Mass and Indiscriminate Immigration as Indirect Corporate Welfare

Executive summary Link to heading

The thesis examined here is analytically plausible, but only under specific institutional conditions. Immigration does not, by itself, amount to “corporate welfare.” The argument becomes strongest when large inflows are combined with one or more of the following policy features: permits or statuses that tie workers to particular employers; weak enforcement of labor standards; recruitment systems that generate debt or legal dependency; sectoral concentration in labor-intensive industries; and fiscal arrangements that socialize a meaningful share of education, housing, health, or infrastructure costs while employers capture the immediate labor-cost gains. Under those conditions, immigration can function as an indirect subsidy to firms by expanding labor supply, weakening worker exit options, slowing the wage response to scarcity, and shifting some costs to workers, households, or the public sector. \[1\]

The evidence does not support a simple universal claim that immigration broadly or always harms native workers. The National Academies concluded that the long-run effect of immigration on native wages and employment overall is “very small,” with any negative effects more likely concentrated among prior immigrants and some lower-education groups. The U.K. Migration Advisory Committee likewise found little overall employment effect and only small wage effects, tilted against lower-wage workers and toward higher-wage workers. The Mariel Boatlift remains a classic case where a large sudden inflow did not generate large average wage losses in Miami, although later reappraisals found more negative effects for very low-skill workers depending on the comparison group and specification. \[2\]

Accordingly, the strongest version of the thesis is conditional rather than universal: mass and indiscriminate immigration can function as indirect corporate welfare chiefly when the migration regime is structured in ways that deliver firms a more compliant, lower-mobility, or lower-cost labor force than would otherwise prevail in a tighter labor market. This is clearest in tied-visa systems such as Gulf kafala arrangements, in agricultural guestworker programs, and in segments of labor markets where unauthorized or highly dependent workers are concentrated. It is weaker where immigrants have broad mobility rights, strong labor protections, or are concentrated in high-skill sectors that also generate innovation, entrepreneurship, and tax revenue. \[3\]

A balanced conclusion is therefore warranted. The available evidence supports the proposition that some migration systems and some labor-market segments operate as state-enabled cost-reduction mechanisms for employers. It does not support the blanket claim that all large-scale immigration is corporate welfare in every context. The empirical burden is to identify when immigration primarily expands productive capacity and demand, and when it instead mainly supplies firms with downward wage flexibility, weaker labor resistance, and publicly externalized costs. \[4\]

Scope, definitions, and methodology Link to heading

This report uses the following analytic definitions. Mass immigration means inflows that are large enough, relative to a national labor market, a local labor market, or a specific occupation, to affect bargaining conditions, labor-market adjustment, or public-service demand. Indiscriminate immigration means admissions or tolerated labor inflows that are not closely calibrated by labor standards, skill composition, sectoral need, absorptive capacity, or local fiscal burden. Corporate welfare is used here in a functional sense: policy-created value transfers or risk transfers that benefit firms by lowering private costs, increasing labor leverage, or socializing part of the input cost structure.

The central question is not whether immigration can raise output. It often does. The narrower question is whether some migration regimes allow firms to capture gains that would otherwise accrue to workers through higher wages, better conditions, greater investment in mechanization, or stronger bargaining power. That question is distinct from whether immigration benefits consumers, raises GDP, or helps offset aging. \[5\]

The report prioritizes primary and near-primary materials: U.S. Department of Labor program data, USDA and USCIS materials, National Academies and Federal Reserve research, U.K. Migration Advisory Committee reports, ILO publications, OECD analyses, World Bank work, and original academic papers by Card, Borjas, Dustmann and coauthors, Clemens-Lewis-Postel, Hunt, and others. \[6\]

The main limitations are familiar in this literature. Immigration effects are often small on average but nontrivial at the lower tail of the wage distribution. Fiscal estimates are highly sensitive to how public goods are allocated, whether children’s education is treated as a current cost or long-run investment, and whether analysis is static or lifetime-based. Several official reports explicitly warn that measured burdens vary with assumptions and that immigrant cohorts differ sharply by route, age, skill, and labor-force attachment. \[7\]

Causal mechanisms Link to heading

The thesis can be organized around six causal channels. None operates mechanically in every setting, but each is well grounded in the literature.

