How EB-5 Jobs Are Actually Counted: The Multiplier, Not the Payroll

There is a fact about EB-5 that surprises almost everyone outside the field the first time they meet it. In a Regional Center project, the number of jobs credited to an investor has very little to do with how many people the project puts on a payroll. The great majority of the jobs are produced by an economic model, derived from a documented dollar of spending run through a regional multiplier. Understand that one mechanic, and the entire architecture of an EB-5 job count — why it works, where it breaks, and what a reviewer is actually testing — comes into focus.

This is a piece about how that count is built, and about the discipline that separates a defensible job study from an indefensible one. The through-line is causation. Every rule USCIS enforces, from the retired tenant-occupancy methodology to the post-2022 project-application regime, traces back to a single question the agency asks of every counted job: can you show that the EB-5 investment, and not something else, is what created it? The model is the engine. The audit trail from a verified dollar of spend to a counted job is the product.

The ten-job rule, and the jobs you never see

The requirement itself is simple to state. Each EB-5 investor must create or preserve at least ten full-time jobs — positions of thirty-five or more hours per week — for qualifying U.S. workers, meaning U.S. citizens, lawful permanent residents, or other work-authorized immigrants. The investor and immediate family do not count toward the ten, and neither do nonimmigrant workers such as H-1B holders. What is not simple, and what does most of the analytical work, is the question of which jobs the program will let you count.

The answer depends entirely on how the investment is structured. In a standalone, or direct, investment, only the direct W-2 jobs at the new commercial enterprise (the NCE) count; indirect and induced jobs are not recognized at all. In a Regional Center investment, direct, indirect, and induced jobs all count. That single distinction is the central advantage of the Regional Center program, and it is the reason economic modeling matters to EB-5 in the first place. The statute underpinning it, INA 203(b)(5)(E)(iv), directs that DHS “shall permit aliens … to satisfy only up to 90 percent of the requirement … with jobs that are estimated to be created indirectly,” while providing that “an employee of the new commercial enterprise or job-creating entity may be considered to hold a job that has been directly created.” For petitions filed on or after May 14, 2022, in other words, as much as ninety percent of an investor’s job-creation burden can be met with indirect jobs, provided at least one job per investor is direct.

USCIS defines the three categories precisely in its Policy Manual (Volume 6, Part G, Chapter 2). Direct jobs establish an employer-employee relationship between the NCE, or the job-creating entity, and the persons it employs. Indirect jobs are held outside the NCE but created as a result of it — the employees of the producers of the materials, equipment, and services the project buys, the supply chain. Induced jobs are a subset of indirect effects, created when direct and indirect employees spend their earnings on consumer goods and services in the broader economy. In nearly every Regional Center project, indirect and induced jobs together are the majority of the count, and they are the jobs no one at the project will ever meet.

Legal “direct” is not economic “direct”

Before going near a model, an analyst has to hold two vocabularies apart, because they use the same words to mean different things, and USCIS adjudicator training flags the confusion explicitly. There is “legally direct” and “legally indirect,” the EB-5 statutory sense, and there is “economically direct” and “economically indirect,” the input-output modeling sense. They do not line up.

The clearest illustration is construction labor. On-site construction workers employed by a general contractor are, in the EB-5 legal sense, indirect jobs — because those workers are not employed by the NCE — even though a layperson watching them pour concrete would call them as direct as jobs get. The input-output multiplier is precisely the tool that captures these effects, the ones that are indirect in the legal sense, and it is why the counted majority of an EB-5 project can be jobs that, to the naked eye, look like the most direct work on the site. Getting this distinction wrong is one of the fastest ways to misclassify a job type in a study, which is itself a documented trigger for USCIS scrutiny.

Inside the input-output model

An input-output model is a matrix representation of the inter-industry transactions in a regional economy. It traces how an initial expenditure — what economists call a final-demand shock, and what EB-5 calls the direct effect — ripples outward to generate indirect effects along the supply chain and induced effects as wages are re-spent. The Bureau of Economic Analysis, which publishes the most widely used of these systems, describes its multipliers as providing “a measure of the effects of local demand shocks on total gross output, value added, earnings, and employment.” A dollar goes in; a set of economy-wide effects, including a number of jobs, comes out.

