Full text of Digital Enforcement : Effects of E-Verify on Unauthorized Immigrant Employment and Population : A Special Report of the Federal Reserve Bank of Dallas

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

digital enforcement

federal reserve bank of dallas

D I G I TA L

ENFORCEMENT

Effects of E-Verify on Unauthorized
Immigrant Employment and Population
1

A special report of the Federal Reserve Bank of Dallas
September 2017

digital enforcement

federal reserve bank of dallas

TABLE OF CONTENTS
Executive Summary.......................................................................................................................................................... 1
Overview.................................................................................................................................................................................2
Background on E-Verify ................................................................................................................................................ 3
State-Level Policies..................................................................................................................................................................................................................... 4
Compliance Mechanisms, Take-Up Rates................................................................................................................................................................ 4

Methodology....................................................................................................................................................................... 5
Modeling E-Verify Requirements’ Impact ..............................................................................................................................................................5
Synthetic Control Method....................................................................................................................................................................................................6
Data Source...................................................................................................................................................................................................................................... 7
Previous Research....................................................................................................................................................................................................................... 7

Likely E-Verify Impact Observed in Several States.......................................................................................... 8
Alabama..............................................................................................................................................................................................................................................9
Arizona...............................................................................................................................................................................................................................................10
Georgia.................................................................................................................................................................................................................................................11

TABLE
OF
CONTENTS

Mississippi.........................................................................................................................................................................................................................................12

Utah...................................................................................................................................................................................................................................................... 13
North Carolina and South Carolina ............................................................................................................................................................................ 14

Unauthorized Immigrant Employment More Affected than Unauthorized Population..........15
Economic Effects and Policy Implications.........................................................................................................16
Conclusion............................................................................................................................................................................16
Appendix A: State E-Verify Requirements.........................................................................................................17
Appendix B: Data and Methods...............................................................................................................................19

Data Source.................................................................................................................................................................................................................................... 19
Interpretation of the Data................................................................................................................................................................................................... 19
Assumptions of the Synthetic Control Method................................................................................................................................................. 19
Statistical Significance........................................................................................................................................................................................................... 19

Notes......................................................................................................................................................................................23
ABOUT THE AUTHORS
Pia M. Orrenius is a vice president and senior economist in the Research Department of
the Federal Reserve Bank of Dallas; Madeline Zavodny is a professor of economics at the
University of North Florida.

ACKNOWLEDGMENTS
Madeline Zavodny thanks The Pew Charitable Trusts for support for this project. The views expressed are
those of the authors and do not necessarily reflect the views of The Pew Charitable Trusts, the Federal
Reserve Bank of Dallas or the Federal Reserve System. The authors thank Sarah Bohn, Daniel Costa and
B. Lindsay Lowell for reviewing a draft of this report, and Adam Hunter, Karina Shklyan, Nicole Svajlenka
and Michele Waslin for helpful comments and assistance.

digital enforcement

federal reserve bank of dallas

e-verify digitally checks workers’ documentation against official records. Although the federal
government requires its own agencies and contractors to use E-Verify, it does not require that other
employers use it. A number of states have stepped
into this void and required that some or all employers use E-Verify. As a result, E-Verify has become increasingly prevalent, with half of newly hired workers nationwide vetted through the system in 2015.
With discretion left to the states, there is large
regional variation in E-Verify laws and usage.
Twenty-one states required use of the program as
of December 2016. Most states only require public-sector employers or contractors to use E-Verify;
only eight states have universal mandates covering
all employers.
Alabama, Arizona, Georgia, Mississippi, North
Carolina, South Carolina, Tennessee and Utah have
mandated that virtually all employers use the system to screen new hires. This study examines the
impact of E-Verify requirements on the number of
likely unauthorized immigrants living and working in seven of those states. (Tennessee adopted its
universal E-Verify mandate relatively recently and
isn’t included in this study.)
Our analysis indicates that the number of unauthorized immigrants and/or unauthorized immigrant workers fell below what would have been
expected absent E-Verify in five states—Alabama,
Arizona, Georgia, Mississippi and Utah. This is
based on counterfactual projections using states
with characteristics resembling those studied.
In addition to these relative declines, the actual
numbers of likely unauthorized immigrants living
and working in Arizona and Mississippi were, as of
2015, below the levels at the time of implementation in 2008. In Alabama and Utah, they are about
the same or slightly higher than when those states’
laws took effect in 2012 and 2010, respectively.
Meanwhile, four years after Georgia implemented
E-Verify in 2012, there were fewer than the projected number of unauthorized immigrant workers but
no measurable change in the unauthorized immigrant population. Finally, there was no statistically
discernable change in the number of likely unauthorized immigrants living or working in North Carolina and South Carolina.

EXECUTIVE
SUMMARY

digital enforcement

federal reserve bank of dallas

DIGITAL
ENFORCEMENT

EFFECTS OF E-VERIFY ON UNAUTHORIZED
I M M I G R A N T E M P L OY M E N T A N D P O P U L AT I O N

OVERVIEW
the 1986 immigration reform and Control Act
prohibited employers from knowingly hiring unauthorized workers. In the ensuing decades, government at all levels and private employers have pursued
various strategies to ensure a legal workforce. One
tool created as part of those efforts is an online federal system, E-Verify. It enables employers to digitally
check eligibility documents provided by new hires
against federal records. The federal government requires its own agencies and contractors to use E-Verify, with requirements for other employers left to the
discretion of the states. Firms not subject to a government rule may still choose to use the system. As of
December 2016, at least some employers in 21 states
had to use E-Verify.
In 14 states—Colorado, Florida, Idaho, Indiana,
Louisiana, Michigan, Minnesota, Missouri, Nebraska,
Oklahoma, Pennsylvania, Texas, Virginia and West
Virginia—E-Verify requirements only apply to certain
public-sector agencies or contractors. Another eight
states—Alabama, Arizona, Georgia, Mississippi, North
Carolina, South Carolina, Tennessee and Utah—currently have universal mandates that require all or nearly all employers to use the system to screen new hires.
States have tended to phase in requirements, beginning
with larger employers and then extending to smaller
ones; some states with universal mandates exempt very
small employers. Louisiana requires employment verification but does not mandate the use of E-Verify for
that purpose. Use of the system has increased steadily
and, in 2015, the share of newly hired workers nationwide run through the system reached 50 percent.
This report estimates the effects of state E-Verify
requirements on the number of likely unauthorized
immigrants living and working in seven states with

universal requirements. (Tennessee’s requirement
began too recently to examine its effects.) It contrasts
the actual changes in population and employment
levels over time with projections of what would have
happened in each state absent the E-Verify requirement. The analysis includes testing of statistical significance, or whether estimated effects are likely to
be distinguishable from zero. Effects are examined
over a range of three to eight years, depending on the
elapsed time between when each state’s law took effect and the latest available data. Compared with the
projections, the analysis found:
●● Reductions in the number of likely unauthorized immigrants living and working in four states: Alabama,
Arizona, Mississippi and Utah.
□□ Alabama’s population of likely unauthorized
immigrants was 10 percent below the projection three years after its mid-2012 implementation, while its number of likely unauthorized
immigrant workers was 57 percent below the
projected level.
□□ Eight years after Arizona’s 2008 implementation, its population of likely unauthorized
immigrants and number of likely unauthorized
immigrant workers were 28 percent and 33 percent below projected levels, respectively.
□□ Mississippi implemented universal E-Verify in
mid-2008, and seven years later, its population
of likely unauthorized immigrants was 70 percent below its projected level, while its number
of likely unauthorized workers was 83 percent
below projection.
□□ Five years after Utah’s mid-2010 implementation, its population of likely unauthorized
immigrants and number of likely unauthorized
immigrant workers were 30 percent and 34 percent below their projected levels, respectively.

digital enforcement

federal reserve bank of dallas

bility to legally work in the United States by comparing information provided by a new hire with federal
records.1 The digital system has its roots in the Immigration Reform and Control Act of 1986, which made it
illegal to knowingly employ people who are not allowed
to work in the United States and required that employers review eligibility documents for all new hires.
To address concerns about widespread availability
of fraudulent eligibility documents, Congress mandated the creation of a system to authenticate workerprovided records as part of the Illegal Immigration
Reform and Immigrant Responsibility Act of 1996.2
E-Verify’s precursor, the Basic Pilot Program, became
available in 1997 to employers in five states with large
immigrant populations—California, Florida, Illinois,
New York and Texas. It expanded to Nebraska in 1999
and to all other states in 2003. The program was renamed E-Verify in 2007.
The federal government has used E-Verify to check
the work eligibility of its employees since 2007 and
has required certain contractors to do so since 2009.
States and localities have authority to regulate the use
of E-Verify by all other public and private employers.3
The E-Verify system confirms work eligibility by
comparing documentation that newly hired workers
provide to their employers with federal government
records. It is generally intended only for checking the
status of new hires; certain government contractors
must apply it retroactively to existing employees. The
process, which employers are only permitted to use
after an applicant accepts an offer of employment,
begins with online entry of information provided on
Employment Eligibility Verification Form I-9.4
All employers, regardless of whether they use
E-Verify, are required to complete and retain a Form
I-9 for each new hire using information from a list of
acceptable documents presented by the employee,
such as a passport, permanent resident card, driver’s license or employment authorization document.
E-Verify then compares that information with Social
Security Administration and, if needed, DHS records
and notifies the employer whether the information
matches that of an eligible worker. Employers are required to inform workers whose information does not
match and give them eight federal workdays to resolve
the discrepancy before terminating employment.
E-Verify ensures that the information a worker
provides is accurate, not that he or she is, in fact, the
person identified by the documents. In response to

