Rand Rra1295 1 PPP Criminal Record Eligibility
Summary
A RAND Corporation research report, Small Businesses, Criminal Histories, and the Paycheck Protection Program (RR-A1295-1, © 2021), by Shawn D. Bushway, Dulani Woods, Denis Agniel and David M. Adamson. The report estimates how many small business owners have criminal history records and how many businesses and employees were affected by PPP felony restrictions and their 2021 revisions. Its key findings state that nearly 4 percent of small businesses had owners with a criminal history, and that under the revised 2021 restrictions the number of businesses affected dropped by 95 percent, to 11,481. The national analysis links business ownership and criminal history records from a consumer and background check data company, and a state-level analysis samples owners in Minnesota and North Carolina. The report was produced with support from Arnold Ventures.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Full text
Research Report
C O R P O R AT I O N
SHAWN D. BUSHWAY, DULANI WOODS, DENIS AGNIEL, DAVID M. ADAMSON
Small Businesses,
Criminal Histories, and
the Paycheck Protection
Program
T
here is a growing realization of the potential importance of business ownership for people
with criminal history records (Hwang and Phillips, 2020). These individuals face many bar-
riers to employment, and entrepreneurship can be a reasonable alternative to taking on wage
work. However, most U.S. government domestic aid programs exclude individuals with
criminal history records from receiving federal assistance—and, because the prevalence of criminal
histories among small business owners is largely unknown, we do not know how many entrepre-
neurs are cut off from such assistance.
A recent example of a domestic aid program with restrictions for people with criminal history
records is the Paycheck Protection Program (PPP), which President Donald Trump signed into law
as part of the Coronavirus Aid, Relief, and Economic Security (CARES) Act on March 26, 2020
(Public Law 116-136, 2020). The PPP provided money for payroll, rent, mortgage interest, and utili-
ties to small businesses (businesses with fewer than 500 employees). In its first two weeks, the PPP
distributed over $349 billion to approximately 1.7 million small businesses. However, small busi-
nesses owned by individuals with a criminal background were ineligible to receive funds.
KEY FINDINGS
Q Nearly 4 percent of all small businesses had owners with a criminal history, and about
1.5 percent of all small businesses had owners with a felony record.
Q An estimated 140,325 disqualifying felonies existed under the original PPP restrictions, and
212,655 small businesses had owners with a record of a felony in the last five years.
Q Under the revised 2021 PPP restrictions, the number of businesses affected dropped by
95 percent, to 11,481.
Q Under the original PPP restrictions, 343,198 employees were affected; this number was
reduced by 95 percent, to 17,533, under the revised restrictions.
Initially, the PPP used a sweeping definition of potentially prevented from accessing PPP aid because
criminal background for its restrictions. The defini- of the original PPP felony restrictions. We also esti-
tion applied to any owner of 20 percent or more of mate how many small businesses and small busi-
the equity of the applying business who ness employees were potentially given access to PPP
• was incarcerated aid after the 2021 felony restriction revisions by the
• was on probation Biden administration.
• was on parole We produce both a national estimate and an
• was subject to formal criminal charges in any estimate focused on two states. At the national level,
jurisdiction we applied an innovative method that used data from
• had been convicted of any felony, been placed a consumer and background check data company.
on pretrial diversion, or been placed on any We worked with a data aggregation company that
form of parole or probation (including pro- independently collects information on business own-
bation before judgment) within the past five ership and information on criminal history records.
years (U.S. Small Business Administration Although these data sets are separate, each is indexed
[SBA], 2020). using the same process to uniquely identify associ-
ated individuals. As a result, at the national level, we
Questions about criminal history appeared on
were able to determine how many people who own
the application form, and applicants were required to
small businesses also have a criminal history record.
give the SBA permission to check their records.1
We were also able to examine the effects of felony
Following a lawsuit by the American Civil
restrictions by age, sex, race, state location, busi-
Liberties Union (Hayashi, 2020), the Trump adminis-
ness size, and industry category. However, because
tration limited the five-year felony restrictions to those
of missing data in these demographic categories, we
convicted of fraud, bribery, embezzlement, or a false
have low confidence in these results.
statement in a loan application or an application for
At the state level, we worked with a different
federal financial assistance. For all other felonies, the
data aggregation company that maintains a database
restrictions were limited to a one-year window after
on small businesses. We asked for a sample of small
conviction for those not incarcerated (SBA, 2020).
business owners in two states, Minnesota and North
In February 2021, the Biden administration
Carolina. We then searched for these individuals in
eliminated the one-year restriction for those with
publicly available state criminal history records data
felony convictions who were not incarcerated (The
to estimate the number of small business owners who
White House, 2021). The restriction for those incar-
have a criminal history as well as those who might
cerated was maintained because of concerns about
have been excluded from PPP aid under the five-year
creditworthiness (SBA, 2020).
felony rules.
We compare our results from both the national-
and state-level approaches with results from a recent
Study Purpose
study of PPP criminal history impacts on sole propri-
In this report, we estimate the number of small etorships (Finlay, Mueller-Smith, and Street, 2020).
business owners who have a criminal history and In this study, researchers drew from the Criminal
the number of small business employees who were Justice Administrative Records System (CJARS) to
estimate the number of sole proprietors who have
Abbreviations criminal history records that would exclude them
from the PPP program; these estimates were made for
CI confidence interval
seven states. In Michigan and Texas, the two states
CJARS Criminal Justice Administrative
for which CJARS has the most-complete criminal
Records System
PPP Paycheck Protection Program history information, researchers estimated that as
SBA U.S. Small Business Administration many as 2.6 to 3.2 percent of sole proprietors were
ineligible for PPP loans based on one of the five
2
criminal history record restrictions described above. ground check on an applicant. Some of these com-
The CJARS study likely played a key role in the revi- panies also are useful resources for an analysis of the
sions of the PPP restrictions because it demonstrated potential reach of PPP rules because they warehouse
the implications and impact of the criminal justice records across a broad variety of topical areas, using
restrictions (Arnold Ventures, 2021). a common identification across these areas. This
We examine business ownership beyond sole common identification allowed us to cross-match
proprietorship and in more areas than seven states, all individuals who were associated with business
although our research focuses more narrowly on records and all individuals who were associated with
felony restrictions. As such, this report is an oppor- criminal history records. The process of creating
tunity to both verify and extend the results of the common identification for different areas is known
CJARS study. as entity resolution. An entity resolution algorithm
helps link possibly ambiguous characteristics (such
as criminal history records) to an object or individual
Results of National Analysis (Talley, 2011). As an example of linking business
owners to criminal history records, if three individu-
Data and Methods
als named William Smith were born in October 1980
We used data from a consumer and background check and three individuals named Bill Smith were born in
data company to link information from individual October 1980 and owned construction companies,
criminal history records to business information the entity resolution algorithm could attempt to cor-
about company ownership. This allowed us to create a relate each one of those records to one William Smith
nationwide estimate of the number of business owners who was born on October 15, 1980, and who was
who have been convicted of a crime in the past. It also convicted of a felony in a county court.
