level data it is more data-driven than ever before.

In this course you will be given a brief overview of what defines marketing as a discipline and learn about the marketing process from a data-driven decision perspective. We will then focus on the use of causal inference methods using experimental and quasi-experimental data to study marketing phenomena. You will learn about the basic requirements to identify causal effects and how to design experiments (e.g., AB testing, randomized control trials). You will be exposed to statistical methods allowing us to derive causal relationships from experimental and quasi-experimental (or observational) data. While these methods are widely applied in many different disciplines (e.g., economics, political science, sociology), we will use applications from the field of marketing research to illustrate the principles, challenges, and opportunities of these methods, as well as how to derive managerial recommendations from this type of analysis.

Course content

The course will roughly center around the following core topics. Details are exemplary and may be subject to change.

• What is marketing?
• Why and when is data science applied in marketing?
• Foundations of causal inference
  • Potential outcome framework
  • Different types of treatment effects
  • Randomization, selection effects, and basic inference
  • Basics of causal diagrams
• Experimental designs
  • Field vs. lab vs. natural experiments
  • Between vs. within subject design
  • Multi-factorial design
  • Internal and external validity
• Quasi-experimental and observational methods for causal inference
  • Concept of endogeneity
  • Instrumental variables (IVs)
  • Propensity score matching
  • Difference-in-difference design
  • Selected other methods