Why the Deadline Matters More Than the Average

The interaction result showed that local concentration can change what additional UI time produces, but an average effect still hides who actually responds. The IZA paper’s German evidence shifts attention to the worker’s position in the benefit spell: a long extension may matter little for most people and a great deal for a small group close to exhaustion. This chapter explains why exhaustion points and changing worker composition make a single duration–wage story unreliable.

The extension acts at the exhaustion point

The IZA paper revisits an age-based discontinuity in potential benefit duration from 12 to 18 months. The institutional setting gives researchers a comparison between workers just below and just above an age threshold, while the six-month difference in potential duration creates the policy variation. The important result is not the average six-month response by itself. It is where in the distribution of durations the response appears.

The paper estimates quantile treatment effects rather than looking only at the mean. For unemployment duration, the effect is close to zero through much of the distribution. It begins to rise around the upper part of the distribution and becomes clearly different from zero around the 87th percentile. At that point, the difference grows rapidly to roughly five or six months. The shift is concentrated around the first benefit-exhaustion point rather than spread evenly across short and long spells.

The nonemployment-duration pattern is weaker. The distribution remains largely unaffected until around the 70th percentile, with the largest difference near the 80th percentile. The confidence bands are wide enough that the paper cannot reject no effect on nonemployment duration with the same confidence. Still, the stable part of the distribution is informative: before the exhaustion points, the authors see no meaningful change in the timing of exits.

stateDiagram-v2 state "Short potential duration" as Short state "Long potential duration" as Long state "Search before first exhaustion" as Before state "Near first exhaustion" as Near state "Unemployment duration response" as UnempResponse state "Nonemployment response for a minority" as NonempResponse state "Reemployment" as Reemployed [*] --> Short [*] --> Long Short --> Before: Benefit clock starts Long --> Before: Benefit clock starts Before --> Near: Spell approaches first limit Before --> Reemployed: Most spells exit without a duration shift Near --> UnempResponse: Small group stays longer in unemployment Near --> NonempResponse: Fewer remain without employment longer UnempResponse --> Reemployed: Later exit after added benefit time NonempResponse --> Reemployed: Delayed reemployment for a minority

The state sequence makes the mechanism visible. The longer-benefit group does not keep every worker in the searching state for six additional months. Most workers leave before the first exhaustion point, so the additional eligibility does not bind their timing. A smaller group reaches the point at which the difference between 12 and 18 months becomes economically relevant. For those workers, the extension can move the exit point; for the rest, it may change the amount of support available without changing when they return.