allowed us to estimate how many might be affected This company’s entity resolution algorithm
by PPP felony restrictions. We do not have informa- matched multiple characteristics to an individual,
tion on other aspects of individuals’ criminal history ensuring that all records—such as business records
records, such as current correctional status or whether and criminal history records—pertaining to an indi-
an individual faces pending charges. vidual were available through the database. After
The consumer and background data warehousing a limited evaluation of the data company’s entity
company that provided our data both aggregates and resolution algorithms, we believe that its methodol-
organizes information from government and com- ogy is more accurate than simpler record-matching
mercial sources across the United States. It sells data approaches that use only a few of the characteristics
services to government and corporate decisionmak- that might be found in each record (e.g., name, date
ers to allow them to make informed decisions about of birth). Although any semiautomated record-
providing credit, employment, rental housing, insur- matching system will have some degree of error, we
ance, and more. Commercial companies that main- do not believe that this degree of error had a signifi-
tain criminal history records play a significant role in cant impact on our results.2
employment decisions in the United States (Bushway To the best of our knowledge, no previous
and Kalra, 2021; Lageson, Webster, and Sandoval, research has made use of nationwide consumer and
2021). Although there are no estimates of the number criminal history record data to examine the impli-
of criminal history record checks done for the explicit cations of a particular restriction on employment
purpose of lending, researchers have estimated that or business activity (Bushway and Kalra, 2021). For
between 75 and 90 percent of all formal job applicants this study, we asked the consumer and background
will face a criminal history record check at some stage check company to search its criminal history and
in the hiring process (Bushway and Kalra, 2021). business records. We were particularly interested in
We selected a consumer and background data the subpopulation of business owners who also had a
company for our research because a bank disbursing criminal history record. We also asked them to iden-
PPP aid might use such a company for a quick back- tify both owners who had a felony or misdemeanor
3
conviction and the time elapsed since the conviction title was “business owner” in one of the 50 U.S. states
(with a preference for the felony if both existed). The or Washington, D.C.; these individuals were affiliated
total number of owners with felony convictions were with 3.4 million small businesses. We used the SBA
also binned by the following categories: business size, definition of small (500 or fewer employees) to iden-
industry, age, race, and sex. We did not receive any tify small businesses from business records. This des-
individually identifying information in this process. ignation potentially included some large businesses
Some individual owners were reported to be whose size is unknown; however, most businesses
affiliated with multiple entities, and our analysis that we identified were noted as having fewer than
takes this into account. (We also were provided with 500 employees in the consumer and data background
counts of individuals affiliated with multiple entities.) company records.3
However, because of the limitations of this aggrega- As is typical for consumer and criminal history
tion procedure, individual owners would count twice data collected from hundreds of public and com-
if they were affiliated with businesses in different mercial sources from across the country, significant
industries, which would place them into different portions of many variables—such as business size,
aggregation bins in the data. Similarly, a single busi- offense type, and race—were incomplete. We used
ness could be counted twice if it has multiple owners statistical methods to impute the missing values for
who have felonies. Depending on how often this the quantity of most interest to this report—whether
occurred in the data, both of these limitations could a record with an unknown felony status could real-
result in an artificially higher estimate of the propor- istically be inferred to be a recent felony. Because of
tion of small business owners with a criminal history. the scale of missing values for other characteristics
We have no direct information on how often this (such as race, sex, and age), we did not impute them.
occurred, but the broad agreement of our prevalence We acknowledge that the statistical model assumes
estimates with Finlay, Mueller-Smith, and Street’s that the information that is available to us—such as
(2020) estimates and with our state-level analyses category of offense (e.g., arson, assault), age of indi-
(which use a different dataset) suggests that the extent vidual, and state—are sufficient for inferring felony
of this double counting may have been small. status. Our model is at best an approximation and is
Using this information, we produce national likely to be subject to mis-specification.
estimates of the number of small business owners who We used an approach known as a doubly robust
have records of any felony conviction and of those estimator to estimate how many offenses of unknown
who have a felony conviction within the past five felony status could reasonably be inferred to be
years (we define these as recent felonies). We estimate recent felonies (hereafter, inferred felonies) (Bang and
these numbers at the national and state levels, within Robins, 2005). In this approach, we used two pieces
industry, for various business sizes, and by charac- of information to infer missing recent felonies.
teristics of the owner. Furthermore, we estimate how First, we used the probability of felony status
many small business owners would have been affected being unknown given everything we already know
by changes in PPP eligibility implemented in the past about the individual and business. For example, if
year to relax the restrictions against those with crimi- most of the records in North Carolina have unknown
nal histories. Specifically, we limit our analysis to felony status, a greater burden would be placed on the
felonies that were known to be related to fraud, coun- observed felonies to inform us about missing ones in
terfeiting, forgery, bad checks, or bribery. that state. This type of model is known as a propen-
sity score model.
Second, we used the probability of a record being
Numbers of Criminal History Records
a recent felony given everything we know about the
and Business Records individual and business. If most of the records that
Criminal offense information was matched to are observed to have assault as the offense category
2.2 million records affiliated with individuals whose are recent felonies, then we might be inclined to
think that a record listed as an assault with unknown
4
felony status could very well be a recent felony. This assumption likely leads to an undercount of employ-
type of model is referred to as an outcome model. ees that could be helped by the policy changes.