There is a technical distinction worth naming because EB-5 depends on it. A Type I multiplier captures interindustry effects only — direct plus indirect. A Type II multiplier adds the household-spending, or induced, effects on top. Because the Regional Center program lets a project count induced jobs, EB-5 work uses Type II multipliers. Three modeling systems dominate the field, and USCIS accepts all of them.

USCIS accepts more than one model because the regulation tells it to. Rather than mandating a system, 8 CFR 204.6 permits “reasonable methodologies … including multiplier tables, feasibility studies … and other economically or statistically valid forecasting devices.” Some practitioners nonetheless prefer RIMS II for a specific reason: it is the only one of the three whose multipliers a USCIS economist can independently reproduce and test, because BEA publishes them, while the others rest on proprietary derivatives. When an adjudicator can check your arithmetic against a government source, there is less room for argument.

Two employment multipliers that get confused

RIMS II provides, in BEA’s description, “final-demand multipliers for output, earnings, employment, and value added; and direct-effect multipliers for earnings and employment” — six types in all. Two of them are employment multipliers, and confusing the two is one of the more consequential errors in an EB-5 study.

The final-demand employment multiplier is the total number of jobs generated per one million dollars of spending delivered to final demand. In BEA’s words, “each column entry indicates the change in employment in each row industry that results from a $1 million change in final demand.” This is the multiplier applied to construction expenditure inputs, and it is the source of the familiar “ten to fifteen jobs per million dollars” convention. The direct-effect employment multiplier is a different animal: it is the total number of jobs generated per one direct job, a jobs-per-job ratio. BEA defines it as indicating “the total change in employment in the region that results from a change of one job.” You reach for it when you are starting from a known direct job count rather than a dollar figure.

The reason to get the pairing right is that the two roads are supposed to arrive at the same place. BEA’s own user handbook demonstrates the reconciliation with a sports-facility construction example for the Kansas City economic area, and it is an ideal teaching device provided one caveat is respected.

The modeling choices that move the number

A model does not hand back a single inevitable answer. The job count is highly sensitive to a handful of choices the modeler makes, and each of them is a place a reviewer will look. The first is industry and sector selection, done on the NAICS classification. Historically RIMS II carried a single “Construction” industry, code 230000; the updated data disaggregated construction into several sub-industries. Choosing the right sector for each cost category matters, and soft costs in particular should be routed to their own specific multiplier categories rather than bundled into general construction — a point USCIS made explicitly in its June 2012 EB-5 economic methodologies question-and-answer guidance.

The second choice is geography, the definition of the county or metropolitan area used as the region. As a rule the multiplier rises as the region grows, because a larger economy captures more of its own supply chain and re-spending locally rather than leaking it elsewhere. BEA’s handbook makes the point with hard numbers: for the Kansas City area, the construction final-demand output multiplier is 2.3270, but for Jackson County alone it is 1.8723. A dollar spent on construction in Manhattan carries a different indirect impact than the same dollar spent in rural Nebraska, and the region you draw is part of what determines the count. The third choice is the input type itself — construction expenditures for construction jobs, revenues or operating expenditures for operational jobs, and construction employment or job-years where a direct-effect approach is used.

Underneath all of this sits a discipline that is easy to neglect: dollar-year indexing. Inputs have to be expressed in the same dollar-year as the multipliers. The RIMS II regional base year has run around 2015, and the national benchmark tables update on years ending in two and seven; as BEA explains, “the annual RIMS II multipliers use current-year regional data in combination with the most-recent-year national benchmark input-output tables, so the 2019 RIMS II multipliers use 2019 regional data with 2012 national benchmark data.” Expenditures must be inflated or deflated to match. Skip the step and the arithmetic silently drifts.

Construction, job-years, and the 24-month line

Construction generates EB-5 jobs in two distinct ways, and keeping them separate is essential. Construction expenditures, run through the model, always produce indirect and induced jobs, and those are available to a Regional Center project regardless of how long the building takes. Direct construction jobs — the on-site building jobs themselves — are different: they may be counted as permanent only if the construction activity lasts two years, twenty-four months, or more.