Labor-supply effects and wage discipline Link to heading

An increase in labor supply can reduce wage pressure in affected occupations, especially where migrants are close substitutes for existing workers or prior immigrants. The strongest empirical pattern is not a large aggregate wage collapse, but a distributional effect: compression at the bottom, small average effects overall, and gains or neutrality higher up. The U.K. MAC concluded that migration is not a major determinant of overall wages but found some evidence of small negative effects for lower-skilled workers and benefits for higher-skilled workers. In the same report, Nickell and Saleheen’s estimates imply that a 10-percentage-point rise in the migrant share of semi- and unskilled service occupations is associated with about a 1.9% reduction in pay in that group. Dustmann, Frattini, and Preston similarly found negative effects at the bottom of the U.K. wage distribution and positive effects above it. \[8\]

For the United States, the National Academies found very small overall long-run effects on native wages and employment, with potential losses most likely for prior immigrants and native-born high-school dropouts. The Mariel literature shows why the debate persists: Card’s original design found little effect from a 7% labor-force shock in Miami, while Borjas’s reappraisal found 10% to 30% wage declines among high-school dropouts depending on the comparison set. The corporate-welfare interpretation is therefore strongest not where broad average effects are absent, but where a policy creates a stable inflow into narrow labor-market segments that disciplines wages there. \[9\]

Legal precarity and employer dependence Link to heading

Where workers’ residence, renewability, or ability to remain lawfully present depends on a specific employer, exit becomes costly and formal bargaining power weakens. The ILO notes that tied-status systems deter complaints because some workers depend on employers not only for wages but also for lawful residence, and may fear becoming irregular, detained, or deported if they challenge abuse. In Gulf kafala systems, the ILO states directly that the employer-linked sponsorship model severely limits workers’ ability to leave employers and creates risks of exploitation and forced labor. \[10\]

This mechanism exists on a spectrum. U.S. H-1B is less restrictive than kafala because portability exists: an eligible H-1B worker may begin with a new employer once a nonfrivolous petition is filed, and laid-off workers may have up to a 60-day grace period to act. Even so, CRS reports that many H-1B workers—especially from India—remain in temporary status for many years while waiting for employment-based green cards, and Bipartisan Policy Center analysis argues that the transition from H-1B to permanent residence can tether workers to employers for more than a decade, discouraging the exercise of rights. Hunt’s NBER paper likewise finds that green-card processing and unemployment risk constrain mobility even though transfers are legally possible. \[11\]

Anti-union and collective-action effects Link to heading

Temporary and dependent migrants often face special barriers to organizing. The NLRA excludes “any individual employed as an agricultural laborer,” leaving farmworkers outside the core federal private-sector collective-bargaining statute. The ILO documents employer retaliation, denial of organizer access to employer-controlled housing, blacklisting, and the added risk that dismissal can trigger deportation where visas are employer-linked. \[12\]

The recent U.S. H-2A experience demonstrates the relevance. DOL’s 2024 H-2A rule expanded anti-retaliation protections, including protections for workers who consult attorneys or assert rights, but those protections were later blocked in part by courts and then suspended in enforcement, underscoring how fragile organizing protections are in a guestworker-dense sector. Where workers are temporary, geographically isolated, housed by employers, or easily replaced, the labor-supply effect is amplified by an organizational effect: firms gain not only more labor, but more governable labor. \[13\]

Anti-automation and anti-investment effects Link to heading

If labor scarcity would otherwise induce mechanization or process innovation, abundant low-cost labor can delay that adjustment. This is one of the strongest channels for the “indirect subsidy” interpretation because it implies a transfer not merely from workers to firms, but from longer-run productivity upgrading toward continued reliance on cheap labor.

The evidence is substantial. Following the end of the Bracero program, Clemens, Lewis, and Postel found no large labor-market improvement for domestic farm workers, while accompanying evidence pointed to mechanization and production changes instead. San’s later work on the same shock finds that more exposed crops experienced a sharp increase in innovation after Bracero exclusion; a standard-deviation increase in exposure raised patenting by about 70.7% relative to the pre-period. For Denmark, Mann finds that immigration and automation are substitutes: a 1-percentage-point increase in the share of non-Western migrants reduced firms’ probability of robot adoption by about 7%. Austria’s recent firm-level evidence is more mixed and points to both substitution and complementarity depending on origin and education, but it still frames migration and automation as alternative responses to labor shortages. Malaysia’s World Bank review likewise records sustained policy concern that reliance on low-skilled migrant labor weakens incentives to automate, even while also noting that the empirical evidence is mixed and context dependent. \[14\]

Externalized public costs Link to heading

If employers capture the gain from lower labor costs while public authorities absorb schooling, health, housing, infrastructure, or enforcement costs, part of labor reproduction is socialized. This is the most literal sense in which immigration can resemble indirect corporate welfare.