We used location (state) and category of offense, Finally, when an owner is known to be affiliated with
sex, race, age of the individual, industry, and size of multiple businesses in the same industry, we only
business in the propensity score and outcome models, count one such business for the purpose of estimat-
each of which was estimated with logistic regression. ing the number of affected employees. Therefore, our
The eventual estimator combines the propensity estimate of affected employees could reasonably be
score and outcome model with the observed data so considered a lower bound.
that, if either model is correct, the final estimate will
be consistent.4
Estimated National Prevalence of
In addition to reporting estimates of the number
of businesses affected by PPP restrictions nationwide, Small Business Owners with a Felony
we also report estimates by owner and business char- Conviction
acteristics. These include state, age, race, business size, We first estimate the number of small business
and industry. However, many of these characteristics owners who had at least one felony conviction at
remain unknown. Because we do not know the extent some point in time. Table 1 presents our estimates
of missing data across races, industries, and ages, we of the numbers of small business owners with any
consider our subgroup estimates possibly to be lower criminal history. We estimate that nearly 4 percent
than reality. For example, our estimate of female busi- of all small businesses have owners with a criminal
ness owners affected by restrictions might be much history and about 1.5 percent of all small businesses
higher if we could identify the sex of every owner. have owners with a felony record. These estimates are
Furthermore, our estimates of the numbers of likely low because they are based on records that are
employees who possibly were affected by PPP aid known to have gaps and that cover a limited period
restrictions are conservative because we do not know (in many states, records more than 20 years old were
the exact number of employees in each business. not warehoused).
Business sizes are known only in broad categories In comparison, Shannon et al. (2017) estimated
(fewer than five employees, five to nine, ten to 19, that 8 percent of all U.S. adults have at least one
20 to 99, and 100 to 499), and many business sizes felony conviction. There are many reasons why busi-
are unknown. Our conservative estimates assign the ness owners might have a lower prevalence of felony
lowest number of employees within each band to a convictions than the general population. One of
business. For example, if the business is categorized those reasons could be the barriers to loans and other
as having five to nine employees, we assume that assistance that make it harder for those who have a
they have five employees. In line with that approach, felony record to become business owners. However, a
businesses of unknown size are assumed to have one nontrivial number of individuals with felony records
employee. This is because we do not have sufficient still have become business owners.
information in the rest of the data to confidently Among the 2.2 million individuals listed as small
estimate what the business size is likely to be. Our business owners in the data, 65,045 were identified as
TABLE 1
Overall Estimates of Criminal History Among Small Business Owners
Business Owners Associated Businesses Prevalence
(95-percent CI) (95-percent CI) (95-percent CI)
Any criminal history 1,136,309 1,730,790 3.83%
(1,090,427–1,182,190) (1,676,849–1,784,730) (3.71%–3.95%)
Any felony record 433,013 661,113 1.46%
(419,612–446,415) (644,981–677,244) (1.43%–1.50%)
NOTE: CI = confidence interval.
5
having recent felonies. These owners were affiliated We estimate that under the revised 2021 PPP
with 97,295 small businesses; therefore, small busi- restrictions, the number of businesses affected
nesses with owners having listed recent felonies were dropped by 95 percent—from 212,655 (0.47 percent
0.22 percent of the 45.2 million total businesses in the of small businesses) to 11,481 (0.03 percent of small
data.5 businesses). This 95-percent drop corresponds to the
In addition, 1.4 million records had unknown drop in disqualifying recent felonies, which fell to
crime status.6 We estimated that about 75,000 recent only 6,956 under the revised restrictions.
felonies were affiliated with about 115,000 businesses We also estimate the number of employees
among these records with unknown felony status. affected by both sets of PPP restrictions. We estimate
We then estimate the number of small business that 343,198 employees (95-percent CI 336,767–
owners who faced the original restrictions embed- 349,629) were affected by the original PPP restric-
ded within the PPP loan program (see Figure 1). The tions; this number was reduced by 95 percent to 17,533
original restrictions barred anyone with a felony (95-percent CI 16,399, 18,666) under the revised PPP.
conviction in the last five years (recent felonies).
Our final estimated total is that there were Effects by Business Size
140,325 disqualifying felonies (95-percent We also examine how business and owner character-
CI 136,148–144,502) and that 212,655 small busi- istics were associated with being affected by both sets
nesses had owners with a record of a recent felony of PPP restrictions.
(95-percent CI 207,450–217,861). This is an estimated Most small businesses affected by PPP restric-
overall prevalence of 0.47 percent of all businesses tions were relatively small (even for small businesses)
(95-percent CI 0.46–0.48 percent). or of unknown size (see Figure 2). For example, the
FIGURE 1
number of very small businesses (with fewer than five
employees) affected under the original PPP is more
Estimate of Disqualifying Felonies and
Affected Businesses Under the Original than 40 times larger than that of small businesses
and Revised Paycheck Protection Plan with over 100 employees.
Restrictions However, when the number of employees is taken
into consideration, affected larger small businesses
have roughly the same number of affected employees
250,000
as ones of a smaller size (see Figure 3). Companies
Original restrictions
with 20 to 99 employees actually have the most
Revised restrictions affected employees under the revised restrictions
200,000
CIs (3,979, 95-percent CI 3,385–4,573). The restriction
revisions caused a drop of up to 97 percent in the
150,000 number of affected businesses (regardless of size).
Effects by State
100,000 We also examine the recent felony prevalence by
state. We find that as many as 1.8 percent and as few
as 0 percent of businesses in a state might have been
50,000
affected by the original PPP restrictions. Under the
revised PPP restrictions, most states experienced a
reduction of between 90 and 97 percent in affected
0
Disqualifying felonies Affected businesses businesses, which is consistent with our findings at the
national level. Under the revised restrictions, the range
NOTES: Under the original felony restrictions, 0.47 percent of our
nationwide estimate of 45,159,461 small businesses were affected. of affected businesses dropped to between 0.14 and
Under the revised restrictions, 0.03 percent were affected—a 0 percent per state. However, given the incompleteness
reduction of 95 percent.
6
FIGURE 2
Estimates of Businesses Affected Under the Original and Revised Paycheck Protection
Plan Restrictions, by Business Size
>5
5 to 9
Business size
Original restrictions
10 to 19
Revised restrictions
CIs
20 to 99
100 to 499
Unknown
0 20,000 40,000 60,000 80,000 100,000 120,000
Number of businesses
NOTE: Employee estimates were conservative: When an owner was known to be affiliated with multiple businesses, only one business was used to
contribute to the employee estimate. In addition, employee estimates were on the lower end of each size band; for example, all companies in the
five-to-nine employee size band were assumed to have only five employees.