The rule flows from the definition of a qualifying job. Employment that is intermittent, temporary, seasonal, or transient does not qualify as permanent and full-time; but jobs expected to last at least two years are generally not treated as temporary. So construction lasting twenty-four months or longer can yield countable direct construction jobs, while construction under twenty-four months generally yields only the indirect and induced jobs derived from the expenditure inputs. The Reform and Integrity Act layered statutory limits on top for petitions filed on or after May 15, 2022. For construction activity lasting less than two years, indirect jobs from that activity may satisfy only up to 75 percent of the job-creation requirement — the Congressional Research Service, in report IF13040, states it plainly: “Indirect jobs created by construction activity lasting less than two years may satisfy 75% of the job creation requirement.” And for direct-job purposes, the count is prorated, multiplying the total number of such jobs by the fraction of the two-year period the construction actually lasts.

This is where the job-years concept enters, and where a nuance trips up newcomers. Construction employment is conventionally measured in job-years — one job held for one year is one job-year, and multiple job-years aggregate. But in EB-5 practice, construction jobs are not counted by multiplying a headcount by duration. The model converts construction expenditures directly into a final job figure. As the advisory firm EB5AN puts it, “these job creation figures are given as final values; they are not multiplied by project duration (i.e., job-years).” The job-years idea matters far more for the twenty-four-month direct-job qualification test than for the expenditure-based count. And the stakes of that test are real. EB5AN’s own illustration is blunt: “Construction projects that last only 23 months … result in far fewer EB-5 eligible jobs than those that last 24 months. In such cases, one month may mean the difference between subscribing 5 EB-5 investors and 20 EB-5 investors.”

A worked example: from a dollar of spend to an investor slot

Assemble the pieces on a clean, deliberately simplified case. Take a fifty-million-dollar hard-construction-cost expenditure on a mixed-use hotel project in a defined metropolitan area, with construction lasting twenty-four months or more so that direct construction jobs qualify. Map the hard costs to the appropriate construction NAICS sector for the project’s county or metro, index the dollars to the multiplier’s base year, and apply the final-demand employment multiplier.

Of that six-hundred-job total, a model would split the result into its direct, indirect, and induced components — on the order of, very roughly, forty percent direct, thirty percent indirect, and the remainder induced, though the exact shares are model output and vary by region, sector, and project. The EB-5 point is not the split; it is that indirect and induced together are the majority, and that the Regional Center structure is what lets the project count them. Divide the qualifying total by ten and you have investor capacity: six hundred jobs supports sixty investors at the bare minimum.

No prudent project subscribes to the minimum. The gap between the jobs a project needs and the jobs it is projected to create is the cushion, and it is the project’s only protection against cost underruns and schedule slippage. As the advisory firm LCR frames it, “if a project needs exactly 500 jobs to support 50 investors, that leaves very little room for error … if the project is projected to create 700 or 800 jobs, the margin — the cushion — is stronger.” A project that needs to support fifty investors, and therefore five hundred jobs, wants to be modeling six hundred to eight hundred, not five hundred and one.

Construction is only half of many projects. Where the asset has an operating business — a hotel, an assisted-living facility, a retail center — stabilized annual operating revenues, or operating expenditures, are run through the model’s operating sector to produce recurring direct, indirect, and induced operational jobs. Construction jobs are one-time; operational jobs recur, and both count. Since the Reform and Integrity Act capped short-construction indirect jobs at seventy-five percent, post-2022 projects increasingly favor an operational or revenue component precisely because it generates genuine direct jobs and reduces reliance on the capped construction-indirect bucket.

What the count looked like in the field

The mechanics are easier to trust against real projects, provided each figure is read for what it is — an as-reported or estimated project number, not an independently audited count. The capital-stack logic is clean, and the law firm Holland & Knight has illustrated it with a sixty-million-dollar hotel expected to generate roughly 1,500 qualifying jobs. At about 12.5 jobs per million dollars, that is 150 investors’ worth of capacity, since 1,500 divided by ten is 150. Flip the same project around: had it generated only 150 jobs, no more than about 7.5 million dollars of EB-5 capital could responsibly have been used. Job creation, not appetite, sets the ceiling on how much EB-5 money a deal can absorb.