The best evidence is from fiscal-incidence work. Dallas Fed’s summary of the National Academies shows that when migrants are assigned the marginal cost of public goods, the long-run fiscal effect is positive overall, but short-run and cross-sectional burdens still fall disproportionately at state and local levels, largely because those governments finance education. Urban Institute’s state-and-local analysis reaches the same conclusion: relative burdens vary by tax structure and demographics, but dependent children and K–12 education explain much of the gap, and state-local effects differ from federal effects. The implication is not that immigration is fiscally negative per se. Rather, benefits and burdens can be split across levels of government and across time, allowing firms and federal coffers to benefit while localities bear adjustment costs. \[15\]

Sectoral concentration and firm sorting Link to heading

A final channel is concentration in low-paying firms and sectors. OECD’s cross-country linked employer-employee evidence finds that immigrants earn 34% less than comparable native-born workers at entry on average across 15 OECD countries, and that about 63% of the initial earnings gap is explained by concentration in lower-paying firms and sectors. This matters for the thesis because it suggests that the firm-level incidence of immigration is not random. Lower-paying firms and sectors are disproportionately able to recruit workers whose outside options are weak, whose credentials are not yet portable, or whose bargaining position is degraded by immigration status and labor-market segmentation. \[16\]

The flowchart summarizes the conditional thesis supported by the evidence above: the argument is strongest when immigration affects bargaining power, not merely headcounts. \[17\]

Empirical evidence on the main claims Link to heading

Claim testing Link to heading

The evidence base supports some claims strongly, others only partially.

Comparative table of migration regimes Link to heading

Fiscal-incidence table Link to heading

Key studies table Link to heading

Case studies across countries and sectors Link to heading

U.S. H-1B and the skilled-labor hierarchy Link to heading

The U.S. H-1B program is often misdescribed as either a pure labor-market necessity or a pure wage-depression program. The evidence suggests a more mixed reality. Formally, employers must pay the greater of the actual or prevailing wage, and workers can port status to a new employer. In practice, however, the program channels large numbers of workers into highly concentrated sectors and firms. In FY 2024, DOL certified 790,845 LCA positions, with software developers alone accounting for 32.4% of positions, and a small set of major firms accounting for large shares. At the same time, the employment-based green-card queue can keep workers in temporary status for years or decades, especially those from India, reducing bargaining power even when nominal portability exists. This is not identical to wage subsidy. It is closer to state-mediated monopsony relief for employers who gain access to skilled labor without facing the full quit threat present in an unconstrained labor market. \[30\]

U.S. agriculture, H-2A, and the bracero legacy Link to heading

Agriculture provides the clearest U.S. example of the thesis. USDA reports that H-2A certifications rose from just over 48,000 in FY 2005 to about 385,000 in FY 2024. In FY 2023, DOL certified 378,513 H-2A positions, with 84.5% in crop, nursery, and greenhouse labor. USDA also reports that in 2020–22, 42% of hired crop farmworkers lacked work authorization, meaning U.S. crop agriculture still depends heavily on legally vulnerable labor outside the guestworker stream as well. \[31\]

The corporate-welfare logic here is multi-layered. Employers receive labor where domestic recruitment is difficult, but the workforce is unusually dependent: employer-provided housing, transportation, contract-specific work, rapid termination reporting, rural isolation, and exclusion from core collective-bargaining law all reduce worker leverage. GAO found that H-2A violations accounted for 54% of back wages assessed to all agricultural employers over the six-year period it examined, which is a striking indicator of enforcement problems in a supposedly regulated legal-labor channel. \[32\]

The bracero literature deepens the point. The Bracero program was defended in part on the idea that Mexican labor was necessary and opposed on the idea that it depressed domestic farm wages. Yet when the program ended, Clemens, Lewis, and Postel found limited gains for domestic farm labor, while mechanization and crop reorganization followed. San’s estimates show a substantial increase in patenting in more exposed crops after bracero exclusion. The implication is not that guest labor always lowers wages dramatically. It is that easy access to low-cost migrant labor can postpone the wage and innovation adjustments that labor scarcity would otherwise force. \[33\]