FIGURE 3
Estimates of Employees Affected Under the Original and Revised Paycheck Protection
Plan Restrictions, by Business Size
>5
95-percent reduction in affected employees
5 to 9
94-percent reduction in affected employees
Original restrictions
10 to 19 Revised restrictions
95-percent reduction in affected employees
Business size
CIs
20 to 99
94-percent reduction in affected employees
100 to 499
97-percent reduction in affected employees
Unknown
95-percent reduction in affected employees
Total
95-percent reduction in affected employees
0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 90,000 300,000 350,000
Number of employees
NOTE: Employee estimates were conservative: When an owner was known to be affiliated with multiple businesses, only one business was used to
contribute to the employee estimate. In addition, employee estimates were on the lower end of each size band; for example, all companies in the
five-to-nine employee size band were assumed to have only five employees.
7
of criminal history reporting at the state level nation- nal justice policies affect the levels of criminal history
wide (Lageson, Webster, and Sandoval, 2021), we penetration separately from offending.
believe our lower bound of 0 percent is more likely to
be the result of the lack of data than an actual finding. Effects by Race
When we modify our nationwide prevalence estimate Among the affected businesses, we estimate that
to the 29 states in which we believe there is good state-
• 93,640 had owners who were White
wide court coverage,7 we estimate the overall preva-
• 31,620 had owners who were Black
lence as 0.65 percent (95-percent CI 0.63–0.67 per-
• 4,212 had owners who were Hispanic
cent) rather than 0.47 percent. As a result, we view our
• 2,753 had owners who were Asian or Pacific
nationwide estimates as a lower bound of the number
Islander
of businesses that have been affected by the changes
• 437 had owners who were American Indian or
in initial PPP restrictions.
Alaskan Native
• 79,279 had owners of unknown race and
Effects by Industry Category
ethnicity.
The impact of the PPP restriction also varies dra-
Under the revised restrictions, those numbers
matically across industry categories. We estimate
were reduced by more than 86 percent for all racial
that about 27,000 retail businesses (95-percent
and ethnic categories (Table 2). Although the changes
CI 26,651–27,721) and about 25,000 construction
in affected businesses were not dramatically differ-
businesses (95-percent CI 24,376–25,3900) had
ent across race and ethnic categories, we recognize
owners with recent felonies, and each of these was
that there is significant potential for variation across
reduced by 94 to 95 percent by the change in PPP
racial and ethnic categories within specific states.
restrictions (Figure 4). The industry with the lowest
This is partly because of the differences in states’
number of recent felonies was management of com-
racial and ethnic makeup and the widely differing
panies and enterprises:8 We estimate that only 125
policies associated with expungements and sealing
businesses were affected by the original PPP restric-
records among states (Burton et al., 2021). We did
tions and that the number was reduced by 89 percent
confirm that the background check data are regularly
under the revised PPP. Only one additional industry
updated to remove records that have been sealed or
had a reduction in affected businesses that was less
expunged. However, because of the number of miss-
than 90 percent (information, which experienced an
ing values, the data provided still were insufficient to
89-percent reduction).
produce reliable race and ethnicity estimates of the
The difference across states mentioned in the
differential impact of the changes in PPP restrictions
previous paragraph also could be caused by variation
at the state level.
in criminal justice policies. For example, the same
Although we do not have complete data on race,
crime could be a misdemeanor in one state but a
we find that the original PPP restrictions differen-
felony in another. This is particularly pertinent to our
tially affected Black individuals. The nationwide data
research because of the current progress of marijuana
that we have in which the race of the owner is known
legalization. Some states, such as Florida and Texas,
suggest that 24 percent of the businesses affected by
offer diversion programs to large numbers of people
the original restrictions were owned by Black indi-
after their first felonies; if an offender completes a
viduals. This percentage could be much higher if
diversion program, their arrest record might be sealed
Black owners are overrepresented among owners who
or expunged. This could radically alter the number of
have a missing racial status in the records.
people affected by felony restrictions. It also is at least
possible that states with more-punitive policies have
higher concentrations of minorities (McElhattan, Effects by Sex and Age
2021). Future research could examine how state crimi- The reduction in businesses affected after the revi-
sion of the PPP restrictions is estimated to be higher
8
FIGURE 4
Estimate of Businesses Affected by Original and Revised Paycheck Protection Plan
Restrictions, by Industry Category
Management of companies and enterprises
89% reduction
Utilities
98% reduction
Mining, quarrying, and oil and gas extraction
96% reduction
Public administration
97% reduction
Educational services Original restrictions
95% reduction
Information Revised restrictions
89% reduction
CIs
Arts, entertainment, and recreation
95% reduction
Finance and insurance
95% reduction
Real estate and rental and leasing
96% reduction
Agriculture, forestry, fishing, and hunting
93% reduction
Health care and social assistance
91% reduction
Wholesale trade
94% reduction
Transportation and warehousing
94% reduction
Manufacturing
96% reduction
Accommodation and food services
94% reduction
Professional, scientific, and technical services
94% reduction
Administrative and support and waste
management and remediation services 95% reduction
Other services (except public administration)
94% reduction
Construction
95% reduction
Retail trade
94% reduction
Unknown
95% reduction
0 20,000 40,000 60,000 80,000
Affected businesses
NOTE: Each industry was identified by its two-digit North American Industry Classification System code.
for businesses owned by men (96 percent reduc- Reductions in affected businesses tended to be
tion) than for those owned by women (92 percent about 95 percent for businesses with owners of any
reduction; see Table 3). Businesses with owners of age (see Table 4).
unknown sex had a reduction of 94 percent. However,
because business with owners of unknown sex might
not be evenly distributed between male and female, it
is important not to rely too heavily on this finding.