●● Fewer likely unauthorized immigrants working
in Georgia but no significant impact on the likely
unauthorized population in that state. Four years
after Georgia’s implementation in 2012, there was
no measurable change in its population of likely
unauthorized immigrants, but the number of likely
unauthorized immigrant workers was 14 percent
below projection.
●● In addition to these relative declines, the actual
numbers of likely unauthorized immigrants living and working in Arizona and Mississippi were,
as of 2015, below their implementation levels. In
Alabama and Utah, they were about the same or
slightly higher than when those states’ laws took
effect. In Georgia, the actual number of likely unauthorized immigrants working was the same as
when that state’s law took effect.
●● Greater impact on the number of likely unauthorized immigrant workers than on the overall likely
unauthorized population in the five states. Consistent with the intent of E-Verify laws to target unauthorized workers, the mandates typically have
larger effects on employment of likely unauthorized immigrants than on their population.
●● Statistically insignificant changes in likely unauthorized immigrant population and employment in
North Carolina and South Carolina. This suggests
that the E-Verify laws in these states have had no
measurable effect.
Because this research explores the seven states
with a universal E-Verify mandate consistently
and across a longer time horizon than any previous
study, it provides new insight into the requirement’s
effects on unauthorized immigrants living and
working in those jurisdictions. Where results were
statistically significant, the duration of the E-Verify
requirement’s impact was mixed. Policymakers can
use this information to consider the short- and medium-term effects on a state’s likely unauthorized
population and employment and to identify fiscal
and economic impacts associated with population
shifts that may warrant further study.

BACKGROUND ON E-VERIFY
e-verify is a free online system operated by
U.S. Citizenship and Immigration Services in the
Department of Homeland Security (DHS). It allows
businesses to electronically check employees’ eligi-

digital enforcement

federal reserve bank of dallas

concerns about the use of other people’s identities,
U.S. Citizenship and Immigration Services added a
photo-matching tool for federally issued employment
authorization documents, permanent resident cards
and U.S. passports. However, photo matching is not
available for driver’s licenses, the most commonly
presented form of photo identification.5

State-Level Policies

States began requiring some employers to use
E-Verify in 2006.6 Georgia passed a law requiring
public employers and government contractors to use
E-Verify; Colorado passed a requirement for government contractors; and North Carolina passed a requirement for state agencies and universities. The following year, Arizona became the first state to require
that all employers—not just public employers and
government contractors—use E-Verify, a mandate upheld by state and federal courts in 2008 and 2009 and
affirmed by the U.S. Supreme Court in 2011.7
Over the next few years, more states required some
or all public employers and government contractors
to use E-Verify, and seven other states—Alabama,
Georgia, Mississippi, North Carolina, South Carolina,
Tennessee and Utah—adopted mandates that apply
to all or nearly all employers.8 Excepting Mississippi,
those states initially required public employers and/

or government contractors to use E-Verify before expanding coverage to most or all other employers. (For
a full list of state E-Verify requirements, see Appendix A.) As of December 2016, 21 states had E-Verify
requirements of varying scope (see Map).
Experts and policymakers disagree about the value of E-Verify requirements, and states have made
different choices regarding the system. A few states
have discontinued their requirements or prohibited
mandatory use of E-Verify.

Compliance Mechanisms, Take-Up Rates

The Immigration Reform and Control Act of
1986 made it illegal to hire unauthorized workers
but reserved enforcement of that law to the federal government. In doing so, it limited the civil and
criminal sanctions that state and local governments
can impose on employers for hiring unauthorized
immigrants.9 Most laws that require universal use
of E-Verify punish noncompliant employers by suspending or revoking their business licenses, while
those covering only government contractors typically cancel existing contracts and prohibit future ones.
Some laws, such as Utah’s universal mandate, do not
enumerate consequences for noncompliance.10
The extent of compliance with E-Verify requirements is unknown. However, several studies of Arizo-

Map
MORE THAN ONE-THIRD OF STATES REQUIRED E-VERIFY IN 2016

No E-Verify requirement
All employers
Public sector/government contractors
Rescinded/expired all or in part

NOTE: Mandates are as of December 2016.
SOURCES: Data compiled from legal firms, other research, advocacy organizations and personal communications with government officials; “E-Verify
3.0, Self-Check, and I-9 Changes on the Horizon,” LawLogix by Hyland, May 11, 2010, www.lawlogix.com/e-verify; “Immigration,” Troutman Sanders,
www.troutmansanders.com/immigration.

digital enforcement

federal reserve bank of dallas

METHODOLOGY

na’s enforcement soon after implementation found
that only about half of new hires from October 2008
through September 2009 were screened and that as
of April 2010, just three employers had been indicted
for violating the law.11 Among the states with universal mandates, only South Carolina randomly audits
employers for compliance.12
Employers that are otherwise not required to use
E-Verify may still have an incentive to voluntarily
do so. For example, businesses operating in multiple
states, including at least one that requires E-Verify,
may choose to use the system for all of their establishments to ensure uniformity. Further, some states
allow firms to cite use of E-Verify as an affirmative
defense against charges of knowingly employing an
unauthorized worker.
The numbers of employers enrolled and cases
run through the system have grown considerably
(Chart 1). More than 600,000 employers were participating as of July 2015, and nationwide, half of all
new hires in fiscal year 2015 were screened using the
system.

Modeling E-Verify Requirements’ Impact

Chart 1
HALF OF NEWLY HIRED U.S. WORKERS
SCREENED WITH E-VERIFY BY 2015
Share of workers, percent
50

0
’02 ’03 ’04 ’05 ’06 ’07 ’08 ’09 ’10 ’11 ’12 ’13 ’14 ’15

NOTES: Figures are based on a comparison of the number of cases run
through E-Verify with data on new hires from Job Openings and Labor
Turnover Survey (JOLTS). Before 2007, data are for Basic Pilot, the
E-Verify predecessor.
SOURCES: U.S. Citizenship and Immigration Services, “What Is E-Verify,”
www.uscis.gov/e-verify/what-e-verify (last modified Feb. 26, 2016); U.S.
Citizenship and Immigration Services, “Estimated Costs and Timeline to
Implement Mandatory E-Verify,” https://goo.gl/XJOCyg (published June
10, 2016); U.S. Citizenship and Immigration Services, “History and Milestones,” www.uscis.gov/e-verify/about-program/history-and-milestones
(last updated March 11, 2016); and Bureau of Labor Statistics, “Job Openings and Labor Turnover (JOLTS),” www.bls.gov/jlt/data/htm.

This report estimates the effect of universal E-Verify mandates on the population and employment levels of likely unauthorized immigrants in seven states
with such requirements by comparing actual levels
with those projected absent the E-Verify policy. A
proxy group was used to identify likely unauthorized
immigrants, namely Mexican and Central American
immigrants ages 20–54 who are not naturalized U.S.
citizens and have at most a high school education.
The analysis used a synthetic control method to create projected counterfactuals to demonstrate what
would probably have occurred in each state had it
not implemented universal E-Verify and compared
those projections to what actually happened.13 The
counterfactuals were developed by identifying the
set of states with the most similar demographic and
economic characteristics to each of the seven studied
states before they enacted their E-Verify policies. The
states in each set were aggregated and weighted via an
algorithm explained briefly below and in greater detail
in Appendix B.
The synthetic control methodology was previously
used to examine the effect of the universal E-Verify
requirement in the state of Arizona, but this is the
first study to apply it to multiple states with similar
mandates. This analysis also examines a longer time
period than the previous studies of Arizona, which
are discussed in more detail later in this report.
Two important events must be considered when examining the effect of an E-Verify law—the date the law
was adopted and the date it took effect. The laws may
have different effects in the near term after they are
adopted relative to the period following implementation. For example, unauthorized immigrants and their
employers may not immediately react to a policy’s
adoption but instead might wait until it takes effect.
The method used here allows researchers to trace
the impact of E-Verify laws over time to see if their
impact grew or diminished in the years after taking effect. This analysis was conducted twice, first
using the laws’ adoption dates and then their implementation dates. Because the results indicated
few differences between the outcomes from the two
dates, the report only shows outcomes relative to
the implementation dates.14 If E-Verify requirements
affect population or employment levels before they

digital enforcement

federal reserve bank of dallas

take effect, the results here will understate their overall impact.
When examining the results, it is important to
take into account the confidence bands around the
findings, which are wider for states with smaller
unauthorized immigrant populations. Arizona, the
state with the largest such group in this analysis, was
home to approximately 350,000 working-age likely
unauthorized immigrants when its E-Verify requirement took effect. On the other end of the spectrum is
Mississippi, whose likely unauthorized immigrant
population numbered around 30,000. However, the
findings are consistent with predicted outcomes and
prior research.
This report focuses on two outcomes: changes in
working-age population and in employment of likely
unauthorized immigrants. Both would be expected to
fall after a state begins requiring widespread use of
E-Verify. With regard to population, inflows of unauthorized immigrants to an E-Verify state may decline
and outflows increase as migrants within and outside
the state experience or anticipate reduced employment opportunities. Employment of unauthorized
immigrants also should decline as the number of jobs
available to these workers falls.
The results highlight changes to each state’s likely unauthorized immigrant population and employment levels according to three measures:
●● The percent difference between the actual outcome and the projection in 2015
●● The projected percent change, absent the E-Verify
requirement, from the implementation year starting point
●● The actual percent change from the starting point
The magnitude of these changes is likely to depend
on a number of factors that vary across states, such as
size and composition of the unauthorized workforce,
employer compliance, whether a state’s neighbors
have E-Verify requirements, the size of states’ informal labor markets and the share of firms exempt from
the mandates.

