Experimental Descriptive Designs

Question 1

Why do we use Experimental Descriptive Design?

Answer 1

We use experiments for descriptive inquiries when units don’t naturally reveal the characteristic of interest.

Experiments act as measurement tools for descriptive purposes,not to establish causal relationships.

Question 2

What are Descriptive Audit Experiments?

Answer 2

Measure discrimination or preference of one group compared to another. (e.g., job interviews, housing selection)

Question 3

What is the method of Audit experiments?

Answer 3

Trained researchers pose as equally qualified applicants from different groups to assess differential treatment.

Question 4

What are the limitations of Audit experiments?

Answer 4

• No potential outcomes doesn’t estimate causal effects.
• Requires considering unit types (e.g., “always discriminators,” “never discriminators”).

Question 5

What is the purpose of List Experiments?

Answer 5

Address sensitivity bias when respondents might underreport socially undesirable traits.

Question 6

What is the method of list experiments?

Answer 6

Embed a sensitive question within a list of unrelated items presented to a control group.

Question 7

What are Conjoint experiments?

Answer 7

Describe preferences for products, candidates, or policies with multiple attributes.

Question 8

What is the method for conjoint experiments?

Answer 8

• Show profiles with varying attribute levels (e.g., candidate profiles with gender, age, experience).
• Randomly assign each respondent a subset of profiles to rate.

Question 9

What is ACME?

Answer 9

Average Marginal Component Effects (ACME): average change in preference for one attribute across all other attribute levels.

Question 10

What are some considerations in Conjoint design?

Answer 10

• Attribute selection: Omitted variable bias can occur if important attributes are excluded (e.g., candidate’s partisanship masked by race or gender).
• Attribute levels: Balance the number of levels per attribute with precision (more levels require more estimation).
• Number of profiles: Consider respondent fatigue and survey cost.
• Choice format: Single vs. forced choice depends on cost and real-world applicability.
• Total profiles rated: Balance survey cost with respondent attention.

Question 11

What are Behavioural Games?

Answer 11

Measure latent or unobservable characteristics like trust, altruism, or risk attitudes.