9
TABLE 2
Estimate of Businesses Affected by Original and Revised Paycheck Protection Plan
Restrictions, by Race/Ethnicity of Owner
Affected Businesses, Original PPP Affected Businesses, Revised PPP
Race/Ethnicity Restrictions (95-percent CI) Restrictions (95-percent CI) Reduction
White 93,640 (90,696–96,583) 4,782 (4,520–5,044) 95 percent
Unknown 79,279 (75,276–83,282) 4,357 (4,014–4,700) 95 percent
Black 31,620 (30,280–32,960) 1,981 (1,861–2,101) 94 percent
Hispanic 4,212 (3,904–4,519) 133 (97–168) 97 percent
Asian/Pacific Islander 2,753 (2,041–3,464) 100 (67–132) 96 percent
Other 715 (612–819) 103 (80–127) 86 percent
American Indian/Alaskan Native 437 (381–492) 24 (14–35) 95 percent
TABLE 3
Estimate of Businesses Affected by Original and Revised Paycheck Protection Plan
Restrictions, by Sex of Owner
Affected Businesses, Original PPP Affected Businesses, Revised PPP
Sex Restrictions (95-percent CI) Restrictions (95-percent CI) Reduction
Male 125,473 (122,099–128,847) 5,584 (5,301–5,868) 96 percent
Female 28,651 (26,910–30,391) 2,361 (2,204–2,518) 92 percent
Unknown 58,532 (54,971–62,094) 3,536 (3,222–3,849) 94 percent
TABLE 4
Estimate of Businesses Affected by Original and Revised Paycheck Protection Plan
Restrictions, by Age of Owner
Affected Businesses, Original PPP Affected Businesses, Revised PPP
Age (95-percent CI) (95-percent CI) Reduction
18–25 3,527 (3,045–4,009) 127 (97–156) 96 percent
26–35 29,698 (28,050–31,345) 1,453 (1,334–1,572) 95 percent
36–45 59,960 (57,410–62,510) 3,308 (3,099–3,517) 94 percent
46–55 54,046 (51,760–56,332) 3,129 (2,938–3,320) 94 percent
56–65 32,224 (30,573–33,874) 1,593 (1,466–1,720) 95 percent
>65 13,223 (12,159–14,287) 580 (481–679) 96 percent
Unknown 19,978 (17,051–22,905) 1,291 (1,005–1,577) 94 percent
State-Level Analysis: Minnesota We chose Minnesota and North Carolina for our
and North Carolina analysis because these states make criminal history
data available freely online, using name and date
Data and Methods of birth as identifiers. To obtain name and date of
Because we lacked direct visibility into the contents birth information for small business owners, we
and comprehensiveness of our national-level data, worked with the company DatabaseUSA to obtain
we conducted an additional analysis at the state level a random sample of data for small business owners
to examine the data and findings in more detail. (DatabaseUSA, undated). DatabaseUSA linked two
10
existing datasets: (1) a marketing dataset with infor- A final decision that we made before sampling
mation about businesses and (2) Exec@Home, an was how to determine an individual’s level of owner-
executive-level dataset that contained information ship in the business. Because of the varied nature of
about business executives. DatabaseUSA’s sources (e.g., business registration
We worked with DatabaseUSA to randomly records, yellow page listings, conference registra-
sample these linked business records across dif- tions), several personal titles (e.g., owner, president,
ferent industry sectors (see Table 5). We attempted vice president, manager, chief executive officer) could
to oversample two industries—construction and indicate a significant ownership interest in a small
hospitality—in which we expected felony prevalence business. However, to minimize potential ambigu-
would be higher because of anecdotal evidence.9 ity, we ultimately only sampled from Exec@Home
The sample we requested from DatabaseUSA records in which the individual’s title was owner.
was limited in its coverage. The initial dataset cov- If we included other titles, our prevalence findings
ered all small businesses. However, because we also would likely change but could be less accurate, as
needed owner names and dates of birth, we limited the PPP restrictions apply only to individuals with a
the sample frame to those Exec@Home records 20-percent or larger ownership stake.
that provide job title, name, and full date of birth. Using this final sample frame, we asked
Furthermore, much of the Exec@Home date of birth DatabaseUSA to randomly sample its records by
data only provide the month and year of birth. In state and business sector using our sampling targets.
those cases, DatabaseUSA populates the data for day The “received” column in Table 5 summarizes the
of birth with the first day of the birth month. As a number of records we received from DatabaseUSA.
result, actual first-of-the-month birth dates are indis- Table 6 provides an overview of the age dis-
tinguishable from placeholder dates. Therefore, we tribution of owners in the DatabaseUSA data
excluded all first-of-the-month birthdates from the versus the distribution of owners in the summaries
sampling frame, which eliminated 56 percent of pos- of the national consumer and background data
sible sample records. discussed in the previous section. One peculiar
TABLE 5
Business Owner Information Requested from DatabaseUSA
State Sector Received Criminal History Match Felony Within Five Years
Minnesota Hospitality 1,395 66 2
Construction 1,393 55 4
All other 1,389 36 1
North Carolina Hospitality 844 14 0
Construction 844 27 0
All other 395 9 1
Total 6,260 207 8
TABLE 6
Selected Age Distributions in Business Owner Data Sources
Percentage of Owners in Age Group
18 to 25 26 to 35 36 to 45 46 to 55 56 to 65 65 and older
DatabaseUSA 0 0.2 4.4 18.5 35.1 41.7
National Consumer and 1.1 11.4 20.8 23.2 19.4 12.5
Background Dataa
a 11.7 percent of ages were unknown.
11
feature of the sample of owners that we received data. Therefore, we likely had fewer false positives
from DatabaseUSA was the age of the individuals and more false negatives for both our North Carolina
included. DatabaseUSA’s age distribution was skewed overall criminal history matches and our North
much more toward older owners. Because older Carolina under-five-year criminal history matches;
individuals are less likely to commit crimes of any our estimate is likely lower than the real-world crimi-
kind, this means that we were likely to find a lower nal history prevalence.
prevalence of recent criminal histories through the For both states, we had information only for
DatabaseUSA owner sample. This skewed age dis- offenses committed within these states. We necessar-
tribution for business owners was a key limitation of ily missed any offenses committed in other jurisdic-
our approaches for Minnesota and North Carolina. tions. Again, this means that our estimates should
To check for criminal histories among the be seen as a lower bound. For the purposes of this
business owners from the DatabaseUSA sample, analysis, for any situation where multiple crimes were
we searched the public criminal history records in returned for a match, we retained only the most recent
Minnesota and North Carolina using name and date record.
of birth. We conducted these searches online for Given these limitations, we believe that false
Minnesota, using the state’s public criminal history positive and false negative problems are likely to be
website (Minnesota Public Criminal History, web- much more prevalent in this matching process than
page, undated-b). North Carolina provides a web- in a whole-record entity resolution algorithm match
based search and a database download option (North (as discussed in the earlier section).
Carolina Department of Public Safety, webpage,
undated-a; North Carolina Department of Public
Results
Safety, webpage, undated-b). For North Carolina,
we downloaded the database and ran the matches Ultimately, we estimate that a minimum of 3.4 per-
offline.10 (For both states, we only had conviction cent of the owners of small construction businesses
data, not nonconviction data.) and a minimum of 2.5 percent of the owners of
There were two key differences between these small hospitality businesses in Minnesota and North
two sources. First, Minnesota removes public crimi- Carolina had some criminal history (see Figure 5).
nal history records 15 years after an individual com- Overall, 2.5 percent of owners across all industries
pletes their sentence (Minnesota Public Criminal had some criminal history. This number is less
History, webpage, undated-a). North Carolina pro- than the 4.5 percent we found in the national data.
vides a much more comprehensive history of indi- Of those, less than 1 percent had felonies on their
viduals’ past criminal activity. This means that our records and less than 0.3 percent had a felony within
estimates of any criminal history for individuals in the past five years.
