Method
This page exists because our entire value to a unit of government rests on the figures we hand across the table surviving scrutiny. So here is the whole method, including the parts that cost us numbers.
Every figure traces to a public filing, an executed document, or a named published source. Nothing is modelled, projected, estimated or inferred unless it is labelled as such on the same line where it appears. Where the underlying filings will not support a figure, we withhold it and explain why rather than publishing something that will not hold.
There is no artificial intelligence anywhere in the path that produces a number on this site or in the Pressure Test. Not because the technology is bad, but because the claim we make is that nothing here is guessed — and a model in the arithmetic would make that sentence false.
Sources
The landscape database behind our benchmarking is assembled from these and only these. Anything a third-party tool generated, we do not load — a dataset that cannot name its filing is not evidence.
| Source | What it gives us | Vintage |
|---|---|---|
| Indiana Gateway — Annual Financial Report, detailed disbursements | Salaries and wages (line D101) and employee benefits (line D107) as filed by each unit. Townships additionally file a “Health Care” line (D808). | 2024 |
| Indiana Gateway — 100R | Position-level compensation, from which we derive a verified full-time headcount after deduplicating people across positions. | 2025 |
| DLGF Form 4B | Adopted budget by fund and unit. | 2025 |
| NCES Common Core of Data, joined to IDOE enrolment counts | School corporation identity and enrolment for the 291 school rows. | Current release |
| Executed contracts obtained through public records requests | Vendor identity, compensation, term and audit rights, where a unit has provided them. | As dated |
Where an advertised figure exceeds the adopted figure, we report the adopted figure, which is the one that governs.
The four flags
A benefits number divided by a headcount is only meaningful if both halves mean what they appear to mean. Four conditions tell us they do not. Any one of them and the per-employee figure is withheld — in what we publish, in what the Pressure Test shows a visitor, and in what our own systems record.
Flag 1
Benefits reported below roughly 15% of salaries. Almost always this means the unit books benefits to a fund outside the reported line, so the figure is not the whole cost. We record the headcount, withhold the per-employee figure, and note that the real number has to be sourced from the unit directly.
Flag 2
Benefits reported at roughly 60% of salaries or above. The line is bundling PERF pension contributions and payroll taxes with medical, so a per-employee figure drawn from it overstates the health plan — sometimes by a factor of two. Flagged and labelled wherever it appears.
Flag 3
We hold a headcount proxy derived from payroll for every unit, and we never use it as a denominator. A cost per employee requires a headcount deduplicated from the 100R and classified full-time. Statewide that condition is met for 28 units, and only those produce a figure we will stand behind.
Flag 4
The county or unit named in a file’s header does not match its filename. It happens more than you would expect. We trust the header and flag the file, because presenting one county’s payroll to another county is the sort of error you do not recover from in a first meeting.
A per-employee-per-month figure computed on benefits-eligible headcount rather than enrolled headcount is a lower bound, not the cost. Where we can only obtain eligible counts, the figure is published as a lower bound and labelled as one on the same line.
The comparison
The Pressure Test compares units on the ratio of reported benefits to reported salaries, because that ratio is computable from the AFR alone — it needs no headcount, so it exists for 863 units rather than 28. Cost per employee appears only as a labelled benchmark drawn from the verified population, never as the comparison itself.
| Unit type | Units in band | Median benefits-to-salary ratio |
|---|---|---|
| County | 81 | 46.7% |
| City / Town | 444 | 21.6% |
| Township | 338 | 7.8% |
Source: 46 North extract of the Indiana Gateway AFR 2024, lines D101 and D107. The wide spread between unit types is itself a finding: counties commonly book pension and payroll taxes into the same line, which is why the flags above matter more than the medians do. There is deliberately no school corporation band — the 291 school rows in the master carry no salary or benefits line, so no honest comparison exists.
The verified population
Across the state, 28 units survive every check and produce a cost per employee we will publish. The median is $33,195 and the range runs from $17,111 to $51,885. Seven of those carry a pension-inflated label and are shown with it attached.
That population is a benchmark, not a target and not a savings estimate. A unit sitting above the median is not overspending; it may be covering more people more generously, which is a policy choice its board is entitled to make. What the figure does is tell you whether the line is worth a closer look this year.
We will not present a per-employee figure to a unit of government as a savings estimate, and we will not quote a figure that came out of the headcount proxy. Where an independent review is likely to find savings, we say that reviews of this kind commonly find five to fifteen percent — an experience range, framed as one, never a guarantee and never attached to a specific unit before the work is done.
Every figure on this site names its source. If one of them does not reconcile to the filing it cites, we want to know — that is a defect in our work and we will correct it and say so.