What methods do economists use to estimate taxes paid by unauthorized workers?
Executive summary
Economists estimate taxes paid by unauthorized (undocumented) workers by combining demographic and tax data, applying behavioral assumptions like the share who file or have payroll taxes withheld (commonly 50–75%), and modeling consumption, payroll and income tax incidence; one recent synthesis estimated nearly $97 billion in federal, state and local taxes for 2022 using these methods [1] [2]. Key inputs include counts of likely unauthorized adults, data on ITIN holders and withholding/filing rates, and models that allocate sales and business-paid consumption taxes ultimately borne by consumers and workers [2] [1] [3].
1. “Counting the invisible”: population and taxpayer proxies
Economists begin by estimating how many unauthorized workers live and work in the U.S. and then link that population to tax records or proxies; for example, analysts use active ITIN counts and estimates of unauthorized population by age and state as core inputs—ITEP cited ITIN data and demographic estimates in producing its nationwide tax estimate [2] [1]. This step is essential because official employment records exclude many unauthorized workers, so researchers rely on demographic series and administrative proxies rather than direct, universal payroll data [2] [1].
2. “Whose tax records count?”: using ITINs, fake SSNs and filing behavior
Researchers use observed filings tied to Individual Taxpayer Identification Numbers (ITINs) and patterns of social security number misuse to infer tax payments; ITEP’s 2024 work used ITIN data to scale up tax estimates, and fact-checkers note that some undocumented workers use ITINs or borrowed SSNs and still pay into Social Security and Medicare [2] [4]. Analysts must also allow for under-the-table employment and fake SSN use, which create measurement noise and require assumptions about what fraction of income was taxed or reported [4].
3. Behavioral assumptions: withholding and filing probabilities
A crucial modeling choice is the fraction of unauthorized workers who have payroll taxes withheld and who file returns. Studies often assume wide ranges—in one literature-based example analysts used a 50–75% probability that an unauthorized worker has payroll and income taxes withheld and a separate 50–75% filing probability; those probabilities materially affect revenue estimates [3]. Sensitivity to these behavioral parameters is why different credible studies can produce different headline numbers.
4. Tax incidence modeling: beyond income—sales, excise and business taxes
Good estimates go beyond payroll and income taxes to attribute consumption taxes paid directly by undocumented consumers and consumption taxes borne indirectly when businesses pay taxes on inputs; ITEP’s model explicitly allocates sales and excise taxes as well as the share of business-borne consumption taxes that ultimately come from undocumented immigrants [1]. That broader incidence approach pushes totals higher than a narrow payroll-only accounting [1].
5. Reconciliation and calibration to administrative aggregates
Researchers preserve micro-level heterogeneity (wage distribution, refund patterns, filing status) but calibrate their modeled aggregates to match observed IRS and refund statistics so totals look realistic; for example, one project adjusted withheld-wage estimates to align with 2023 IRS aggregate refund and filing data while keeping rank ordering from initial wage imputations [3]. Calibration to administrative aggregates reduces some model error but depends on the quality and relevance of the calibration data [3].
6. Sensitivity, scenarios and policy counterfactuals
Authors present ranges and scenarios because small changes in filing or employer behavior change revenues a lot; Budget Lab’s exercise shows an IRS-ICE data-sharing shock could reduce revenues by billions depending on whether withholding and filing fall, illustrating how scenario analysis exposes fiscal sensitivity to enforcement and legal-status changes [3]. ITEP and other groups supply central estimates and discuss uncertainty ranges rooted in their behavioral assumptions [1] [2].
7. Sources of disagreement and hidden assumptions
Disagreement among studies often traces to three hidden levers: (a) the unauthorized population denominator and its age/workforce composition; (b) assumed filing/withholding rates and the prevalence of “under-the-table” pay; and (c) whether models include indirect taxes borne via businesses. Those assumptions are rarely neutral—advocacy or policy goals can subtly shape choice of parameters and which taxes are counted, a dynamic visible when different organizations emphasize different components [1] [2] [3].
8. Bottom line for readers and policymakers
Available, recent reporting shows that comprehensive estimates using these methods reached roughly $97 billion in federal, state and local taxes for 2022, but that figure depends on the population and behavioral assumptions described above [2] [1]. Evaluations should judge both the data sources used (ITINs, administrative aggregates) and the assumptions about withholding, filing and consumption tax incidence—changes to any of those inputs produce materially different fiscal conclusions [3] [1].