What Makes a Data Center Economic Impact Study Credible?

Data centers have become one of the most economically significant infrastructure investments in the United States. As states and localities compete to attract them — and as policymakers debate tax incentives, utility rate structures, and land use approvals — economic impact studies have become a standard part of the policy conversation.

But not all impact studies are created equal. Some are rigorous, transparent, and defensible. Others are advocacy documents dressed up as analysis. Knowing the difference matters — whether you are commissioning a study, reviewing one, or deciding whether to rely on it in a policy debate.

The Methodology Question

The first question to ask about any data center impact study is: What model was used, and is it appropriate for this application?

SRC EvalMetrics uses IMPLAN (Impact Analysis for Planning) and RIMS II (Regional Input-Output Modeling System, produced by the U.S. Bureau of Economic Analysis) for regional economic impact analysis of data center projects. Both are established input-output frameworks that support transparent treatment of construction, operations, local supply chains, and induced household-spending effects.

Studies that use proprietary models, undisclosed methodologies, or simple employment multipliers without a documented regional model are not making an estimate — they are making an assertion. The difference matters when the study is presented to a legislative committee or a skeptical journalist.

The Equipment Problem: Where Most Inflation Hides

Data centers are capital-intensive. A hyperscale facility may spend $500 million to $2 billion on servers, networking equipment, and cooling infrastructure — most of it manufactured outside the region and often outside the United States. This is where many studies quietly inflate their numbers.

The correct treatment in IMPLAN is to model equipment purchases as commodity output events using margined values — not purchaser prices. The BEA benchmark data suggest that for electronic equipment of this kind, the retail and wholesale margin is roughly 12 cents on the dollar. The remaining 88 cents represents the manufacturer's price, and for imported equipment, that value leaks out of the regional economy almost entirely. A study that models the full purchaser price of imported servers as locally circulating spending is overstating the regional economic impact by a factor that can easily exceed two.

The correct procedure is to split the equipment spend into its margin component (which stays local, supporting wholesale and retail trade activity) and its basic price component (which must be assigned a regional purchase coefficient reflecting actual local sourcing). The retail margin itself should be reclassified to the appropriate trade sector, not left in the manufacturing sector where IMPLAN's default commodity event would place it.

The Local Component: Regional Purchase Coefficients

IMPLAN's default regional purchase coefficients (RPCs) estimate the share of each commodity that is sourced locally versus imported from outside the study region. For most commodities in most regions, the default RPCs are reasonable starting points. For data center equipment, they are not — and a credible study will not use them without examination.

The Bill-of-Goods Discipline

A credible data center impact study does not model the facility's spending as a single aggregate. It decomposes the construction budget and operating expenditures into their component sectors — structural work, electrical systems, mechanical systems, IT infrastructure, staffing, utilities, maintenance — and models each sector separately with appropriate RPCs and multipliers.

For operations specifically, the production function embedded in IMPLAN's industry averages is unlikely to match a data center's actual cost structure. Data centers are unusually capital-intensive and unusually light on labor relative to their revenue. A study that uses an industry-average production function without editing it to reflect the facility's actual staffing and spending mix will misstate both the direct impact and the multiplier effects.

Single Facility, Cluster, or Region-Wide?

The unit of analysis matters. A study of a single facility is not the same as a study of the data center industry in a state, and the two should not be confused — or combined without explicit acknowledgment.

Cluster and industry-wide studies face an additional methodological challenge: agglomeration effects. Data centers tend to co-locate near power infrastructure, fiber networks, and each other. Some of the economic activity attributed to the industry in a region may reflect geographic concentration rather than net new activity. A credible study is explicit about whether it is measuring gross activity or net new activity, and whether the counterfactual (what would have happened without the data centers) has been considered.

The geographic scope of the study should also match the policy question being answered. A state-level study is appropriate for state tax policy; a county-level study is appropriate for a local abatement negotiation. Studies that use state-level multipliers to answer county-level questions are producing numbers that do not belong to the decision being made.

The Multiplier Problem

Multipliers are the mechanism by which direct spending generates indirect and induced economic activity. They are also the most commonly misused element of economic impact analysis.

Two issues arise repeatedly in data center studies. First, Type I versus Type II multipliers: Type I multipliers capture supply-chain (indirect) effects only; Type II multipliers add household spending (induced) effects. Type II multipliers are larger, and some studies use them without disclosure. A credible study states which multiplier type is used and why.