Minnesota will necessarily be lower than Minnesota’s Table 7 provides a further insight into these
real-world equivalent; as a result, our estimate for any numbers. The majority of data matches were for
criminal activity for Minnesota should be viewed as a owners with misdemeanors in their history who
minimum estimate. owned businesses with fewer than 25 employees.
Second, name-based matching approaches will Although we have information on businesses with
inevitably produce false positives (individuals incor- 100 to 500 employees in our “owner” data set, none of
rectly matched because they have the same name/ those owners were matched to criminal history data.
aliases and date of birth) and false negatives (e.g., Although these numbers of owners are small,
individuals with histories who were not matched they could represent a large number of employ-
because of inconsistency in name/aliases or date ees who could be affected if the small business
of birth). Minnesota’s criminal history search site they work for is disqualified from PPP benefits.
searches across known aliases associated with a According to the U.S. Census Bureau (2020), approxi-
particular date of birth. However, we did not have mately 6,000,000 small businesses employed nearly
additional aliases to match for the North Carolina 61,000,000 individuals. If all 0.3 percent of those
12
FIGURE 5
Estimated Prevalence of Criminal Histories Among Small Business Owners in
Minnesota and North Carolina
5.0
4.0
Prevalence (percentage)
3.43
3.0
2.48 2.54
2.0
1.0
0.82 0.86
0.38
0.22
0 0.09 0.13
All criminal All Within All criminal All Within All criminal All Within
history felonies 5 years history felonies 5 years history felonies 5 years
Construction Hospitality Other
NOTE: Data weighted to reflect the distribution of small businesses in the 2012 Survey of Business Owners.
TABLE 7
Estimates of Business Owners with Criminal History, by Sector, Severity of Crime, and
Business Size
Business Size
Sector Severity Under Five Employees Five to 24 Employees 25 to 99 Employees
Hospitality Misdemeanor 17 46 8
Hospitality Felony 1 5 1
Construction Misdemeanor 26 43 2
Construction Felony 2 9 —
Other Misdemeanor 23 16 1
Other Felony 4 — —
NOTES: These data include all criminal history records, not only felonies and misdemeanors committed with the past five years. These data include
records from both Minnesota and North Carolina.
13
businesses were at risk (or excluded), then roughly qualifying events under the original restrictions. This
180,000 businesses employing nearly 1.8 million indi- result is a combination of five factors:
viduals could be affected by felony restrictions. 1. being in prison
These figures are likely to be a lower bound in 2. being on parole
terms of the criminal history prevalence of small 3. being on probation
business owners. In particular, we are much more 4. having a pending felony charge
likely to have experienced false negatives in terms of 5. having one or more felony in the past five
name matching, and we do not know of crimes that years.
might have been committed in other jurisdictions
The final factor had the largest single impact.
(e.g., federal, other states).
Results varied by state: Texas had the highest
estimate of disqualified owners (3.2 percent) and
Comparison with Previous New Jersey and Pennsylvania had the lowest estimate
(0.6 percent).
Research Estimates
This important study had two limitations. First,
When the original PPP felony restrictions were cre- the researchers only examined data from seven states;
ated in March 2020, there were no direct estimates second, the researchers only considered sole propri-
of how many businesses might be affected by these etorships. Individuals involved in partnerships or
restrictions. In June 2020, Finlay, Mueller-Smith, other small businesses were not included. As a result,
and Street used the new CJARS to estimate that as this study did not provide a national picture of how
many as 1.7 percent of sole proprietorships might many small business owners have criminal history
be ineligible for the first version of the PPP because records. Therefore, federal programs that restrict
of current or prior criminal justice involvement. small business eligibility using owners’ criminal his-
CJARS has administrative criminal justice data from tory records lacked an information base for inform-
seven states; the U.S. Census Bureau then links the ing policy and program decisions about the impact of
criminal justice information with information on criminal history record restrictions. It makes sense
individual tax returns maintained at the U.S. Census to compare our estimates with the relevant numbers
Bureau. The tax data identifies sole proprietorships, a from Finlay, Mueller-Smith, and Street’s study.
subset of the small business types affected by the PPP In Table 8, we compare our state-specific esti-
restrictions. mates of prevalence of felony convictions in the past
According to Finlay, Mueller-Smith, and Street’s five years with the CJARS estimates of those with a
study, more than 250,000 self-employed business recent felony.
owners in these seven states had one or more dis-
TABLE 8
Comparison of RAND and Criminal Justice Administrative Records System Estimates of
Prevalence of Recent Felonies
State RAND National Estimate (percentage) RAND State Estimate (percentage) CJARS Estimate (percentage)
Texas 1.16 — 1.00
North Carolina 0.87 0.23 0.70
Pennsylvania 0.49 — 0.60
Wisconsin 1.63 — 0.70
Michigan 0.44 — 2.00
Minnesota 0.20 0.11 —
New Jersey 0.05 — 0.60
14
Our estimates are in broad agreement with the were explicitly prevented from participating in the
CJARS estimates. Their estimate in North Carolina PPP loan program as announced in March 2020,
(0.70 percent) is close to our estimate of 0.87 percent which disallowed applicants who had had a felony
(CI 0.67 to 1.06 percent), and our estimates are even conviction in the past five years (0.47 percent of small
closer to Finlay, Mueller-Smith, and Street’s estimates businesses have owners who have a felony within the
for Texas (1.16 percent and 1.00 percent, respectively) past five years, and 1.5 percent have owners who have
and Pennsylvania (0.49 percent and 0.60 percent, any felony record). After pushback from advocates
respectively). We found much lower rates than and policymakers, some of which was fueled by the
Finlay, Mueller-Smith, and Street did in Michigan timely research done by Finlay, Mueller-Smith, and
(0.44 percent versus 2.00 percent) and New Jersey Street, the Trump administration and then the Biden
(0.05 percent versus 0.60 percent) and higher rates in administration removed many of these restrictions.