Synthetic Control Method

Because the studied states enacted their laws in
different years, the available length of their post-implementation periods varies. Although universal
E-Verify mandates began taking effect as the U.S.
entered the Great Recession, the synthetic control
methodology used in this report implicitly controls

for business cycles and national policy changes common across states; this ensures to the extent possible
that the seven studied states and the states used to
create the counterfactuals reflect similar trends.15
The synthetic control method’s major advantage
is that the comparison group is selected via a data-driven process. A computer algorithm creates the
combination of states that best mirrors the treatment
state during the pre-intervention period instead of a
researcher choosing ad hoc which states should compose the comparison group.
This synthetic control method involves creating a
counterfactual of what might have occurred in a state
absent the policy change (the “control”) and then comparing that counterfactual to what actually occurred
in that state (the “treatment”). The counterfactual is
created by identifying the set of states that is the most
similar to the treatment state before the policy change
(the “intervention”) and then creating a weighted average of outcomes in those states. In essence, this method compares the actual outcome after a state implements an E-Verify law with the projected outcome had
the state not implemented the law. The difference between the actual and counterfactual gives an estimate
of the effect of the E-Verify law.
The first step in the synthetic control method is to
identify states that can be used to create the counterfactual, or the “donor pool.” 16 In this study, the donor
pool is the other 43 states and the District of Columbia, which had not enacted a universal E-Verify law.
The second step is to combine states in the donor pool that are most similar to the treatment state
during the pre-intervention period. For this analysis,
the combination was based on several characteristics
or predictor variables that are important when considering unauthorized immigration:
●● The outcome under examination (likely unauthorized immigrant population or employment)
●● The share of the population ages 20–54 composed
of likely unauthorized immigrants
●● Four measures of business-cycle conditions: real
gross domestic product per capita, the unemployment rate, single-family construction starts per
capita and single-family construction permits per
capita17
The last two variables proxy for the extent of construction activity in a state. The construction sector
is a major employer of unauthorized immigrants and
collapsed after experiencing rapid growth in many

digital enforcement

federal reserve bank of dallas

grants, 47 percent lack a high school diploma and another 27 percent are high school graduates but have
not attended college. Additionally, 70 percent of all
unauthorized immigrants were born in Mexico or
Central America.18 Although not all people who meet
these criteria are unauthorized immigrants, many
are, and economic researchers commonly use this
group as a proxy for unauthorized immigrants.19
This report refers to this population as “likely unauthorized immigrants.” Because this baseline does
not capture more educated unauthorized immigrants
or those born outside of Mexico and Central America
and may include some legal immigrants, the analysis
may understate the impacts of requiring E-Verify use.

states during the period examined. The analysis also
includes several demographic characteristics of likely unauthorized immigrants in the state as predictor
variables: the shares that are female, from Mexico
and have not completed high school.
The synthetic cohort method assigns each state in
the potential donor pool a relative weight. The relative weights minimize the mean-squared prediction
error (the squared deviations between the outcome
for the treatment state and the synthetic control unit
totaled over all pre-intervention periods). The relative weights total to 100 percent across the donor pool.
Appendix Tables B1 and B2 indicate the states that received positive weight in each specification. The preand post-intervention values for the synthetic control
were then created by applying the weights to the donor
states’ population and employment levels during each
period and calculating percent changes.
Statistical significance can be measured in several
ways when using the synthetic cohort method. This
report focuses on results that are statistically significant—meaning that a researcher has a reasonable
degree of confidence that they are not due to mere
chance—using difference-in-differences regressions,
as explained in Appendix B.

Previous Research

Several analyses have examined the impacts of Arizona’s universal E-Verify law and another anti-immigrant measure adopted there. The state has the largest
unauthorized population among those with universal
requirements and was the first to implement a universal mandate. Using the synthetic control method, the first study to examine the effects of Arizona’s
2008 E-Verify law found that it led to a substantial
decline in the number of likely unauthorized immigrants living in the state.20 Subsequent research confirmed that finding.21 However, two studies found that
Arizona’s 2010 omnibus immigration law, Support
Our Law Enforcement and Safe Neighborhoods Act
(S.B. 1070), which aimed to further reduce the number of unauthorized immigrants, appears to have had
little additional long-run effect.22 Arizona’s E-Verify
mandate also had a substantial negative impact on
employment among likely unauthorized immigrants
and prompted a large share to shift into self-employment from wage-and-salary employment.23
Previous research found evidence of a significant
drop in the number of likely unauthorized immigrants across the seven states with universal E-Verify
mandates in place by 2012.24 Rather than employing
the same synthetic control method used here and in
research specific to Arizona, that analysis conducted fixed-effects regressions, which measured how
much the number of likely unauthorized immigrants
changed within states after they required E-Verify,
controlling for the time trend in those states’ unauthorized immigrant populations. The regressions did not
compare changes in states that required E-Verify with
those that did not. The ability to do so is a key advantage

Data Source

This analysis uses data from the Census Bureau’s
Current Population Survey (CPS) to examine the
population and employment effects of E-Verify requirements. The large, nationally representative
survey is administered monthly and captures information about respondents’ places of residence, labor
market outcomes and demographic characteristics.
It captures all workers employed and does not distinguish between formal and informal sectors. The CPS
includes all U.S. residents regardless of legal status
and does not indicate whether someone is an unauthorized immigrant. This report therefore focuses on
a group of survey respondents who meet all of the following criteria:
●● Non-U.S. citizens
●● Born in Mexico or Central America
●● Ages 20–54
●● Possessing at most a high school education
When analyzing datasets such as the CPS, these
characteristics are typically used to define the unauthorized population. According to the Pew Research
Center, among working-age unauthorized immi-

digital enforcement

federal reserve bank of dallas

of the synthetic control method, although it requires
assuming that a treatment state would have looked like
its synthetic control absent the policy change.

LIKELY E-VERIFY IMPACT OBSERVED
IN SEVERAL STATES
in four of the seven studied states—Alabama,
Arizona, Mississippi and Utah—the number of likely
unauthorized immigrants is substantially lower than
the projected counterfactual, suggesting that universal E-Verify led to a much smaller unauthorized
immigrant population than if the policy had not been
enacted. Employment among the likely unauthorized
was also far lower than the projected counterfactuals in five of the seven states—the above-named four
states plus Georgia. The shortfall in employment was
larger than that of population in all cases, suggesting
as one might expect, that the laws more closely target
workers than the population at large.
The analysis also shows, however, that mandatory
E-Verify does not always result in an actual reduction in the likely unauthorized population over time.
Policies in Arizona and Mississippi resulted in lower
actual populations of likely unauthorized immigrants
in those states, but Utah’s likely unauthorized population hovered around its original size immediately
after that state’s requirement was implemented. And
in Alabama, the likely unauthorized population actually increased slightly over time.

Similarly, the number of likely unauthorized immigrants working in these five states was lower than the
projected counterfactual, but as with the population
levels, those decreases did not necessarily indicate
an actual decline in likely unauthorized workers over
time. The number of likely unauthorized immigrant
workers was lower in Arizona and Mississippi, slightly higher in Alabama and higher in Utah relative to
pre-E-Verify levels.
In all of these states, the likely unauthorized immigrant population and number of workers were lower
than they would have been without the E-Verify requirements. However, in Georgia, the likely unauthorized immigrant population was unaffected, and
in North Carolina and South Carolina, there were
no statistically significant effects. This section first
presents the results for the five states with a significant effect on population or employment and then the
results for the two states with no significant effects.

digital enforcement

federal reserve bank of dallas

ALABAMA
Effects on Likely Unauthorized Population
(Chart 2A)
• The actual number of likely unauthorized
immigrants in the state was 10 percent below
the projected level three years after the 2012
implementation.
• The model projects that without universal E-Verify,
the likely unauthorized population would have
grown 16 percent between 2012 and 2015.
• The actual number of likely unauthorized
immigrants dropped in the first 12 months after
implementation and remained below that level for
a year, before rebounding. By 2015, it was 4 percent
higher than in 2012.

Effects on Likely Unauthorized Workers
(Chart 2B)
• The actual number of likely unauthorized workers
was 57 percent below the projected level three years
after the 2012 implementation.
• The model projects that without universal E-Verify,
the number of likely unauthorized workers would
have grown 137 percent between 2012 and 2015.
• The actual number of likely unauthorized workers
initially declined but then recovered. It ultimately
increased 3 percent between 2012 and 2015.