Second, the commuter leakage problem: data centers often employ highly paid workers who live outside the immediate study region. If the study uses a regional multiplier but a significant share of labor income is spent outside the region, the induced effects are overstated. A credible study either adjusts for commuter patterns or acknowledges the limitation explicitly.

The Scope Definition Problem

A credible study defines its scope precisely and applies it consistently. This means being explicit about:

What is included — construction-phase impacts and operations-phase impacts are fundamentally different and should be reported separately. Construction is a one-time event, properly reported in job-years (one job for one year); operations are a recurring annual flow. A study that blends them — or reports construction job-years in a way that invites readers to mistake them for permanent jobs — is overstating the ongoing economic contribution of the facility. The honest presentation states the job-years total and its annualized equivalent over a realistic build period.

What geography is covered — as above, the scope should match the policy question being answered.

What time period is covered — one-time construction impacts are not the same as annual operations impacts. Both are legitimate; conflating them produces misleading totals. Where a study compares recurring annual benefits against a one-time upfront cost (as fiscal analyses of front-loaded tax exemptions must), the comparison belongs in present value, with the discount rate disclosed.

The Fiscal Analysis Component

For data centers specifically, fiscal impact analysis is often as important as economic impact analysis. State and local governments are frequently asked to provide significant tax incentives — property tax abatements, sales tax exemptions on equipment purchases, and utility rate preferences. A credible study should quantify:

• Property tax revenues generated (net of any abatements)
• Sales and use tax revenues from construction and equipment
• Income tax revenues from employment
• The cost of public services required to support the facility
• The net fiscal position over a defined time horizon

Two refinements separate serious fiscal work from boosterism. First, do not put the same instrument on both sides of the ledger: if the policy being evaluated is a sales tax exemption, crediting the firm's own sales tax remittances as a benefit while counting the foregone sales tax as the cost is circular. Second, the most defensible fiscal basis before a skeptical legislative audience is often the narrowest one — the indirect and induced state and local taxes the activity generates through its supply chain and household spending, plus employees' income tax — with the subsidized firm's own direct remittances set aside. A study built on that basis cannot be accused of counting the subsidy's beneficiary as its own justification.

Studies that report only gross revenue figures, without accounting for incentive costs, abatements actually in force, or public service demands, are presenting an incomplete fiscal picture — and usually not by accident.

The Energy and Infrastructure Dimension

Modern data center impact studies increasingly need to address energy demand, water use, and infrastructure requirements — not just jobs and GDP. A hyperscale data center can consume 100–500 megawatts of power, requiring significant transmission and generation investment. Water-cooled facilities can draw millions of gallons per day.

A credible study will quantify these demands, assess the capacity of existing infrastructure to accommodate them, and identify any public investment required. Ignoring these dimensions does not make them go away — it simply means the analysis is incomplete.

What to Look For in a Credible Study

The Credibility Standard

The ultimate test of a data center impact study is whether it can withstand scrutiny from a skeptical audience — a legislative committee, a competing developer, an environmental group, or a journalist. Studies that are designed to produce a predetermined number will not survive that scrutiny. Studies that are designed to answer a question honestly — with transparent methodology, margins where margins belong, regional purchase coefficients the region can actually support, and caveats stated rather than buried — will.

At SRC EvalMetrics, our data center impact work has been used in legislative testimony, regulatory proceedings, and public policy debates. The standard we apply is simple: every number in the report should be one we are prepared to defend in public. The corollary is just as simple: every number we are not prepared to defend stays out of the report.

References

MNS Analytics & IMPLAN Group. (2023). IMPLAN system data and modeling framework. IMPLAN Group LLC.

U.S. Bureau of Economic Analysis. (2023). RIMS II: Regional input-output modeling system. U.S. Department of Commerce.

Ambargis, Z. O., Mead, C. I., & Rzeznik, S. J. (2014). University contribution studies using input-output analysis (BEA Working Paper No. 0105). Bureau of Economic Analysis.

U.S. Energy Information Administration. (2023). Electric power annual. U.S. Department of Energy.

Synergy Research Group. (2024). Data center market share and hyperscale capacity reports. Synergy Research Group.

Broadway Consulting Group. (2022). Best practices in data center economic impact analysis. Industry white paper.