Mariel Boatlift as a stress test of the thesis Link to heading

The Mariel Boatlift is indispensable because it prevents overgeneralization. About 125,000 Cubans arrived in 1980, and roughly half settled in Miami, expanding the local labor force by about 7%. Card found virtually no significant effect on less-skilled wages or unemployment, and emphasized Miami’s prior immigrant adaptation, industry structure, and the possibility of other labor-market adjustments. For a general thesis that “mass immigration lowers wages and therefore subsidizes capital,” Mariel is the canonical challenge. \[29\]

Borjas’s later reappraisal, however, argued that the correct comparison group is the least-educated workers, among whom wages fell sharply. The dispute suggests a useful refinement: the thesis is strongest when the relevant labor market is narrowly defined, when immigrants are close substitutes for incumbents, and when institutions do not let wages, technology, and migration margins adjust smoothly. Mariel does not disprove the thesis; it shows that the thesis fails when local demand, prior immigrant experience, or other equilibrating forces absorb the shock. \[34\]

U.K. and Germany Link to heading

The British evidence points to a mixed but intelligible pattern. MAC finds little aggregate employment effect from EEA migration and small overall wage effects, but also some lower-tail wage losses and reduced prices in personal services, particularly in middle- and lower-skilled services. A rise in migrant shares also appears to have put upward pressure on house prices where supply was constrained. This combination is central to the corporate-welfare interpretation: employers and consumers may gain from cheaper services, while lower-wage workers and housing markets bear more of the adjustment. \[35\]

Germany reveals a different margin. Its recent policy debate is dominated by labor shortages and aging, and OECD emphasizes those shortage pressures. At the same time, IAB finds that employment bans, place-based residence restrictions, and reception-center residence are all associated with lower refugee employment probabilities. Refugee employment rises substantially over time—nearly two-thirds of refugees in Germany for seven years were in paid employment in the 2022 survey—but legal frictions measurably slow integration. This does not fit a simple “cheap labor” story as neatly as Gulf or H-2A. It does, however, show that migration policy can produce worker segmentation that matters for labor-market power and wage trajectories. \[36\]

Gulf migrant labor and low-skill labor markets in Malaysia Link to heading

If one seeks the strongest real-world example of migration operating as state-backed labor discipline for capital, the Gulf sponsorship model is the benchmark. The ILO states that kafala severely limits workers’ opportunity to leave employers, creates risks of forced labor and exploitation, and allows employers in some systems to trigger loss of status through “absconding” claims. The report’s own recommendation is that migrant entry, residence, and work authorization should not be tied to a specific employer. That recommendation is telling: if the employer tie is the problem, employer leverage is not incidental but structural. \[24\]

Malaysia offers a more mixed developing-country case. The World Bank reports that migrant workers are heavily concentrated in low-skilled jobs and in agriculture, manufacturing, and construction; in 2022, migrant workers accounted for 44.7% of employment in low-skilled jobs. The same report records active domestic concern that low-skilled migration weakens innovation incentives and depresses wages, but its own bottom line is more nuanced: the negative wage effects found in the literature are generally small, some studies find complementarities with Malaysian workers, and evidence on migration’s relationship to automation remains limited. This is a useful reminder that the “indirect corporate welfare” mechanism can be real even when aggregate welfare effects are mixed. \[37\]

Counterarguments and alternative explanations Link to heading

The strongest counterargument is the demand-side one. Immigrants are not only workers; they are also consumers, renters, entrepreneurs, and founders. Migration Observatory summarizes the British evidence by noting that the number of jobs is not fixed and that migrants can increase labor demand as well as labor supply. NBER work on immigrant-founded firms finds higher rates of innovation and patenting, while related research finds that high-skilled immigrants make large contributions to U.S. innovation. In the U.K., skilled-worker routes also appear fiscally positive on a lifetime basis. These points limit any attempt to treat firms’ labor-cost gains as the whole story. \[38\]