Chart 2
ALABAMA: MANDATORY E-VERIFY CORRESPONDS WITH DECLINE,
SLOWER GROWTH IN NUMBER OF LIKELY UNAUTHORIZED IMMMIGRANTS
A. Population: Likely unauthorized immigrants
living in Alabama

B. Employment: Likely unauthorized immigrants
working in Alabama

Percent change since implementation

140
120

100
10%

Actual
0

57%

Actual

–20

Projected

–40

Projected

–20
–40

–60

–60
–80

–80
–8

–7

–6

–5

–4

–3

–2

–1 Apr. 1 1
2
2012
Years before and after E-Verify implementation

–8

–7

–6

–5

–4

–3

–2

–1 Apr. 1 1
2
2012
Years before and after E-Verify implementation

NOTES: Alabama’s E-Verify law took effect April 1, 2012. Each year represents pooled data from the preceding 12 months. Bracketed number denotes
the percent shortfall in the actual number in 2015 relative to the projection. Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey, 2003–15.

digital enforcement

federal reserve bank of dallas

ARIZONA
Effects on Likely Unauthorized Population
(Chart 3A)
• The actual number of likely unauthorized
immigrants was 28 percent below the projected
level eight years after the 2008 implementation.
• The model projects that without universal E-Verify,
the likely unauthorized population would have
grown 15 percent through the end of 2015.
• The actual number of likely unauthorized
immigrants declined for the first five years after
implementation and has since grown but remains
17 percent lower than in 2008.

Effects on Likely Unauthorized Workers
(Chart 3B)
• The actual number of likely unauthorized workers
was 33 percent below the projected level eight years
after the 2008 implementation.
• The model projects that without universal E-Verify,
the number of likely unauthorized workers would
have grown 13 percent between 2008 and 2015.
• The actual number of likely unauthorized workers
initially dropped and is recovering, remaining 24
percent lower than in 2008.

Chart 3
ARIZONA: MANDATORY E-VERIFY CORRESPONDS WITH DROP
IN NUMBER OF LIKELY UNAUTHORIZED IMMMIGRANTS
A. Population: Likely unauthorized immigrants
living in Arizona

B. Employment: Likely unauthorized immigrants
working in Arizona

Percent change since implementation

28%

–10

Actual

33%

–10
Projected

–20

Actual

–30

Projected

–20
–30

–40

–50

–50
–8

–6

–4

–2

Jan. 1
2
4
6
2008
Years before and after E-Verify implementation

–8

–6

–4

–2

Jan. 1
2
4
6
2008
Years before and after E-Verify implementation

NOTES: Arizona’s E-Verify law took effect Jan. 1, 2008. Each year represents pooled data from the preceding 12 months. Bracketed number denotes the
percent shortfall in the actual number in 2015 relative to the projection. Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey, 1999–2015.

digital enforcement

federal reserve bank of dallas

GEORGIA
Effects on Likely Unauthorized Population
(Chart 4A)
• The actual number of likely unauthorized
immigrants was not different from the projected
level four years after the 2012 implementation.
Georgia was the only state in which there was
a significant impact on one but not both of the
metrics examined.
• The model projects that without universal E-Verify,
the likely unauthorized population would have
grown about 6 percent through the end of 2015,
which is not significantly different from the actual
growth of about 8 percent.
• The actual number of likely unauthorized
immigrants declined the first year after
implementation but then rose.

Effects on Likely Unauthorized Workers
(Chart 4B)
• The actual number of likely unauthorized workers
was 14 percent below the projected level four years
after the 2012 implementation.
• The model projects that without universal E-Verify,
the number of likely unauthorized workers would
have grown 15 percent between 2012 and 2015.
• The actual number of likely unauthorized
workers dropped in the year following the law’s
implementation. It then rebounded over the next
three years, ending about 1 percent below its level at
the time of implementation.

Chart 4
GEORGIA: MANDATORY E-VERIFY HAS LITTLE EFFECT ON NUMBER
OF LIKELY UNAUTHORIZED IMMIGRANTS BUT EMPLOYMENT DECLINES
A. Population: Likely unauthorized immigrants
living in Georgia

B. Employment: Likely unauthorized immigrants
working in Georgia

Percent change since implementation

Actual

–10

0
Projected

–20

Actual

14%

–10

–30

–20

–40

–30

Projected

–40

–50
–8

–7 –6

–5

–4

–3 –2

–1 Jan. 1 1
2
2012
Years before and after E-Verify implementation

–8

–7 –6

–5

–4

–3 –2

–1 Jan. 1 1
2
2012
Years before and after E-Verify implementation

NOTES: Georgia’s E-Verify law took effect Jan.1, 2012. Each year represents pooled data from the preceding 12 months. Bracketed number denotes
the percent shortfall in the actual number in 2015 relative to the projection. In Chart A, no shortfall is denoted because there is no statistically significant
effect of E-Verify. Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey, 2003–15.

digital enforcement

federal reserve bank of dallas

MISSISSIPPI
ARIZONA
Effects on Likely Unauthorized Workers
(Chart 5B)
• The actual number of likely unauthorized immigrant
workers was 83 percent below the projected level
seven years after the 2008 implementation.
• The model projects that without universal E-Verify,
the number of likely unauthorized workers would
have grown 145 percent between 2008 and 2015.
• The actual number of likely unauthorized workers
fell 59 percent since 2008.

Effects on Likely Unauthorized Population
(Chart 5A)
• The actual number of likely unauthorized
immigrants was 70 percent below the projected
level seven years after the 2008 implementation.
• The model projects that without universal E-Verify,
the likely unauthorized population would have
grown 93 percent between 2008 and 2015.
• The actual number of likely unauthorized immigrants
grew slightly in the year after implementation but
then fell substantially beginning in the second year.
Despite a small recent rebound, the total remains 43
percent below 2008 levels.

Mississippi has far fewer unauthorized immigrants than the other states studied, so estimates may be less reliable.
However, the data are consistent with a sizable and enduring decline in the state’s unauthorized immigrant population.

Chart 5
MISSISSIPPI: MANDATORY E-VERIFY CORRESPONDS WITH DROP
IN NUMBER OF LIKELY UNAUTHORIZED IMMIGRANTS
A. Population: Likely unauthorized immigrants
living in Mississippi

B. Employment: Likely unauthorized immigrants
working in Mississippi

Percent change since implementation

180

320
280

140

240
200

100

160
60

120
70%

80
83%

40
Actual

–20

Projected

–40

–60

Projected

–80

–100

Actual

–120
–8

–6

–4

–2

July 1
2
4
2008
Years before and after E-Verify implementation

–8

–6

July 1
2
4
2008
Years before and after E-Verify implementation
–4

–2

NOTES: Mississippi’s E-Verify law took effect July 1, 2008. Each year represents pooled data from the preceding 12 months. Early spikes
in the projections are the result of a small likely unauthorized population in Vermont, which was included in the set of states that determined Mississippi’s counterfactual. Bracketed number denotes the percent shortfall in the actual number in 2015 relative to the projection.
Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey,1999–2015.

digital enforcement

federal reserve bank of dallas

UTAH
Effects on Likely Unauthorized Population
(Chart 6A)
• The actual number of likely unauthorized
immigrants was 30 percent below the projected
level five years after the 2010 implementation.
• The model projects that without universal E-Verify,
the likely unauthorized population would have
grown 43 percent between 2010 and 2015.
• The actual number of likely unauthorized
immigrants dipped in the year after
implementation but then rose steadily for three
years before dropping back to 2010 levels.

Effects on Likely Unauthorized Workers
(Chart 6B)
• The actual number of likely unauthorized immigrant
workers was 34 percent below the projected level
five years after the 2010 implementation.
• The model projects that without universal E-Verify,
the number of likely unauthorized workers would
have grown 74 percent between 2010 and 2015.
• The actual number of likely unauthorized workers
grew between 2011 and 2014. It then declined
sharply in 2015 but remained 15 percent higher
than in 2010.

Chart 6
UTAH: MANDATORY E-VERIFY CORRESPONDS WITH SLOWER GROWTH
IN NUMBER OF LIKELY UNAUTHORIZED IMMIGRANTS
A. Population: Likely unauthorized immigrants
living in Utah

B. Employment: Likely unauthorized immigrants
working in Utah

Percent change since implementation

Percent change since implementation
80

50
30%

Actual

0
–10

34%

10
Projected

Projected

0
–10

–20

–30
–8 –7

–6 –5 –4 –3 –2 –1 July 1 1 2 3
2010
Years before and after E-Verify implementation

–30
–8 –7

–6 –5 –4 –3 –2 –1 July 1 1 2 3
2010
Years before and after E-Verify implementation

NOTES: Utah’s E-Verify law took effect July 1, 2010. Each year represents pooled data from the preceding 12 months. Bracketed number denotes the
percent shortfall in the actual number in 2015 relative to the projection. Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey, 2001–15.

digital enforcement

federal reserve bank of dallas

NORTH
ARIZONA
CAROLINA & SOUTH CAROLINA
In North Carolina and South Carolina, changes
resulting from E-Verify were not statistically
significant, meaning that E-Verify requirements
appear to have had no measurable effect. Population
results are shown in Charts 7A and 8A, and
employment results are shown in Charts 7B and 8B
for North Carolina and South Carolina, respectively.
Actual population and employment of likely
unauthorized immigrants in North Carolina spiked
following implementation of E-Verify, while projected
levels initially rose more slowly. Three years after
implementation, however, the actual numbers were
slightly lower than projected levels, although the
difference is not statistically significant (Appendix B).
In South Carolina, actual and projected levels of
unauthorized immigrants living and working there
tracked one another closely, strongly suggesting the
law had no impact. Four years after the law, actual
numbers spiked above the projected, contrary to the
expected E-Verify effect.

