Against the backdrop of the recently introduced Anti-Waste and Circular Economy Law (2020), in two days, we will examine the design and potential impact of the repairability index by going through all major stages of a discrete choice experiment (DCE).
In Day 1, we will start with an introductory lecture on DCE. We will discuss the potential of DCE by comparing it with other traditional impact evaluation techniques such as randomized control trials. After that we will design a DCE together, going through the experimental design, questionnaire design and online setup. Our target is to come up with a ready-to-go survey so that you and your friends can fill in the survey and provide data for the activities in Day 2.
We kick start Day 2 with a mini lecture on random utility theory. The knowledge will prove to be essential in understanding how the data set should be structured and analyzed. By this time, you will understand better why a rather simple looking logistic regression model can inform us how people valuate things. We will then spend time to prepare the data for analysis. This is the time to apply your Stata skills to clean and re-structure the data to make them fit for further data analysis. We will then analyze the data and conclude by discussing our findings related to the value and/or design of the repairability index and reflecting on the original survey design.
After completing this part of the course, students will be able to design a discrete choice experiment, conduct it in an online environment, analyze survey data, and provide recommendations to firms or policymakers based on the findings. Students will also have ample opportunities to apply and sharpen their data management skills using Stata. At a more general level, students should be able to reflect on the possibilities and limitations of the method.
Course overview:
DAY 1
What are discrete choice experiments (DCE)? How may we assess the impact of the repairability index on green transitions? What kind of questions can and cannot be answered by DCE?
- Session 1: Identify Objective (0.5 hours)
We will come up with a question answerable with DCE.
- Session 2: Experiment Design (2 hour)
We will identify relevant attributes and levels, and then decide what to keep in the survey. We will also discuss issues related to label and visual display.
- Session 3: Questionnaire Development (1 hour)
We add additional questions to the survey. Other practical issues such as the number and sequence of questions will also be covered.
- Session 4: Final Preparation (1-2 hours)
It is time to set up randomization on Qualtrics. We will also prepare a consent statement and conduct a pilot study. We will also talk about sampling strategy and how to distribute online surveys.
- Data Collection (overnight)
DAY 2
- Session 5: Random Utility Theory (1 hour)
It is not so straightforward to see why a simple logit model can tell us how people valuate certain things. Some knowledge about the theory is going to help us understanding how to analyze the data. The data downloaded from Qualtrics are not yet ready for analysis. We need to reshape the data in a way that fits the command that everyone knows. An understanding of the theory will shed us some light on how the data set should look like.
- Session 6: Data Restructuring (2-3 hours)
Once we know how the data should look like, we are ready to make our hands dirty. The seemingly simple task could be daunting. You think you are good at Stata? This is the time to test it.
- Session 7: Data Analysis and Interpretation (2 hours)
Time to run the model now. We will estimate the willingness to pay and use the estimation results for some scenario comparison. Note that they may not make perfect sense given our research objective and experimental design. But at least you know the steps and can apply them when they are appropriate.
- Session 8: Final Discussion (1 hour)
We will also discuss our findings related to the value and/or design of the repairability index. To what extent can information induce behavioural changes? Finally, we will have had relatively little time to design our experiment. Things may go wrong in the design process that makes the design not measure up to our objectives. It is time to reflect on the original survey design and propose ways to improve it.