Surveys Flashcards
L11 (24 cards)
what is a surbey
experiments embedded in sruvets
forms of interviewa
face to face, computer assisted telephone, web –based
how are surveys different from RCTs?
useful to elicit respondents truthful beliefs
advantages of surveys
cheap, enables causal identification of effects, representative samples
disadvantages
weak, not representative, hard to link attitudes to behaviour
vignette experiment
priming, framing or endorsement
what is a vignette
randomly assign treatments, question wording or information differs, we identify the effect of question wording
identification assumptions behind vignettes
treatment independent of potential outcomes, groups are balanced
same as RCT
difference in means estimator, regression estimator, hetero-robust se’s / conservative estimator
framing experiment
how you frame the question
priming
how priming information affects respondents attitudes, respondents get the same question, but different prior information
vignette experiment can help us to understand the…
effect of different phrasings of a question or different types of information being provided
list experiments are used to
learn about attitudes/experiences that respondents would be reluctant to share
list experiments hide individual responses from the researcher and thus…
can induce truthful responses
list experiment set up
randomly assigned respondents, control group say how many N items on a list they agree with/experience, treatment group have an extra item. TE identiifies the support for or experience of an extra iterm
identification assumptions of lists
Treatment is randomly assigned , no design effects, no liars
what is a design effect
that an additional item does not affect the respondents response to the control items
If assumptions are met, what treatment effect do lists provide
Provide an unbiased estimate of the proportion of those who agree with the sensitive item
conjoint experiments
how do people make trade-offs between different options?
what does conjoint experiment help with?
social desirability bias
what the estimator of the conjoint experiment
the average marginal component effect - the causal effect of a particular attribute on a profile on the probability that that profile is chosen, holding all other attributes fixed
how to solve for a conjoint experiment
subclassification regression estimator
what do you do with SE’s?
cluster at the individual level
conjoint identification assumptions
randomised treatment, no carry over effects, no profile order effects