Several factors may explain these findings. First,
employer compliance with E-Verify mandates
or unauthorized immigrants’ perception of their
vulnerability may have been lower in North Carolina
and South Carolina than in the other states with
universal requirements. Second, North Carolina and
South Carolina may have larger informal labor markets,
enabling unauthorized immigrants to continue to
work off the books while still being counted in CPS.
Further, the Carolinas were among the last states
to implement universal requirements, so the
effects may have been muted by the presence of
E-Verify policies in several nearby states or because
employers with operations in other states might
already have been using E-Verify—making it harder
to detect an effect. In addition, North Carolina’s law
was phased in based on employer size and exempts
small employers, which may reduce the impact. This
analysis is unable to distinguish between these or
other potential explanations.

Chart 7
NORTH CAROLINA: MANDATORY E-VERIFY HAS NO SIGNIFICANT IMPACT ON NUMBER OF
LIKELY UNAUTHORIZED IMMIGRANTS
A. Population: Likely unauthorized immigrants
living in North Carolina

B. Employment: Likely unauthorized immigrants
working in North Carolina

Percent change since implementation

50
40

Projected

30
20

Projected

–10

Actual

–10

Actual

–20

–20
–8

–7

–6

–5

–4

–3

–2

–1 Oct. 1 1
2
2012
Years before and after E-Verify implementation

–8

–7

–6

–5

–4

–3

–2

–1 Oct. 1 1
2
2012
Years before and after E-Verify implementation

NOTES: North Carolina’s E-Verify law took effect Oct. 1, 2012. Each year represents pooled data from the preceding 12 months. No shortfall is
denoted because there is no statistically significant effect of E-Verify. Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey, 2003–15.

digital enforcement

federal reserve bank of dallas

Chart 8
SOUTH CAROLINA: MANDATORY E-VERIFY HAS NO SIGNIFICANT IMPACT
ON NUMBER OF LIKELY UNAUTHORIZED IMMIGRANTS
A. Population: Likely unauthorized immigrants
living in South Carolina

B. Employment: Likely unauthorized immigrants
working in South Carolina

Percent change since implementation

80
60

Actual

20
0
0

Projected

–20

Projected

–20

–40

–60

–60
–8

–7 –6

–5

–4

–3 –2

–1 Jan. 1 1
2
2012
Years before and after E-Verify implementation

–8

–7 –6

–5

–4

–3 –2

–1 Jan. 1 1
2
2012
Years before and after E-Verify implementation

NOTES: South Carolina’s universal E-Verify law took effect Jan. 1, 2012. Each year represents pooled data from the preceding 12 months. No shortfall
is denoted because there is no statistically significant effect of E-Verify. Shaded area represents years after law took effect.
SOURCES: Census Bureau, Current Population Survey, 2003–15.

Chart 9
FIVE MANDATORY E-VERIFY STATES SAW LARGER
DECLINES IN LIKELY UNAUTHORIZED WORKERS THAN
AMONG OVERALL UNAUTHORIZED POPULATION

UNAUTHORIZED IMMIGRANT EMPLOYMENT MORE
AFFECTED THAN UNAUTHORIZED POPULATION
in the four states where effects were seen
on both the likely unauthorized population and
workers (as discussed previously), the levels were
consistently below the projections, suggesting that
requiring E-Verify drove down the number of likely unauthorized immigrants living and working in
these states (Chart 9).
However, in each state, the decline in workers was
greater than the drop in population. This suggests
that the E-Verify requirement has a bigger impact on
workers than on the population, which makes sense
given that the system and the mandates for its use
focus on worksites. The larger drop in the number of
likely unauthorized workers than in the likely unauthorized population probably means that a smaller
share of these immigrants is working.

Percent difference between actual and projected
by immigrant group, since implementations
10
1*

0
–10

–10

–14

–20
–30

–28

–30

–33

–40

–34

–50
–60

–57
Population

–70

Employment

–80
–90

–70
–83

Alabama

Arizona

Georgia

Mississippi

Utah

*Change in the unauthorized population in Georgia was not statistically
different from zero.
SOURCES: Census Bureau, Current Population Survey, 1999–2015.

digital enforcement

federal reserve bank of dallas

ECONOMIC EFFECTS AND POLICY IMPLICATIONS

tifies as ineligible, and delays in filling vacancies due
to mismatches in the system.
Current federal policy requires only the federal
government and its contractors use E-Verify; however, Congress has considered expanding the mandatory use of E-Verify or a similar verification system several times in the past few years, including
in the Border Security, Economic Opportunity and
Immigration Modernization Act of 2013 that the
Senate passed in June 2013 and the Accountability
Through Electronic Verification Act introduced in
the Senate in January 2017.26 A nationwide E-Verify
requirement would probably have a larger economic impact than state laws because it would reduce
opportunities for unauthorized immigrants or their
employers to avoid the mandate by relocating to another state. While state laws may displace some economic activity, lowering it in one area while raising
it in another, a national policy would not have such
offsetting potential.

the findings of this analysis show that the
numbers of unauthorized immigrants living and
working in a state tend to fall after adoption of a universal E-Verify law compared with what those counts
would have likely been without the requirement.
This suggests the laws can be effective in reducing
the population and employment of unauthorized immigrants. However, these laws may have broader economic and fiscal effects and policy implications that
warrant further study.
When the numbers of unauthorized immigrants or
workers in a state change, government revenue collections and spending are also likely to change. For example, most unauthorized workers have income and
payroll taxes withheld from their paychecks and file
tax returns using individual tax identification numbers or borrowed or false Social Security numbers.25
If unauthorized immigrants and their families work
less and earn less income, they will pay less in taxes
to the local, state and federal governments. Further,
demand for public assistance could also increase.
Although unauthorized immigrants themselves are
ineligible for most cash and noncash assistance programs, their U.S.-born children would qualify, assuming they meet income and other eligibility criteria.
Changes in unauthorized immigrant populations
or employment can also have broader economic impacts. For example, unauthorized immigrants are
a small share of the labor force in most of the states
examined in this study, but they have represented an outsized share of labor force growth in recent
decades. If universal E-Verify requirements affect
immigrant inflows or outflows, they could also affect
the supply of labor, and in turn, economic activity. In
addition, if unauthorized workers leave the state or
turn to self-employment, jobs they once held could
be available to low-skilled native and legal immigrant
workers. However, if legal workers hold jobs that
complement and rely on, rather than compete with,
unauthorized immigrants, then those legal workers
could be adversely affected by changes in unauthorized immigrant employment.
The cost of doing business also may be affected by
unauthorized population or employment changes.
For example, some companies might incur costs associated with longer searches for authorized workers, hiring and then replacing workers E-Verify iden-

CONCLUSION
this analysis shows that, compared with what
would probably have otherwise occurred, states with
universal E-Verify policies typically experienced
large reductions in the number of likely unauthorized immigrants and even greater declines in the
number of unauthorized workers. The impact on the
number of employed likely unauthorized immigrants
outweighed the effect on the likely unauthorized population in all five states with statistically significant
results, suggesting that though some unauthorized
immigrants may choose to avoid or leave a state with
a mandate, job opportunities for those who do reside there decrease. In addition, although the laws
corresponded with fewer unauthorized immigrants
and workers compared with projected estimates absent the E-Verify requirements, in some cases, the
mandates appear to have succeeded only in slowing
growth rather than producing a lasting reduction in
the actual number of likely unauthorized immigrants
living or working in a state.
Taken together, these findings suggest that these
laws’ primary impact is preventing or delaying
growth in the number of unauthorized immigrants
living and working in a state. The fiscal and economic
implications for the state as a whole are unclear and
warrant more research.

digital enforcement

federal reserve bank of dallas

APPENDIX A: STATE E-VERIFY REQUIREMENTS

E-Verify laws that apply only to public employees or
government contractors, in part, because they affect
only the small share of unauthorized immigrants
who work in the public sector.27 Additionally, public
employers and government contractors are likely to
require identification documents confirming citizenship or lawful immigration status even in the absence
of an E-Verify requirement.

as of december 2016, 21 states had some type
of E-Verify policy, ranging from universal in the
seven states examined in this analysis to those that
only require a subset of government contractors to
use E-Verify.
Previous research has found limited effects of

Table A1
STATES WITH E-VERIFY REQUIREMENTS
State

Law/Policy
(enacted date)

Effective Date and Employers Covered

Alabama

H.B. 56 (June 9, 2011)

Jan. 1, 2012; government contractors
April 1, 2012; all employers

Arizona

H.B. 2779 (July 2, 2007)

Jan. 1, 2008; all employers

Colorado

H.B. 06-1343 (June 6, 2006)

Aug. 7, 2006; state government contractors
Amended by S.B. 08-193 (May 13, 2008)

E.O. 11-02 (Jan. 4, 2011)

