Overtime Intercept Constraint Extrapolation
RCTField Experiment
High minimum overtime pay mandates under labor laws force experimental wage offers for full-time work to be high, creating severe statistical extrapolation error when estimating reservation wages for hours beyond standard workweeks.
Picture this
Imagine trying to draw a line through points on a graph where all the dots for high hours are grouped at the very top edge; by anchoring the starting height of the line based on where all previous lower-hour lines started, researchers can accurately measure the line's upward slope.
What the evidence says
Unconstrained logit models yield an imprecisely estimated marginal value of time of $43.10 (s.e. = 15.90) for 40 vs 45 hours, whereas imposing linear intercept constraints reduces the standard error tenfold, yielding a precise MVT of $21.89 (s.e. = 1.70).
- Who was studied
- N = 250 unemployed job applicants across 80 U.S. metropolitan areas in overtime choices (40 vs 45 and 45 vs 50 hours per week).
- How
- Logit discrete choice estimation with intercepted parameter constraints derived from linear extrapolation of non-overtime choice intercepts (5 vs 10 through 35 vs 40 hours per week).
What to do
Constrain logit choice model intercepts using linear regressions of baseline choice probabilities when extrapolating high-wage discrete choice parameters.
From the source
"In the '40 vs. 45 Hours' and '45 vs. 50 Hours' panels, we constrain the intercepts, $Pr(1[h_i]=1|e_{ih}=0)$, to values predicted using the estimated intercepts at lower hour treatments... dramatically increasing precision."
Labor_Supply_and_the_Value_of_Non_Work_Time_Experimental_Estimates.pdf