Wisconsin (1.63 percent versus 0.70 percent). The restriction is now limited to those with convic-
Our estimates might differ from those of Finlay, tions for fraud, bribery, embezzlement, a false state-
Mueller-Smith, and Street for two reasons. First, we ment in a loan application, or an application for fed-
are using different sources of information. Second, we eral financial assistance (Hayashi, 2020).
use different definitions, particularly in the case of Building on Finlay, Mueller-Smith, and Street’s
business ownership. These differences have different research, we directly estimate that 201,174 more
effects. We might expect that the tax data on sole pro- businesses were eligible for PPP funding as a result
prietorship is more comprehensive than an assembled of the PPP revisions. The revisions also potentially
list of businesses, meaning that the estimates derived affected 325,665 employees. The 2021 policy change
from CJARS should be larger than ours. However, we represents an order of magnitude reduction in the
have a broader definition of business ownership than prevalence of businesses affected by the restrictions
Finlay, Mueller-Smith, and Street do. (0.47 percent to 0.03 percent).
We also note that our consumer and background Given the limitations of our data, we believe that
data provide only a limited window into the data this estimate is a lower bound on the total number
maintained in each state while CJARS has extensively of affected businesses. Using the states where we
curated their data in the seven states studied, with are more confident in the statewide court records
the most complete data in Michigan and Texas. As a coverage, our prevalence estimates increase 38 per-
result, we are comfortable concluding that our esti- cent (from 0.47 percent to 0.65 percent). As many as
mates are low for such states as Michigan (for which 81,000 businesses not captured in our data might be
our estimates are substantially lower than the CJARS positively affected by the changes in restrictions.
estimates). Although we did not have complete racial data,
the nationwide data that we do have suggest that
Black-owned small businesses were more heav-
Conclusions ily affected by the original PPP felony restrictions.
We estimate that 1.5 percent of businesses have Twenty-four percent of the businesses affected by
owners who have been convicted of a felony; 8 per- these restrictions were owned by Black individuals,
cent of the general adult population have ever been and the percentage could be much higher if Black
convicted of a felony (Shannon et al., 2017). The SBA owners are overrepresented among owners with a
has a long record of imposing restrictions on loans missing racial status. The restrictions also did not
and other aid to small businesses owned by individu- affect all industries equally (Lageson, Webster, and
als with criminal history records. This history could Sandoval, 2021). The impact was particularly large
partly account for business owners being less likely in the retail, construction, waste management, and
to have felony records than the general population. manufacturing sectors—sectors that historically
Among those small business owners who did have include high numbers of people with criminal history
felony convictions, we estimated that about a third records (Holzer, Raphael, and Stoll, 2006).
15
To the best of our knowledge, this is the first evidence to support conjectures about the ripple
time that commercially collected and indexed effect of PPP restriction decisions on employees of
criminal history data have been used by researchers the affected businesses. If small business owners with
to generate national estimates of the prevalence of criminal history records are more likely to hire those
criminal history records in the United States. In this with criminal history records, then the employees
report, we have compared our results from this new disproportionately affected by these decisions could
approach with results from other, more traditional be from racial minority groups that are more likely to
efforts, including research using CJARS data and our have criminal history records.
own attempts to directly estimate the numbers using Future research might help answer these impor-
publicly available information in North Carolina and tant questions by targeting industries, such as con-
Minnesota (Finlay, Mueller-Smith, and Street, 2020). struction, that have higher percentages of small busi-
The main challenge with any attempt to calculate nesses affected by criminal history record restrictions.
the prevalence of criminal history in any subpopu- Consumer and background check data could be used
lation is the completeness of the criminal history to identify businesses that have an owner or owners
record information in the background check data. In with criminal history records and that also received
the current study, we identified states in which the PPP loans. However, because consumer and back-
criminal history record data aggregator appeared to ground check companies expect to receive a fee for
have statewide access to court information. A com- each search they conduct, project costs when search-
parison of the prevalence estimates showed that most ing for collections of businesses (e.g., from the list of
of the states with the lowest estimates also were states businesses that received PPP loans) could rise more
in which the data aggregator did not have compre- rapidly than when searching the summary data that
hensive statewide court information. we used for this analysis. At the very least, estimates
Missing data about key characteristics of indi- from this report could be combined with information
viduals were a separate challenge. Without complete about the number of eligible businesses in each indus-
information in most or all records, it is difficult to try that actually received a PPP loan to provide a real-
answer some of the more detailed questions about istic number of the potential applicants who might
potential biases resulting from missing data. Further have been turned away because of restrictions.
documentation of the strengths and weaknesses of Because felony convictions often are a barrier to
the consumer and background data would make employment, entrepreneurship may well be a reason-
similar analyses with this kind of data more valuable. able alternative for individuals with felony records.
The initial goal of this project was to measure However, even these opportunities could be elimi-
the impact of PPP restrictions on small business nated if such agencies as the SBA do not help people
owners with criminal history records and on their with felony records. We estimate that only 1.5 percent
businesses. One key component of the restrictions— of businesses have owners with a felony record; how-
felony restrictions—was almost eliminated during ever, 8 percent of the U.S. population has a felony
the course of our research. Therefore, we were able record. Although there are many reasons why those
to quantify the potential benefit of this relaxation in with felonies might not own a business, aggressive
restrictions. Our data, which provided detailed infor- enforcement of criminal history record restrictions
mation on convictions, were particularly appropriate might further alienate an already marginalized
for estimating the impact of this repeal. However, subpopulation that needs access to the formal labor
because we had not anticipated the repeal, we were market to complete integration into society (Brayne,
unable to further estimate the positive effects of this 2014). Most research on re-entry after imprison-
decision. For example, we could not measure how ment does not fully engage with the possibilities
many now-eligible small businesses were initially and potential hurdles that face would-be entrepre-
denied PPP loans or how many small businesses neurs. As opportunities for gig work and other self-
closed because of the inability to get assistance under employment opportunities continue to expand, more
the original PPP restrictions. We also do not have needs to be done to understand the viability of this
16
path for those seeking to desist after a conviction.
This research could expand on existing work that
looks at the capital use and needs of minority-owned
firms as a separate and distinct group (Robb, 2012).
One unique (if accidental) feature of this work
is that it highlights the potential benefits of a policy
intervention (the elimination of the original PPP
felony restriction) designed to reduce the obstacles
faced by those with criminal history records rather
than another policy designed to limit their opportu-
nities. It is our hope that, as more policies are imple-
mented that aim to expand opportunities for those
with criminal history records, researchers can more
directly and purposefully estimate the harms and
benefits of these policies for the people and commu-
nities most directly affected by them.
