Uncensored Aggregate Income Tracking
Evaluating labor market policy success solely through accepted post-unemployment wages introduces sample selection bias because wage data is only observable for individuals who find employment before the observation window closes.
Picture this
Imagine rating a math program only by the test scores of students who passed the final exam. Because struggling students who dropped out are excluded, the calculated average score is artificially inflated; measuring total financial support and earnings across every enrolled student gives a complete and unbiased picture.
What the evidence says
While conditional wage comparisons suffered from duration-dependent sample censoring, total 365-day cumulative income tracked across all randomized participants provided a selection-free metric of total participant financial outcomes across treatment arms.
- Who
- N = 4,163 male unemployed job seekers aged 25–64 tracked in German administrative social security databases.
- How
- Methodological duration evaluation comparing conditional accepted wage distributions against uncensored 12-month cumulative gross income (combining net unemployment benefits and active labor earnings).
What to do
Track cumulative gross income combining public benefit transfers and labor earnings over fixed calendar windows rather than relying on conditional accepted wage rates when evaluating active labor market interventions.
From the source
"Inference on post-unemployment outcomes is hampered for the reason that those are only observed if exit to work occurs before the end of the observation window... We do examine the total earnings obtained in t periods after inflow into unemployment. These earnings add UI benefits received to labor earnings in employment and are observed for every individual."
Mandatory_integration_agreements_for_unemployed_job_seekers_a_randomized.pdf