Every data center hearing eventually reaches the fiscal argument, and it goes about the same way everywhere.
The applicant’s consultant presents an economic impact study with a large number in it. Someone from the audience says the study was paid for by the applicant. Someone else cites a state audit finding that these projects don’t pay for themselves. The applicant’s side cites a different study finding that they do. Everyone leaves with the position they arrived with.
Meanwhile the single variable that most determines what your locality actually collects has not been mentioned once.
The variable
It’s your tangible personal property tax rate and depreciation schedule.
Data centers are unusual among large developments in that most of their assessed value isn’t the building. It’s the equipment inside — servers, switchgear, cooling plant — which is personal property, refreshed on a three-to-five-year cycle, and assessed under rules that vary enormously between jurisdictions and get almost no public attention.
The same $150 million equipment investment yields somewhere between $10.8 million and $0.4 million in local revenue over five years depending on the locality’s rate and depreciation treatment.
A twenty-seven-fold spread. Same project. Same equipment. Same everything except a schedule in your code that most elected officials have never read.
If you take one thing from this piece: establish that number before you have any other fiscal conversation. Everything else — the abatement debate, the impact study argument, the jobs numbers — is downstream of it, and arguing about them first is like negotiating a mortgage without knowing the interest rate.
Both halves of the audit finding
Virginia’s legislative audit body, JLARC, produced the most-cited fiscal analysis in this field. Both sides quote it. Each quotes half.
Per million dollars of incentive spending, data centers produced 84 jobs and 48 cents of returned revenue. Statewide averages across all of that state’s incentive programs: 58 jobs and 41 cents.
So: data centers performed above average among that state’s incentives, while still returning less than half of every dollar spent.
Both of those are true. They belong in the same sentence, and a room that hears only one of them has been handled rather than informed. If you’re an official, the useful move is to say both yourself, before anyone else can deploy their preferred half. It costs you nothing and it makes every subsequent number you cite more credible.
The school funding thing everyone gets wrong
This one gets misstated constantly, in both directions, and it’s worth getting right because it’s usually the most emotionally loaded part of the discussion.
In states where school funding is equalized, new taxable value raises the local share of school funding — and state aid falls roughly dollar for dollar. So the district’s operating budget doesn’t grow the way people assume. In that context, an abatement is largely a statewide transfer rather than a local one.
But debt service revenue generally sits outside those equalization formulas. New taxable value genuinely expands a district’s bond capacity.
The honest sentence, which I’ve used in more than one meeting:
“This will meaningfully help your district build buildings, and it will not meaningfully help it hire teachers.”
That lands hard in both directions, which is how you know it’s the right sentence. It disappoints people who were promised a windfall for the classroom, and it corrects people who said the district would see nothing.
And here’s the part nobody expects: you don’t have to argue about any of this. Every district in the country publishes its abatement losses under GASB 77 in its annual financial report. That’s a public document. The exact dollars are already printed. An argument that has consumed three meetings can become a number in an afternoon.
Reading the impact study
When the applicant’s economic impact study lands, three questions settle most of it:
Who paid for it? Not disqualifying — most such studies are applicant-funded, and that’s normal. But it goes on the record, along with the funding of every study your side cites. Applying the disclosure standard in one direction only is how a community loses the room.
Does it have a control group? The best-identified work in this area — Bahar and Wright at Brookings — compared counties that got a facility against counties where one was announced and then cancelled. That’s a real counterfactual. Their finding: a typical county gains 100–200 permanent jobs, wages unchanged, home prices up 2–5%. The authors explicitly warn that naive estimates failing to account for preexisting growth substantially overstate effects. Most impact studies are naive estimates.
Does it report job-years as jobs? “800 jobs” that means 750 construction workers for eighteen months and 50 permanent technicians is the most common inflation in the industry. There’s nothing wrong with 50 good permanent jobs. There’s something wrong with marketing them as 800.
Two findings that rarely surface
Facility type matters more than size. Hyperscale facilities generate measurable telecommunications ecosystem gains. Colocation facilities largely don’t. A bigger building is not automatically a bigger economic story.
Metro counties capture measurable gains; non-metro counties show none. If you’re a rural community being sold a growth narrative, the best available evidence does not support it. Tax base, yes. Broader economic development, mostly no.
That’s an uncomfortable finding for the industry and a useful one for a rural county — not because it argues against the project, but because it tells you to negotiate for the thing that’s actually on offer (revenue, infrastructure, specific commitments) rather than for a growth story the evidence doesn’t back.
Why this is a neutral’s job
Notice that none of the above tells you whether to approve anything.
It tells you what you’re actually deciding, in numbers you can verify yourself, from documents that already exist. Your TPP schedule. Your district’s GASB 77 disclosure. The study’s funder and method.
Most of what looks like a fiscal disagreement in these hearings is people arguing about national averages because nobody did the twenty minutes of local arithmetic that would have made the argument unnecessary.
Do the arithmetic first. Then have the argument, if you still need one.