A second counterargument is that many observed harms are better explained by domestic institutional weakness than by immigration itself. Weak union law, low minimum wages, lax housing supply, fragmented local finance, and poor labor inspection can all magnify the adverse effects of migration or even create them. MAC explicitly warns that migration’s effect on house prices cannot be understood apart from planning and building policy. The National Academies and Urban both stress that measured fiscal burdens are highly sensitive to institutional accounting choices. On this view, immigration may be the proximate shock, but the real policy failure lies in labor-market governance and fiscal federalism. \[39\]

A third counterargument is that some sectors genuinely face labor shortages that domestic workers will not fill at existing wage-productivity combinations. USDA documents rising use of H-2A and persistent labor scarcity, while Germany’s official debate centers on aging and shortages. In such settings, it is possible to interpret migrant labor as a necessary input to maintain output rather than as a subsidy to employers. The rejoinder, however, is that “shortage” can mean either a true absence of labor at any feasible price or a reluctance to pay the wage, improve conditions, or invest in mechanization needed to attract labor. The bracero and automation evidence shows why those meanings should not be conflated. \[40\]

The most balanced position, then, is not that immigration is harmless or that it is inherently corporate welfare. It is that the incidence of immigration depends on route design, labor-law enforcement, sectoral exposure, and fiscal architecture. Where those institutions favor employers, the “corporate welfare” frame becomes empirically stronger. Where migrants are mobile, highly skilled, entrepreneurial, and protected by strong labor standards, the frame becomes weaker and can be misleading. \[41\]

Policy implications and testable hypotheses Link to heading

If the policy objective is to prevent migration from functioning as indirect corporate welfare, the first principle is to reduce employer-created dependency. The ILO’s recommendations are direct: decouple residence from a single employer, allow workers to change jobs without losing status, let workers hold or renew their own permits where possible, and eliminate “absconding” mechanisms. In the U.S. context, that implies stronger portability and anti-retaliation rules for guestworkers, more credible labor inspection, and legal pathways that do not force workers to remain with a sponsor for years to preserve their place in a permanent-residence queue. \[19\]

A second implication is that fiscal federalism matters. If migration has positive federal effects but negative local incidence, central governments should compensate receiving localities for schooling, housing, and infrastructure costs rather than allowing firms and federal revenues to benefit while municipal systems absorb the strain. The Urban and Dallas Fed work implies that the right policy question is not “Is immigration fiscally positive?” in the abstract, but “Positive for whom, at what level of government, and over what horizon?” \[15\]

A third implication concerns technology and labor incentives. Where migration policy is explicitly used to relieve labor scarcity in sectors with low wages and weak productivity growth, policymakers should expect weaker incentives for mechanization and job redesign. One policy response would be to pair labor inflows with automation targets, training standards, or rising wage floors, so that migration fills transitional shortages rather than permanently substituting for capital deepening. The bracero, Danish, and Malaysian evidence suggests that this is at least a serious hypothesis, not a speculative one. \[42\]

The following hypotheses are especially testable in future research:

Open questions and limitations Link to heading

The evidence is strongest for tied or highly dependent migration systems and weaker for broad claims about immigration in general. Several important areas remain uncertain. H-1B, for example, is clearly more mobile than Gulf sponsorship or H-2A, so the appropriate label is not “bonded labor” but “partial dependency.” Malaysian and Austrian evidence on automation remains mixed. Fiscal evidence is sensitive to assumptions about public goods, public debt, and children’s future tax contributions. And some of the largest adaptation margins—native mobility, firm relocation, credential recognition, and entrepreneurial spillovers—are difficult to measure cleanly. \[43\]

The most defensible final formulation is therefore conditional and institutional: mass and indiscriminate immigration can function as a form of indirect corporate welfare when the state supplies firms with labor under conditions of excess labor supply, weak worker mobility, weak labor-law enforcement, or publicly socialized adjustment costs. The proposition is strongest in guestworker, tied-visa, and informal or semi-formal labor systems; weakest in high-mobility, high-skill, high-innovation migration regimes with strong worker protections. \[44\]

\[1\] \[7\] \[15\] \[26\] Working Paper No. 1704 - New Findings on the Fiscal Impact of Immigration in the United States - Dallas Fed

https://www.dallasfed.org/-/media/documents/research/papers/2017/wp1704.pdf

\[2\] \[4\] \[5\] \[9\] The Economic and Fiscal Consequences of Immigration

https://www.nationalacademies.org/projects/DBASSE-CNSTAT-13-03/publication/23550

\[3\] \[17\] \[19\] \[24\] \[44\] Sponsorship reform and internal labour market mobility for migrant workers in the Arab States | International Labour Organization

