Inattention Mixture Model Correction
RCTField Experiment
Online job application field experiments suffer from participant inattention and accidental clicks, which threaten to distort structural logit choice parameters and bias economic estimates.
Picture this
Imagine a multiple-choice survey where 10% of participants choose answer A or B by flipping a coin without reading; by mathematically separating the random coin flippers from the attentive readers, researchers isolate the true choices of attentive participants.
What the evidence says
Inattention affects choice variance without biasing central tendencies; the inattention-corrected flow value parameter ($\tilde{z} = 0.58, s.e. = 0.04$) matches the standard logit baseline ($\tilde{z} = 0.58, s.e. = 0.04$) identically across all hour choices.
- Who was studied
- N = 1,152 unemployed applicants across 80 U.S. metropolitan areas in a nationwide hiring experiment.
- How
- Two-component mixture logit model ($Pr(1[h_i]=1|e_{ih}) = P_e(1-\alpha) + (1-P_e)\alpha$) where the inattention rate ($2\hat{\alpha} = 10.6\%$) is empirically identified from post-choice job detail recall errors.
What to do
Implement mixture choice models incorporating empirical error recall rates to verify whether cognitive noise distorts structural economic parameters.
From the source
"We estimate $2\hat{\alpha}=10.6\%$ using the fraction of applicants who incorrectly recalled their job choice later in the application... inattention-corrected estimates give similar results."
Labor_Supply_and_the_Value_of_Non_Work_Time_Experimental_Estimates.pdf