Jan. 4, 2011; executive branch agencies and (Jan. 4, 2011) government contractors;
superseded by E.O. 11-116

Florida

E.O. 11-116 (May 27, 2011)

May 27, 2011; executive branch agencies and government contractors

S.B. 529 (April 17, 2006)

July 1, 2007; public employers and government contractors, with size phase in

H.B. 87 (May 13, 2011)

Jan. 1, 2012; all employers with more than 10 employees, with size phase in

Idaho

E.O. 2009–10 (May 29, 2009)

July 1, 2009; state agencies and state government contractors

Indiana

S.B. 590 (May 10, 2011)

July 1, 2011; state and local governments and government contractors

H.B. 342 (June 30, 2011)

Jan. 1, 2012; government contractors

H.B. 646 (July 1, 2011)

Aug. 15, 2011; private employers must use E-Verify or retain a copy of certain documents

H.B. 5365 (June 26, 2012)

March 1, 2013; state agencies and contractors

E.O. 08-01 (Jan. 7, 2008)

Jan. 29, 2008; executive branch agencies and state government contractors;
expired April 4, 2011

S.F. 12 (July 20, 2011)

July 21, 2011; state government contractors

Mississippi

S.B 2988 (March 17, 2008)

July 1, 2008; all employers, with size phase in

Missouri

H.B. 1549 (July 7, 2008)

Jan. 1, 2009; public employers and government contractors

Nebraska

L.B. 403 (April 8, 2009)

Oct. 1, 2009; public employers and government contractors

S.B. 1523 (Aug. 23, 2006)

Jan. 1, 2007; state agencies and universities

Georgia

Louisiana
Michigan

Minnesota

North Carolina

H.B. 36 (June 23, 2011)

Oct. 1, 2011; county and city governments
Oct. 1, 2012; all employers with 25 or more employees, with size phase in
Nov. 1, 2007; state and local governments

Oklahoma

H.B. 1804 (May 8, 2007)

Pennsylvania

S.B. 637 (July 5, 2012)

Jan.1, 2013; public works contractors

H.B. 4400 (June 4, 2008)

Jan. 1, 2009; public employers

S.B. 20 (June 27, 2011)

Jan. 1, 2012; all employers

E.O. RP-80 (Dec. 3, 2014)

Dec. 3, 2014; executive branch agencies and contractors; superseded by S.B. 374

S.B. 374 (June 10, 2015)

Sept.1, 2015; state agencies and universities

July 1, 2008; state contractors (enjoined until Feb. 2, 2010)

South Carolina

Texas

digital enforcement

federal reserve bank of dallas

Table A1 (Continued)
State

Law/Policy
(enacted date)

Effective Date and Employers Covered

S.B. 81 (March 13, 2008)

July 1, 2009; public employers and government contractors

S.B. 251 (March 31, 2010)

July 1, 2010; all employers with 15 or more employees

H.B. 737 (April 11, 2010)

Dec. 1, 2012; state agencies

H.B. 1859 (March 25, 2011)

Dec. 1, 2013; government contractors

S.B. 659 (April 2, 2012)

June 10, 2012; new hires working in the Capitol Complex

Utah

Virginia
West Virginia

SOURCES: LawLogix (www.lawlogix.com/e-verify) and Troutman Sanders (www.troutmansanders.com/immigration).

digital enforcement

federal reserve bank of dallas

APPENDIX B: DATA AND METHODS

evolved in the same manner as in the counterfactual.
This study created the counterfactual based on how
similar states’ economic conditions and demographic
characteristics were before a policy change occurred.
If that similarity did not hold after the policy change
for reasons unrelated to the E-Verify policy, then
comparing the treatment state with the counterfactual should not be used to assess the impact of the
E-Verify requirement.
Second, the synthetic control method assumes that
the policy change in a given state did not spill over
into the states that compose the counterfactual. The
states used to create a counterfactual to those with
universal requirements were selected because of their
similar demographic characteristics and economic
conditions. They represented the best fit in the eight
years before E-Verify implementation. Few states that
border on E-Verify states contributed to the synthetic
controls, reducing concerns about possible spillover.
For southern states, most of their neighbors were not
in the donor pool because of the regional concentration of universal mandates.
If there were spillover effects, the synthetic control
method is likely to have overestimated the effect of an
E-Verify law since population and employment levels
would be moving in opposite directions in the treatment state and in the states that compose the counterfactual. Previous research suggests that universal
E-Verify requirements may divert some newly arriving
unauthorized immigrants to nearby states, but they do
not appear to cause unauthorized immigrants already
present in the United States to move to other states,
although some appear to leave the country.28 Previous
analyses of Arizona concluded that spillovers to other
states did not appear to bias results using the synthetic
control method.29 None of the remaining states that
have enacted universal E-Verify laws had sufficiently
large numbers of unauthorized immigrants to create
sizable spillovers to other states, further reducing this
concern.

Data Source

This analysis uses data from the Current Population
Survey (CPS) for 1999 to 2015. The CPS is a nationally
representative survey administered monthly by the
Census Bureau and the U.S. Bureau of Labor Statistics,
which includes information about respondents’ places
of residence, labor market outcomes and demographic
characteristics. About 60,000 households participate
each month.
Because the monthly CPS sample sizes are small
for some states, this study used data collapsed into
12-month periods for which the start date depends on
when a law took effect. For example, Alabama’s universal E-Verify mandate took effect in April 2012, so the
12-month periods run from April to March for that state.
The CPS includes all U.S. residents regardless of
legal status but does not indicate whether someone is
an unauthorized immigrant. Few government surveys
ask about legal status. This report therefore focuses
on a group that research has shown predominately
consists of unauthorized immigrants: immigrants
from Mexico and Central America ages 20–54 (primeworking-age adults) who have at most a high school
education and are not naturalized U.S. citizens. This
report refers to this population as “likely unauthorized
immigrants,” and the subset of these who are employed
are considered “likely unauthorized workers.”

Interpretation of the Data

The main analysis in this report examines population and employment levels. The figures given
reflect the percentage change in the number of likely
unauthorized immigrants living and working in each
studied state and the synthetic control relative to the
year before a state’s E-Verify law took effect. Because
population and employment levels vary considerably
across states, these variables were normalized to one
in the 12-month period before an E-Verify law took
effect. Population or employment levels in other years
were then scaled relative to that baseline, creating the
percentage differences shown.

Statistical Significance

The estimated differences between the treatment
and synthetic control states resulting from E-Verify
laws are visually apparent, but the law’s effects can
be measured more formally by testing the differences for statistical significance. To do so, this analysis
measured the “difference-in-differences,” the average gap between the treatment and the synthetic con-

Assumptions of the Synthetic Control Method

The synthetic control method requires making
several important assumptions:
First, the method assumes that, absent the policy
change, outcomes in the treatment state would have

digital enforcement

federal reserve bank of dallas

trol in the post-intervention period minus the average gap in the pre-intervention period.
The statistical significance of this difference-in-differences can be measured in two ways: by estimating
regressions using the data for the treatment state
and its synthetic control and generating traditional
estimates for the E-Verify requirement’s effect, or by
applying the synthetic control method to every state
that does not have a universal E-Verify mandate and
generating placebo estimates. The latter is akin to a
falsification test; it examines whether states that did
not implement an E-Verify law experienced population or employment changes, relative to their own
synthetic control, that coincided with the timing of
the E-Verify law. If several states experienced difference-in-differences as large or larger than the E-Verify
state did, it suggests that the analysis is not measuring
a causal effect of the E-Verify law. These two methods
differ somewhat in their approaches.
The first method estimates traditional difference-in-differences ordinary least squares regressions
with the data on population and employment in each
of the seven target states as compared with their
synthetic counterparts to determine whether the
relative difference between the two is statistically
significant. It is based on classical inference testing,
the traditional way economists test hypotheses, while
also incorporating the synthetic control.
These regressions combine observations for the
treatment state and its synthetic control and include
indicator variables for the treatment state, the post-intervention period and an interaction among those variables.30 The estimated coefficient on the interaction is
the difference-in-differences, that is, the average change
in the number of likely unauthorized immigrants
living or working in a state before and after E-Verify
implementation compared with the counterfactual.
That coefficient relative to its standard error gives a
measure of whether the difference-in-differences is
statistically significant. The difference-in-differences
parameters and their standard errors are reported in
Appendix Tables B1 and B2.
The second method assesses statistical significance
by using the same synthetic control method to generate
placebo estimates for each state in the donor pool and
then calculating how many of those estimates are at
least as large as the estimated difference-in-differences
for the treatment state.31 It compares the difference
between the E-Verify state and its synthetic control with