https://www.ilo.org/resource/other/sponsorship-reform-and-internal-labour-market-mobility-migrant-workers-arab

\[6\] \[20\] \[23\] dol.gov

https://www.dol.gov/sites/dolgov/files/ETA/oflc/pdfs/H-2A_Selected_Statistics_FY2023_Q4.pdf

\[8\] \[18\] \[35\] \[39\] assets.publishing.service.gov.uk

https://assets.publishing.service.gov.uk/media/5ba26c1de5274a54d5c39be2/Final_EEA_report.PDF

\[10\] itcilo.org

https://www.itcilo.org/sites/default/files/inline-files/Final_ILO%20FRI%20Access%20to%20Justice%20Working%20Paper.pdf

\[11\] FAQs for Individuals in H-1B Nonimmigrant Status

https://www.uscis.gov/archive/faqs-for-individuals-in-h-1b-nonimmigrant-status?utm_source=chatgpt.com

\[12\] National Labor Relations Act | National Labor Relations Board

https://www.nlrb.gov/guidance/key-reference-materials/national-labor-relations-act

\[13\] Federal Register/Vol. 89, No. 83/Monday, April 29, 2024/ …

https://www.govinfo.gov/content/pkg/FR-2024-04-29/pdf/2024-08333.pdf?utm_source=chatgpt.com

\[14\] Immigration Restrictions as Active Labor Market Policy

https://www.aeaweb.org/articles?id=10.1257%2Faer.20170765&utm_source=chatgpt.com

\[16\] \[41\] Immigrant integration: The role of firms: International Migration Outlook 2025 | OECD

https://www.oecd.org/en/publications/2025/11/international-migration-outlook-2025_355ae9fd/full-report/immigrant-integration-the-role-of-firms_db745b4c.html

\[21\] \[42\] Labor Supply and Directed Technical Change: Evidence from the Termination of the Bracero Program in 1964

https://mulysan.github.io/San_bracero.pdf

\[22\] \[30\] dol.gov

https://www.dol.gov/sites/dolgov/files/ETA/oflc/pdfs/LCA_Selected_Statistics_FY2024_Q4.pdf

\[25\] Labor market integration of refugees: Improved institutional settings promote employment

https://doku.iab.de/kurzber/2024/kb2024-10_en.pdf

\[27\] State and Local Fiscal Effects of Immigration

https://www.urban.org/sites/default/files/publication/90796/state_and_local_fiscal_effects_of_immigration.pdf

\[28\] Migration Advisory Committee (MAC) annual report, 2025 (accessible) - GOV.UK

https://www.gov.uk/government/publications/migration-advisory-committee-annual-report-2025/migration-advisory-committee-mac-annual-report-2025-accessible

\[29\] davidcard.berkeley.edu

https://davidcard.berkeley.edu/papers/mariel-impact.pdf

\[31\] \[40\] Farm Labor | Economic Research Service

https://www.ers.usda.gov/topics/farm-economy/farm-labor

\[32\] Fact Sheet #26: Section H-2A of the Immigration and Nationality Act (INA) | U.S. Department of Labor

https://www.dol.gov/agencies/whd/fact-sheets/26-H2A

\[33\] nber.org

https://www.nber.org/system/files/working_papers/w23125/w23125.pdf

\[34\] w21588 copy.pdf

https://www.nber.org/system/files/working_papers/w21588/revisions/w21588.rev0.pdf

\[36\] Addressing skilled labour shortages: OECD Economic …

https://www.oecd.org/en/publications/2025/06/oecd-economic-surveys-germany-2025_b395dc9b/full-report/addressing-skilled-labour-shortages_9edb78e6.html?utm_source=chatgpt.com

\[37\] World Bank Document

https://documents1.worldbank.org/curated/en/099110424210537110/pdf/P181093112d5840451b60a12134739d28b3.pdf

\[38\] The Labour Market Effects of Immigration - Migration Observatory

https://migrationobservatory.ox.ac.uk/resources/briefings/the-labour-market-effects-of-immigration/

\[43\] nber.org

https://www.nber.org/system/files/working_papers/w23529/w23529.pdf