Public-Sector Budgeting: State Agency Modernization

A budget that only ever moved at the margin

A state health and human services agency administered a 1.9 billion dollar budget across roughly 140 programs. Like most public bodies, it built each year's budget from the prior year's baseline plus a legislated percentage, adjusted line by line. The method was defensible and nearly immovable. A program that had underperformed for a decade kept its funding because it had funding, and a program with strong results could not grow faster than the across-the-board increase allowed. Reallocation happened only in a crisis, and even then it was contested line by line without a shared basis for deciding which lines should move.

The agency's leadership had circled this for several budget cycles. The barrier was not appetite, it was defensibility. In public budgeting, a reallocation that cannot be justified to the legislature, the auditor, and the affected constituencies is worse than no reallocation at all, because it invites a challenge the agency will lose. The Stratenity engagement was scoped to rebuild the budget cycle around the decisions the agency actually needed to make, moving money toward evidence, while producing a reallocation that would survive legislative scrutiny and an independent audit.

The starting evidence was uneven across the 140 programs. Some carried rigorous outcome data, a few even had randomized follow-up, while others reported only the inputs they consumed and the people they enrolled. The first two weeks built a shared evidence base by cataloguing, for every discretionary program, what outcome it claimed, what it cost, and how strong the proof behind the claim actually was. The exercise surfaced an uncomfortable truth the leadership already half-knew: a significant share of the discretionary budget funded programs that had never demonstrated an outcome at all, protected only by the inertia of their own baseline.

Programs compete on outcome-per-dollar, with a protected floor

We separated the budget into what the agency was legally obligated to fund and what it chose to fund, because only the discretionary portion can be reallocated and pretending otherwise breaks the exercise. Statutory and entitlement obligations sat in a protected floor, funded first and untouched. The discretionary portion, about 320 million dollars, became a contestable pool where programs competed on evidence of outcome-per-dollar rather than on their historical baseline. Each program submitted its outcome measure, its cost, and the strength of the evidence behind its results, scored on a common rubric so a strong claim with weak evidence did not outrank a modest claim that was proven.

A worked case showed the method's teeth. Two workforce-training programs sought the same 8 million dollars. The older program had run for years on a full baseline but could show only enrollment counts, not employment outcomes. The newer program had a randomized follow-up showing a measurable placement lift per dollar spent. Under the old cycle the older program kept its money by default. Under the new one, the evidence-backed program received the funding to scale, and the older program was given a single cycle to produce outcome evidence or be wound down. It did not, and it was retired, its funding redirected on a documented, auditable basis. The documentation was the point. In a public setting the wind-down of a long-standing program invites a political and legal challenge, and the agency won that challenge not by arguing but by producing the evidence score, the cost comparison, and the single cycle the program had been given to prove itself.

We designed the budget cadence to repeat before the first cycle closed. An annual review re-scored every discretionary program on the same rubric, so a program that scaled had to keep earning its funding and a new proposal entered as a bounded pilot rather than a permanent baseline. A single budget executive owned the rubric and the reallocation decisions, which kept the process from fragmenting back into 140 separately defended lines the moment the pressure of the engagement lifted.

Money followed evidence, defensibly

• The agency reallocated 47 million dollars across the discretionary pool in the first cycle, moving funds toward programs with demonstrated outcome-per-dollar.
• Two long-running, low-evidence programs were wound down against evidence, and three high-performing programs received the funding to scale.
• Statutory and entitlement obligations were funded in full through the protected floor, so no reallocation ever touched a legal mandate.
• Every funding decision carried a documented basis, outcome measure, cost, and evidence score, which held up when the legislative budget office and the state auditor reviewed the cycle.
• The cycle now repeats without the consulting team, so reallocation is a standing capability of the agency rather than a one-time exercise.

What the engagement taught

• Separate the protected floor from the contestable pool first. Reallocation is only credible when it visibly never touches statutory or entitlement obligations.
• Score evidence, not ambition. A common rubric stops a confident but unproven claim from outranking a modest result that is actually demonstrated.
• Defensibility is the constraint that matters in public budgeting. A move that cannot survive the auditor and the legislature is not worth making.
• Give losing programs a defined path, fund a proof or retire, so a wind-down is a documented decision, not an ambush that invites a political fight.
• Build the cycle to repeat. A one-time reallocation reverts to baseline drift; a standing cadence keeps money following evidence year over year.

How to run this in your agency

• Split the budget into a protected floor of legal obligations and a discretionary pool that is genuinely contestable, and be honest about the size of each.
• Require every discretionary program to submit an outcome measure, its cost, and the strength of its evidence on a common rubric.
• Fund high-evidence programs to scale and give low-evidence programs one cycle to prove results or be wound down.
• Document the basis for every funding decision so the cycle survives review by the legislature and the auditor.
• Institutionalize the cadence so evidence-based reallocation repeats each cycle rather than reverting to incremental baseline adjustment.