the analogous difference for every other state that has
not implemented a universal E-Verify requirement—
even those that are not part of the synthetic control. In
other words, the placebo estimates incorporate states
that are very different from the E-Verify state.
This method provides a ranking of states’ difference-in-differences, which indicates the number and
share of states with a difference-in-differences at least
as large as that of the E-Verify state. If the change in
enough other states is at least as large as in the treatment state, it suggests that the difference-in-differences
for the treatment state is not statistically significant. In
that case, the observed change in the treatment state
is unlikely to be due to its E-Verify law but instead
is probably the result of other factors shared across
several states, such as changes in economic conditions.
Appendix Tables B1 and B2 rank the treatment states
by the resulting difference-in-differences estimates.
In essence, the placebo difference-in-differences
form a sampling distribution for the treatment states’
difference-in-differences, allowing for inference testing
under the assumption that E-Verify laws are randomized across states. Statistical significance can be gauged
based on the p-value from a one-tailed test of a state’s
rank; these p-values are a state’s rank divided by the
total number of states included in the ranking.
This is classical inference testing only if the intervention—the E-Verify law—is randomized across
states, which may not be the case.32 Further, the
ranking says little about the relative magnitudes of
the difference-in-differences.
Using this placebo method, just a few of the estimates in this analysis are statistically significant at
conventional levels. Mississippi is the only state with
a significant relative change in its population of likely
unauthorized immigrants; none of the states that did
not implement a universal E-Verify policy had a larger
relative change in the number of likely unauthorized
immigrants around the time that Mississippi’s law
took effect. Alabama and Mississippi both experienced
significant relative declines in the number of likely
unauthorized workers.
Because this analysis relies on states that are dissimilar to the state with the universal requirement
as well as those that comprise the counterfactual, it
uses the traditional difference-in-differences to assess
statistical significance.
As discussed in the main text, previous studies of
Arizona using the synthetic control method found

digital enforcement

federal reserve bank of dallas

Table B1
ESTIMATED IMPACT OF UNIVERSAL E-VERIFY LAWS ON NUMBER OF LIKELY UNAUTHORIZED IMMIGRANTS
Average
Pre-Intervention
Difference

Average Post-Intervention Difference

Difference-inDifferences
(Standard Error)

Rank Among
Placebos

States Contributing to
Synthetic Control

Alabama

0.001

–0.248

–0.249
(0.089)

7/43

AR (4.5%), FL (17.7%), KY (39.8%),
MI (10,8%), MD (8.5%, VT (16.1%),
WV (2.5%)

Arizona

–0.000

–0.245

–0.245
(0.039)

15/45

ID (23.5%), NV (47.3), ND (7%),
TX (22.2%)

Georgia

–0.001

–0.057

–0.056
(0.060)

20/44

DE (11.1%), FL (17.2%), ID (2.3%),
KY (24.5%), ME (7.3%), NV
(34.2%), WV (3.4%)

Mississippi

–0.022

–1.119

–1.097
(0.156)

1/45

KY (52.6%), LA (8.8%), ND
(28.8%), TN (2.8%), VT (7%)

North Carolina

0.001

0.133

0.132
(0.140)

35/44

FL (28.6%), ID (13.1%), ME (2.3%),
MI (4.8%), NV (32.4%), WA (18.7%)

South Carolina

–0.043

0.082

0.124
(0.083)

30/44

FL (24.2%), ID (26.3%), KY
(15.3%), MI (33.5%), VT (0.8%)

0.011

–0.219

–0.230
(0.065)

8/43

ID (33.1%), NE (16%), NV (22.7%),
MI (26.6%), WV (1.6%)

State

Utah

NOTES: The difference-in-differences is the average post-intervention difference (the average difference between the treatment state
indicated and its synthetic control during the post-intervention period) minus the average pre-intervention difference. The standard error
for the difference-in-differences is calculated from an ordinary least squares regression using observations for the treatment state and its
synthetic control. The rank is the treatment state in a lowest-to-highest ordering of the difference-in-differences estimates for the treatment
state and the placebos.
SOURCE: Authors' calculations.

Table B2
ESTIMATED IMPACT OF UNIVERSAL E-VERIFY LAWS ON NUMBER
OF LIKELY UNAUTHORIZED IMMIGRANTS EMPLOYED
Average
Pre-Intervention
Difference

Average Post-Intervention Difference

Difference-inDifferences
(Standard Error)

Rank Among
Placebos

States Contributing to
Synthetic Control

Alabama

–0.003

–0.887

–0.884
(0.211)

2/44

FL (2.8%), ID (4.9%), KY (33.4%),
MI (25.1%), NV (11.7%), ND (15.7%),
VT (6.5%)

Arizona

–0.015

–0.331

–0.316
(0.055)

15/45

ID (35.3%), NV (25.3%), TX (28.7%),
VT (0.6%)

Georgia

0.000

–0.161

–0.161
(0.037)

21/42

DE (17.8%), FL (21.1%), KY (26.8%),
NV (30.6%), VA (3.7%)

–0.001

–1.465

–1.465
(0.374)

1/45

ME (2.6%), MT (17.4%), ND (38.8%),
TN (10.6%), VT (12.8%)

North Carolina

0.006

0.131

0.125
(0.170)

33/43

FL (28.4%), ID (9.8%), ME (1.1%),
MI (10%), NV (36.7%), WA (13.6%)

South Carolina

–0.038

0.112

0.150
(0.092)

33/42

FL 38.8%), ID (47.3%), KY (13.9%)

Utah

–0.000

–0.264

–0.264
(0.094)

12/44

ID (31.8%), NE (2.5%), NV (22.5%),
MI (24.3%), NM (12.1%), ND (0.8%),
OK (6%)

State

Mississippi

digital enforcement

federal reserve bank of dallas

a significant drop in the state’s population of likely
unauthorized immigrants, whereas the drop is not
significant here when evaluated using the placebo
method (although it is using the difference-in-differences regression method). There are several
methodological differences between this and earlier
research that can explain this seeming disparity in
significance levels. Most notably, this study examined
the number of likely unauthorized immigrants relative
to the year before an E-Verify law took effect, whereas
other research has focused on the population share

of likely unauthorized immigrants.33 Using either
measure, the unauthorized immigrant population in
Arizona fell after the state’s E-Verify law took effect,
but the drop is statistically significant only when using
population share, not when using the level.34 In short,
the statistical significance of the placebo results for
Arizona depends on the preferred measure of the
unauthorized population—the number or the proportion.35 Importantly, however, both measures suggest
a substantial decline in unauthorized immigration in
the state after implementation.

digital enforcement

federal reserve bank of dallas

NOTES
“About E-Verify,” U.S. Citizenship and Immigration
Services, www.uscis.gov/e-verify/about-program (last
modified Feb. 11, 2014).

“Did the 2007 Legal Arizona Workers Act Reduce the
State’s Unauthorized Immigrant Population?” by Sarah
Bohn, Magnus Lofstrom and Steven Raphael, Review of
Economics and Statistics, vol. 96, no. 2, 2014, pp.258–69.

For a discussion of E-Verify’s history, see “Electronic
Employment Eligibility Verification,” by Andorra Bruno,
March 2013, Congressional Research Service report
R40446, www.fas.org/sgp/crs/misc/R40446.pdf.

South Carolina S.B. 20 (June 27, 2011).

The canonical study using the synthetic control method
is “Synthetic Control Methods for Comparative Case
Studies: Estimating the Effect of California’s Tobacco
Control Program,” by Alberto Abadie, Alexis Diamond
and Jens Hainmueller, Journal of the American Statistical
Association, vol. 105, no. 490, 2010, pp. 493–505.

At least 19 cities and counties in Alabama, California,
Colorado, Florida, Nebraska, Pennsylvania, Utah and
Washington have enacted E-Verify laws that cover
some or all workers. These city- or county-level E-Verify
requirements typically apply either to all firms or only
to government contractors. Conversely, California state
lawmakers in 2011 invalidated at least six local laws that
required all employers or local government contractors
to use E-Verify. Because of the difficulty of accurately
ascertaining local E-Verify policies across the country, this
report focuses on state-level policies only.

Previous research on the timing of the effect of E-Verify
laws reaches mixed conclusions. “The Labor Market
Impact of Mandated Employment Verification Systems,” by
Catalina Amuedo-Dorantes and Cynthia Bansak, American
Economic Review: Papers & Proceedings, vol. 102, no.
3, 2012, pp. 543–48, concludes that the laws reduced
employment among likely unauthorized immigrants soon
after they were adopted, with no additional impact after they
went into effect. “The Impact of E-Verify Mandates on Labor
Market Outcomes,” by Pia M. Orrenius and Madeline Zavodny, Southern Economic Journal, vol. 81, no. 4, 2015, pp.
947–59, suggests that employers may adapt to laws during
the period between passage and implementation because
the effects on labor market outcomes are somewhat smaller
when examining implementation dates than when studying
adoption dates. Finally, “Do State Work Eligibility Verification
Laws Reduce Unauthorized Immigration?” by Pia M.
Orrenius and Madeline Zavodny, IZA Journal of Migration,
vol. 5, no. 1, 2016 (http://izajom.springeropen.com/
articles/10.1186/s40176-016-0053-3), found that universal
E-Verify laws reduced the population of likely unauthorized
immigrants more a year after implementation as opposed to
immediately after the laws took effect.

This discussion of local laws relies on information generously provided by Huyen Pham and Pham Hoang Van and
by LawLogix, www.lawlogix.com/e-verify-map/. For details
of how Pham and Van created their database of policies,
see “Measuring the Climate for Immigrants: A State-byState Analysis,” by Huyen Pham and Pham Hoang Van
in Strange Neighbors: The Role of States in Immigration
Policy, eds. Carissa Byrne Hessick and Gabriel J. Chin,
New York: New York University Press, 2014, pp. 21–39.
“What’s the Hire Date for E-Verify,” U.S. Citizenship and
Immigration Services, www.uscis.gov/e-verify/employers/
verification-process/whats-hire-date-e-verify (last modified
May 19, 2011).