Two anchor cases come from a 2014 report by the Initiative for a Competitive Inner City and Harvard, Increasing Economic Opportunity in Distressed Urban Communities with EB-5, by Zeuli and Hull (June 2014). The NYLO Dallas South Side Hotel was a 76-room property costing about 19.8 million dollars, of which 5.5 million came from eleven EB-5 investors, who together needed 110 jobs. In the report’s words, “the project created a total of 161 jobs, including 35-40 jobs at competitive wages in the hotel and the indirect and induced jobs from the construction phase” — comfortably past the 110 required, though the report publishes no separate indirect-versus- induced breakdown. The University of Miami Life Science & Technology Park, whose job estimate was prepared by the Jacob France Institute at the University of Baltimore, ran larger: a roughly 107-million-dollar first phase, with 20 million dollars of EB-5 capital from forty investors needing 400 jobs, estimated to produce “over 1,200 jobs at an average salary of $34,000 during the construction phase and over 1,500 jobs (including the multiplier effect) at an average salary of $50,000 after the building becomes operational.” In both, the construction and operational multiplier effects, not the on-site payroll, carry the count.

The “reasonable methodology” standard, and garbage in

None of this modeling is worth anything to an adjudicator unless it is defensible, and the standard for defensibility is written into the regulation. 8 CFR 204.6(m)(3) requires that a Regional Center proposal be supported by “economically or statistically valid forecasting tools, including, but not limited to, feasibility studies, analyses of foreign and domestic markets … and/or multiplier tables,” and 8 CFR 204.6(j)(4)(iii) provides that indirect job creation may be shown “by reasonable methodologies.” The Policy Manual carries the standard forward: a project application “may rely on economically and statistically valid methodologies,” and the I-956F must include “a credible economic analysis regarding estimated job creation that is based upon economically and statistically valid and transparent methodologies” (Volume 6, Part G, Chapter 5).

The economic report does not stand alone. It supports the comprehensive business plan required by Matter of Ho (22 I&N Dec. 206, Assoc. Comm. 1998), which held that the plan must be “comprehensive, detailed, and credible,” detailed enough to support reasonable inferences about job creation and going well beyond “mere conclusory assertions.” It also lives under Matter of Izummi (22 I&N Dec. 169, Assoc. Comm. 1998), the at-risk-capital and material-change precedent, which holds among other things that the full amount of capital must be made available to the businesses most closely responsible for job creation, and that eligibility must be established at the time of filing.

Here is the analytical hook, and the reason a job study is a professional product rather than a formality: USCIS reviews the reasonableness of the inputs and the methodology, not merely the model output. An accepted model fed unrealistic expenditure or revenue assumptions produces an indefensible count. Garbage in, garbage out. Where the model’s inputs reflect direct jobs at the NCE or job-creating entity, USCIS requires the petitioner to demonstrate that the direct-jobs input is itself reasonable, with I-9s, payroll records, or the business plan. And the agency’s own materials name the recurring triggers: overly optimistic multipliers, misclassification of job types, inadequate linkage between the EB-5 capital and the job creation, and reliance on outdated or national data where local and regional inputs were called for. Every one of those is an input problem, not an output problem, which is exactly the point.

The cautionary tale: tenant occupancy and the causation boundary

The clearest proof that USCIS enforces a causation boundary on modeling is a methodology it retired. Tenant-occupancy modeling credited an EB-5 project — typically the developer of commercial, retail, or office space — with the indirect and induced jobs of the future tenants who would one day occupy that space. In effect, it counted the jobs of businesses that would lease space in a building EB-5 capital had helped construct.

The arc of its rise and fall is worth tracing, because the dates mark the boundary. USCIS issued a “Tenant Occupancy” notice on February 17, 2012, and began sending requests for evidence on pending Regional Center applications and exemplar petitions using the model. In May 2012, in guidance numbered GM-602-0001, and again on December 20, 2012, the agency issued operational guidance defining criteria for crediting tenant jobs, offering two routes to establish causation: mapping a specific amount of direct, imputed, or subsidized investment to new jobs, or a “facilitation- based” approach that demonstrated the project removed a significant market constraint, an excess of demand or a supply bottleneck, that would draw a specified number of tenant jobs into the space. From 2013 onward USCIS tightened its RFE template around three questions: would there be tenants to occupy the space; would the tenant jobs be genuinely new rather than merely relocated; and were the estimates built on a reasonable, transparent methodology. Many petitioners simply abandoned their tenant-job claims. In November 2016 the agency consolidated the tenant-occupancy guidance into the Policy Manual, and then, on May 15, 2018, in Policy Alert PA-2018-03, it rescinded the methodology outright, effective immediately, with comments accepted until May 29, 2018.

