Endogeneity

Question 1

Information equivalence (IE)

Answer 1

If we can’t attribute the effect to the treatment, i.e. if the information you give does something more than it is supposed to do

Question 2

Intend to treat (ITT)

Answer 2

What we intend to treat, not necessarily the same as the actual alteration of beliefs in the respondent

Question 3

Abstract encouragement

Answer 3

Telling people explicitly that they shouldn’t think about it as a real life example, i.e. so they do not associate for example a “democratic country” with a specific region

Question 4

Realistic Bayesian model

Answer 4

First, it holds that the relevant prior beliefs of survey respondents are realistic, in that they reflect the relationships among different attributes in the real world. For example, because democracy and European location are positively correlated in the real world, respondents should believe that a country described as “a democracy” is more likely to be in Europe than one described as “not a democracy.” Second, the model holds that survey respondents are Bayesian updaters—that is, given their priors, they respond to new information by updating their beliefs according to the laws of conditional probability.

Question 5

Embedded natural experiments

Answer 5

Block every intended association by making sure the status of the thing in the vignette seems random to the respondent.

Dafoe et. al: Here is a country that just randomly happens to be a democracy or dictatorship.

Question 6

Types of preventions for IE violations

Answer 6

Abstract encouragement

Covariate control

Embedded natural experiments