“Driver’s License Verification,” U.S. Department of
Homeland Security, www.uscis.gov/e-verify/employers/
drivers-license-verification (last modified May 17, 2017).

At that time, E-Verify was still called Basic Pilot. Although
the name was changed to E-Verify in 2007, the tenets of
the program remained the same.

See note 13.

Tennessee is included as a donor state since it had not
yet enacted its universal mandate. In some specifications,
the number of likely unauthorized immigrants observed in
the data for certain states during each 12-month period
was too small for those states to be included in the donor
pool. Appendix Tables B1 and B2 report the number of
states in the donor pool in each specification.
16

Chamber of Commerce of the United States of America v.
Whiting, 563 U.S. 582 (2011).

Utah’s 2010 law (S.B. 251) allows employers to use either
E-Verify or the Social Security Number Verification Service,
a similar online verification process implemented by the
United States Social Security Administration. This analysis
considers the alternative to be equivalent to E-Verify
because, in each case, the employer is made aware of
an invalid Social Security number. Colorado’s law applies
to government contractors and was amended in 2008
under S.B. 08-193 (May 13, 2008) to allow an alternative
“Department Program,” in which the state’s Department of
Labor and Employment investigates complaints and may
conduct random audits. South Carolina’s 2008 law (H.B.
4400) required government contractors to use E-Verify
or employ only workers with a valid driver’s license or
identification card; for the purposes of this analysis, this
is not considered equivalent to mandating use of E-Verify.
Tennessee’s 2011 law (H.B. 1378) directs employers to
either use E-Verify or require all newly hired employees
to provide an identity and employment authorization
document from a specified list of documents, an either-or
approach. Louisiana’s 2011 law (H.B. 646) also has a
similar either-or approach.

Utah S.B. 251 (March 31, 2010).

17
The population and demographics data are calculated
from the Census Bureau’s Current Population Survey. The
real gross domestic product per capita data are from the
Bureau of Economic Analysis. The unemployment rate
data are from the Bureau of Labor Statistics. The construction starts and permits data are proprietary data from
Haver Analytics. ”Regional Economic Accounts,” Bureau
of Economic Analysis, https://bea.gov/regional/index.htm;
“Local Area Unemployment Statistics,” Bureau of Labor
Statistics, www.bls.gov/lau/.
18
“A Portrait of Unauthorized Immigrants in the United
States,” by Jeffrey S. Passel and D’Vera Cohn, Pew
Research Center, April 2009, www.pewhispanic.
org/2009/04/14/a-portrait-of-unauthorized-immigrants-inthe-united-states/.

For examples of papers that use this proxy, see note
14, “The Labor Market Impact of Mandated Employment
Verification Systems”; “Employment Verification Mandates
and the Labor Market Outcomes of Likely Unauthorized

8 U.S.C. §1324a(h)(2).

digital enforcement

federal reserve bank of dallas

“The Labor Market Effects of a Refugee Wave: Applying the
Synthetic Control Method to the Mariel Boatlift,” by Giovanni
Peri and Vasil Yasenov, National Bureau of Economic
Research Working Paper no. 21801, 2015.

and Native Workers,” by Catalina Amuedo-Dorantes and
Cynthia Bansak, Contemporary Economic Policy, vol. 32,
2014, pp. 671–80; “On the Effectiveness of S.B. 1070 in
Arizona,” by Catalina Amuedo-Dorantes and Fernando
Lozano, Economic Inquiry, vol. 53, 2015, pp. 335–51;
“The Response of the Hispanic Noncitizen Population to
Anti-Illegal Immigration Legislation: The Case of Arizona
S.B. 1070,” by Gonzalo E. Sánchez, unpublished paper,
Texas A&M University Department of Economics, 2015.

31
Other studies that use the synthetic control method to
examine the effect of E-Verify laws also generate placebo
estimates to gauge statistical significance. The universal
E-Verify mandate states, including the treatment state, are
not included in the donor pool to avoid biasing the placebo
estimates. See note 11, “Did the 2007 Legal Arizona
Workers Act Reduce the State’s Unauthorized Immigrant
Population?” for a discussion.

See note 11, “Did the 2007 Legal Arizona Workers Act
Reduce the State’s Unauthorized Immigrant Population?”

See note 19, “On the Effectiveness of S.B. 1070 in
Arizona.”
21

32
For further discussion, see “Synthetic Control Methods for
Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program,” by Alberto Abadie, Alexis
Diamond and Jens Hainmueller, Journal of the American
Statistical Association, vol. 105, no. 490, pp.493–505.

See note 19, “On the Effectiveness of S.B. 1070 in
Arizona,” and “The Response of the Hispanic Noncitizen
Population to Anti-Illegal Immigration Legislation: The Case
of Arizona S.B. 1070.”

This study also examines a longer time period after the law
went into effect, uses a slightly different donor pool and uses a
somewhat different set of variables to determine which states
compose the synthetic control. However, these differences do
not appear to drive the difference in significance levels.

“Employment Effects of State Legislation,” by Sarah
Bohn and Magnus Lofstrom, in Immigration, Poverty, and
Socioeconomic Inequality, eds. David Card and Steven
Raphael, New York, Russell Sage Foundation, 2013, pp.
282–314.

If we examine the population share of likely unauthorized
immigrants instead of the relative level using the same
time period, donor pool and predictor variables as in this
analysis, the difference-in-differences for Arizona ranks
third out of 45 states, which is statistically significant below
the 10 percent level. Only two states—California and
Nevada—experienced a larger drop in their population
shares of likely unauthorized immigrants compared with
their counterfactuals during the period examined here.

See note 14, “Do State Work Eligibility Verification Laws
Reduce Unauthorized Immigration?”
24

25
See, for example, Congressional Budget Office, “The
Impact of Unauthorized Immigrants on the Budgets of
State and Local Governments,” December 2007, www.
cbo.gov/sites/default/files/110th-congress-2007-2008/
reports/12-6-immigration.pdf. The size of the Social
Security Administration’s Earnings Suspense File—over $1.2
trillion—also suggests that a large number of unauthorized
immigrants have taxes withheld from their paychecks.

See S. 744 of the 113th Congress (2013–14), www.
congress.gov/bill/113th-congress/senate-bill/744/text; S.
179, 115th Congress (2017), www.congress.gov/115/bills/
s179/BILLS-115s179is.pdf.

See, for example, note 14, “The Impact of E-Verify
Mandates on Labor Market Outcomes,” and “Do State Work
Eligibility Verification Laws Reduce Unauthorized Immigration?” Only 1 percent of likely unauthorized immigrants who
work report being in the public sector, compared with 14
percent of all workers (authors’ calculations from 2005–06
Current Population Survey from Merged Outgoing Rotation
Group data).

“Piecemeal Immigration Enforcement and the New
Destinations of Interstate Undocumented Migrants:
Evidence from Arizona,” by Catalina Amuedo-Dorantes and
Fernando Lozano, unpublished paper, Pomona College
and San Diego State University (2014); see note 14, “Do
State Work Eligibility Verification Laws Reduce Unauthorized
Immigration?”

29
See note 11, “Did the 2007 Legal Arizona Workers Act
Reduce the State’s Unauthorized Immigrant Population?”

The regressions also include year fixed effects and are
estimated with robust standard errors. Several recent
studies examining the effect of the influx of Cuban migrants
as a result of the Mariel boatlift also take this traditional
regression difference-in-differences approach in addition to
doing placebo difference-in-differences testing. See “The
Wage Impact of the Marielitos: A Reappraisal,” by George
J. Borjas, National Bureau of Economic Research Working
Paper no. 21588, 2015; “The Wage Impact of the Marielitos:
Additional Evidence,” by George J. Borjas, National Bureau
of Economic Research Working Paper no. 21850, 2016;

35
The statistical significance of the population results for
the other states are not sensitive to whether levels or shares
are examined, with one exception: Unlike the results for
its population level, Mississippi’s population share of likely
unauthorized immigrants rose relative to its synthetic control,
but the increase is not statistically significant using the
placebo method (ranking 26th out of 45 states).

digital enforcement

federal reserve bank of dallas

THE STAFF
AUTHORS
PIA M. ORRENIUS
MADELINE ZAVODNY
EDITOR
MICHAEL WEISS
ASSOCIATE EDITOR
DIANNE TUNNELL
DESIGNERS
DAVIAN HOPKINS
ELLAH PIÑA
"Digital Enforcement: Effects of
E-Verify on Unauthorized Immigrant
Employment and Population" is
a special report of the Research
Department at the Federal Reserve
Bank of Dallas, P.O. Box 655906,
Dallas, TX 75265-5906. It is available
on the web at www.dallasfed.org/
research/pubs.aspx#tab4. Portions
may be reprinted on the condition
that the source is credited and a
copy is provided to the Research
Department.
Federal Reserve Bank of Dallas
2200 N. Pearl St.
Dallas, TX 75201

digital enforcement

federal reserve bank of dallas

dallasfed.org
@DallasFed

Full text of Digital Enforcement : Effects of E-Verify on Unauthorized Immigrant Employment and Population : A Special Report of the Federal Reserve Bank of Dallas

FRASER