The stated rationale is the whole lesson. The tenant-occupancy methodologies, USCIS concluded, “result in a connection or nexus between the investment and jobs that is too tenuous” and “do not provide reasonable predictions of indirect job creation.” On the facilitation- based route in particular, the agency decided that letting petitioners avoid establishing an equity or direct financial connection between the EB-5 capital and the prospective tenants’ employees had been ill-advised, “because a direct financial connection between the EB-5 capital investment and the job creation is necessary to determine a sufficient nexus.” Reliance on a supply constraint or excess demand alone did not establish a causal link. The EB-5 practitioner Suzanne Lazicki has quoted the economic critique at the heart of a 2012 USCIS RFE, which observed that tenant jobs “would be more appropriately attributed to the tenants themselves and not to the development itself … Because the demand for labor precedes the decision about where to house that labor as a general economic principle.” The deeper problem was job-shifting: tenant jobs might be merely relocated from elsewhere in the same market rather than genuinely new, failing the net-new, but-for test of causation.

The story has a coda that keeps the boundary in view. Under the Reform and Integrity Act, as of May 14, 2022, Congress reopened a narrow door: Regional Center petitioners may again include jobs attributed to prospective tenants, but only if the number is determined by an economically and statistically valid methodology and the jobs are not existing jobs that have merely been relocated. A guardrailed revival, in other words, that leaves the causation and net-new principles exactly where USCIS put them. Tenant occupancy is the cautionary tale because it defines the outer edge of defensible methodology: the further a counted job drifts from documented direct expenditure, the less defensible it becomes. A rigorous report stays close to verifiable direct spending and its immediate multiplier effects.

The post-RIA world: methodology on trial first

The Reform and Integrity Act, enacted March 15, 2022 and effective May 14, 2022 for the codified program, changed not just the rules but the order in which they are tested, and it raised the stakes on the job study accordingly. The Regional Center now files Form I-956F, the project application that replaced the pre-RIA I-924 exemplar, and USCIS reviews it — the business plan, the offering, and the economic methodology — at the project level before any investor’s I-526E petition can be approved. The economics are examined up front, once, for the whole project. As CanAm’s general counsel Walter Gindin has put it, “USCIS will focus on the credibility of the business plan and the credibility of the economic impact analysis,” and, crucially, “the budget and the schedule will tie into the jobs.” A disconnect between the job study, the construction budget, and the construction schedule is now a direct path to a request for evidence or a denial.

Around that review sits the Act’s integrity apparatus: an EB-5 Integrity Fund financed by annual fees of 20,000 dollars for regional centers with more than twenty investors and 10,000 dollars for those with twenty or fewer, along with audits and site visits. And because the I-956F must clear review before any I-526E can be approved, deficient projects are filtered earlier in the process. The effect on outcomes is visible in the approval data. Citing IIUSA’s Director of Policy Research and Data Analytics, Lee Li, CanAm Enterprises reported in February 2026 that post-RIA I-526E approval rates for Regional Center investors ran above ninety percent in 2025, against roughly twenty-five percent for standalone I-526 investors. “For 2025, we have a big spike in approvals,” Li observed; “the approval rate is still over 90%, which I think is very high.” That gap is the clearest single measure of what project-level methodological vetting is worth. The Regional Center program itself is currently authorized through September 30, 2027, which is the horizon against which all of this is being underwritten.

What a defensible count looks like

Strip the subject to its essentials and a defensible EB-5 job count has four properties, each of which a skeptical reviewer can check. It rests on an accepted input-output model — RIMS II, IMPLAN, or REMI — rather than a bespoke device. It is fed realistic, documented expenditure and revenue inputs, tied precisely to the construction contract, the budget, and the pro forma, because the model is only as good as what goes into it. It uses the correct geographic and industry selections, indexed to the right dollar-year. And it stays causally close to the direct spending, never straying so far from a verifiable dollar that the nexus between the investment and the job becomes, in the agency’s word, tenuous. The 2018 tenant-occupancy retirement is the proof that USCIS enforces that last principle, and the post-RIA I-956F regime is the proof that it now enforces it before an investor ever files.

For the analyst and the issuer, the lesson is that economic-impact modeling is a rigorous analytical discipline, not a rubber stamp. The multiplier is a powerful engine, and in a Regional Center project it will do most of the hiring on paper. But the number it produces is defensible only to the extent the inputs behind it are documented and the causation is clean. The deliverable is not the job count. It is the audit trail from a verified dollar of spend to a counted job — the record that answers, line by line, the one question USCIS always asks: what makes you say this investment created that job?