In Simple Words
Imagine estimating how crowded a gym is by counting members on paper and assuming everyone stays all day; if people actually leave early or drop off over time, the real crowding level is lower, meaning any observed overcrowding must be caused by even stronger individual crowding effects than assumed.
Research question
Calibrating baseline treatment intensity ($\tau$) under steady-state assumptions ignores negative duration dependence—the decline in job exit rates over elapsed unemployment duration—and macro-inflow shifts, resulting in miscalibrated general equilibrium structural policy simulations.
Participants
N = 89,466 benefit spells (40,403 experiment cohort and 49,063 pre-experiment cohort) across 15 Danish counties.
Study Design
Sensitivity analysis of general equilibrium search model calibrations across varying baseline treatment intensity parameter values ($\tau^e = 0.20, 0.25, 0.30$).
Results
Accounting for negative duration dependence reduces estimated baseline treatment intensity from 30% to 26% (and down to 21% when adjusting for reduced benefit inflow). Lower assumed baseline treatment intensity ($\tau^e = 0.20$) implies stronger displacement externalities on nonparticipants, causing social welfare to decline faster as program scale increases.
Takeaway
Adjust baseline policy treatment intensity parameters for negative duration dependence and cohort inflow dynamics before estimating general equilibrium search models.