Policy One-Pager Medicaid Drug Rebates
Working Paper · 2026

Pooling is not enough: do interstate Medicaid drug-purchasing pools actually move rebates?

33 states are in a drug-purchasing pool. Once you control for pharmacy-benefit governance — uniform PDLs, carve-outs — the standalone pool effect collapses to roughly zero.
Jonathan Palisoc Ph.D. Candidate, Health Services Org. & Policy
University of Michigan
The bottom line

In the lead saturated event study (SSDC), the uncontrolled post-adoption effect on supplemental-rebate share is +3.37 pp. After pharmacy-benefit governance controls, it attenuates to +1.28 pp. A stable-design benchmark dropping reform-dense windows gives −2.05 pp. The standalone any-pool coefficient is 0.06 pp (p = 0.95). Nominal pool membership is not separable from the governance bundle.

+0.06 pp
Standalone any-pool coefficient on supplemental-rebate share. p = 0.951
01

The premise — and why it’s never been clean.

Three pool families (NMPI, TOP$, SSDC) emerged in the mid-2000s on a simple intuition: more covered lives ⇒ more bargaining power ⇒ bigger supplemental rebates. By 2024, 33 states belonged to one.

But pool adoption almost always travels with a broader pharmacy-benefit overhaul: uniform PDLs, pharmacy carve-outs, MCO formulary autonomy, vendor restructuring. The pool, the carve-out, and the uniform PDL move together — and nobody had tried to separate them.

02

Three estimands, one story.

SpecificationΔ rebate share
SSDC, uncontrolled saturated event study+3.37 pp
SSDC, + governance controls+1.28 pp
Stable-design benchmark (drop reform-dense)−2.05 pp
Static bundle: SSDC alone+1.27 pp (p=0.57)
Static bundle: any pool+0.06 pp (p=0.95)
03

Where the positive signal lives.

The strongest suggestive evidence is for a specific governance bundle — pool + uniform PDL + pharmacy carve-out from MCOs:

+8.91 pp
Bundled pool + uniform PDL + carve-out vs. all other state-years (p = 0.005)

Treated as a support-limited descriptive concentration, not a causal headline — the cell sample is small.

A

Why it matters.

  • Net Medicaid drug spending grew 72% from FY17–FY23 — rebate levers are first-order fiscal policy.
  • Supplemental rebates generate $1.7B+ per year and vary from ~0% to 28%+ of drug spend across states.
  • If pooling is the lever, more states should join. If governance is the lever, joining a pool isn’t enough.
B

What we did.

Built a new 50-state, 2002–2024 panel linking CMS-64 supplemental-rebate data to a hand-coded, spell-level pharmacy-benefit governance dataset — uniform PDL status, carve-out, MCO formulary autonomy, vendor type, negotiation model.

Used SSDC’s late-adopter set as the lead heterogeneity-robust saturated event study; NMPI and TOP$ as appendix extensions.

C

Robustness.

  • Switcher analyses, federal-rebate headroom checks, SDUD drug-mix controls.
  • Rambachan–Roth honest sensitivity bounds.
  • No specification recovers a meaningful standalone pool effect.

So what?

Joining a pool is not, on its own, a lever for higher rebate capture. The actionable story is the governance bundle — preferred drug lists, carve-out architecture, negotiation model — that historically traveled with pool adoption. State fiscal staff should price that bundle, not the pool label.

Jonathan Palisoc · Ph.D. Candidate, Health Services Organization & Policy, University of Michigan
Working paper, 2026. Cite as: Palisoc, J. (2026). Pooling Is Not Enough.
Read the paper
papers.jonpalisoc.com