Statistical Targeting vs. Profiling Systems
Profiling models predict only an individual's baseline risk of long-term unemployment without active treatment, whereas targeting models estimate net individual impacts across multiple heterogeneous treatment options. Profiling fails when baseline risk does not correlate with net programme benefits, leading to sub-optimal program allocation.
Picture this
Think of profiling like a hospital sorting patients solely by how sick they are, giving the most intensive care to the sickest person regardless of whether that care helps them. Targeting is like testing which specific medicine works best for each patient, ensuring people get the exact treatment that improves their health.
What the evidence says
Empirical analysis demonstrated that long-term unemployment risk is not highly correlated with individual programme impacts, proving that targeting systems provide superior allocation efficiency over traditional profiling models.
- Who
- National population sample of N = 460,442 Swiss jobseekers aged 25 to 55 from administrative databases (2001–2003).
- How
- Econometric evaluation comparing baseline unemployment risk predictions E[Y^0|X] against potential outcome estimations E[Y^r|X] across 6 to 8 active programme categories.
What to do
Transition public service allocation tools from passive risk profiling models to multi-treatment targeting models that evaluate conditional potential outcomes across distinct interventions.
From the source
"Targeting is preferable to profiling if a variety of heterogeneous labour market programmes are offered... and if the long-term unemployment risk is not highly correlated with programme impacts."
Targeting_Labour_Market_Programmes_Results_from_a_Randomized_Experiment.pdf