Endnotes
lIt is unknown how these restrictions were implemented in prac-
tice. The SBA Inspector General’s office criticized the SBA for
generally lax administration (SBA, Office of Inspector General,
2021), and felony restrictions possibly were ignored. Nonetheless,
written restrictions could have discouraged applicants who did
not want to run afoul of federal rules (Brayne, 2014).
2 For example, when working directly with state criminal history
data, we checked for duplicate name and date of birth matches.
We found one name-date of birth duplicate match in 4,177
checks for Minnesota and 8,514 name-date of birth duplicates
in 1,347,810 records from North Carolina (a 0.02-percent and
0.6-percent duplicate rate, respectively). Given what we know
about the commercial entity resolution algorithm, we have every
reason to believe its error rates would be far smaller than what we
found with our much more naive approach.
3 The data from the aggregator indicated employee size as fewer
than 500, and therefore technically excluded companies with 500
employees, a group that would have been included in the SBA
definition.
4 The estimator can be written
YS WAAY, IT, — LA, —1;) IT},
i=]
in which A, is an indicator of knowing felony status, Y, is an indi-
cator of being affected by PPP restrictions, 7, is an estimate of the
propensity score (probability of knowing felony status), and yu, is
an estimate of the outcome model (probability of being affected
by PPP restrictions). W, indicates the count of businesses in the
grouping by business and owner characteristics.
The variance of the estimator can be written as
n
Y= 2,)W? (Y,- My 7,
i=l
This equation treats the felony missingness as the only source of
randomness because the consumer data are not a random sample
but a (near) census of criminal history records.
5 This number could be compared with the 27.6 million total
businesses with and without employees in the 2012 Survey of
Business Owners conducted by the U.S. Census Bureau (U.S.
Census Bureau, 2016).
6 An additional 130,000 felonies had occurred earlier than the
five-year cutoff, and 591,000 records were listed as misdemeanors
or infractions.
7 The company provided us with a list of sources from which they
collect their data. We identified states in which the aggregator
stated that they collected court data from a statewide source,
usually the state administrative office of the courts. We are not
publishing the list of states that the company considers to have
good statewide court coverage because the company considers
this information to be confidential and proprietary.
8 Management of companies and enterprises is a North American
Industry Classification System categorization that mostly applies
to offices of holding companies and bank holding companies
(North American Industry Classification System Association,
undated).
° We did not have the consumer and background check results
when we asked for the oversample and therefore did not focus on
retail, which was highlighted in the national-level consumer and
background data analysis.
10 Ror North Carolina, we conducted only exact name matches.
We did not attempt to use common nicknames, Soundex (which
searches for similar-sounding names), or any other name
alternatives for matching. The Minnesota website search has the
advantage of returning results across all known aliases (alterna-
tive name matches).
References
Arnold Ventures, “A Statement from Arnold Ventures on the
Biden-Harris Administration’s Expansion of the Paycheck
Protection Program,” press release, February 22, 2021. As of
May 21, 2021:
https://www.arnoldventures.org/newsroom/a-statement-
from-arnold-ventures-on-the-biden-harris-administrations-
expansion-of-the-paycheck-protection-program
Bang, Heejung, and James M. Robins, “Doubly Robust
Estimation in Missing Data and Causal Inference Models,”
Biometrics, Vol. 61, No. 4, December 2005, pp. 962-973.
Brayne, Sarah, “Surveillance and System Avoidance: Criminal
Justice Contact and Institutional Attachment,” American
Sociological Review, Vol. 79, No. 3, 2014, pp. 367-391.
17
Burton, Alexander L., Francis T. Cullen, Justin T. Pickett, Public Law 116-136, Coronavirus Aid, Relief, and Economic
Velmer S. Burton, Jr., and Angela J. Thielo, “Beyond the Eternal Security Act, March 27, 2020. As of May 14, 2021:
Criminal Record: Public Support for Expungement,” Criminology https://www.congress.gov/116/plaws/publ136/
and Public Policy, Vol. 20, No. 1, February 2021, pp. 123–151. PLAW-116publ136.pdf
Bushway, Shawn D., and Nidhi Kalra, “A Policy Review of Robb, Alicia, Access to Capital Among Young Firms, Minority-
Employers’ Open Access to Conviction Records,” Annual Review Owned Firms, Women-Owned Firms, and High-Tech Firms, San
of Criminology, Vol. 4, 2021, pp. 165–189. Rafael, Calif.: SBA Office of Advocacy, April 2012. As of April 15,
2021:
DatabaseUSA, homepage, undated. As of May 18, 2021: https://www.sba.gov/sites/default/files/files/rs403tot%281%29.pdf
https://databaseusa.com
SBA—See U.S. Small Business Administration.
Finlay, Keith, Michael Mueller-Smith, and Brittany Street,
Criminal Disqualifications in the Paycheck Protection Program, Shannon, Sarah K. S., Christopher Uggen, Jason Schnittker,
Washington, D.C.: U.S. Census Bureau, ADEP-WP-2020-04, Melissa Thompson, Sara Wakefield, and Michael Massoglia,
June 2020. As of May 14, 2021: “The Growth, Scope, and Spatial Distribution of People With
https://www.census.gov/content/dam/Census/library/working- Felony Records in the United States, 1948–2010,” Demography,
papers/2020/econ/cjars-ppp-adep-working-paper-20200622.pdf Vol. 54, No. 5, October 2017, pp. 1795–1818.
Hayashi, Yuka, “SBA Sued over Rule Barring Convicted Felons Talley, Terry, “Foreword,” in John R. Talburt, Entity Resolution
from PPP Loans,” Wall Street Journal, June 16, 2020. As of and Information Quality, Burlington, Mass.: Morgan Kaufmann
May 14, 2021: Publishers, 2011.
https://www.wsj.com/articles/sba-sued-over-rule-barring-
convicted-felons-from-ppp-loans-11592320399 U.S. Census Bureau, "Survey of Business Owners (2012)," data
set, September 2016. As of June 9, 2021:
Holzer, Harry J., Steven Raphael, and Michael A. Stoll, “Perceived https://www.census.gov/data/developers/data-sets/business-
Criminality, Criminal Background Checks, and the Racial owners.html
Hiring Practices of Employers,” Journal of Law and Economics,
Vol. 49, No. 2, October 2006, pp. 451–480. ———,“2017 SUSB Annual Datasets by Establishment Industry,”
data set, March 2020. As of May 14, 2021:
Hwang, Jiwon, and Damon J. Phillips, “Entrepreneurship as https://www.census.gov/data/datasets/2017/econ/susb/
a Response to Labor Market Discrimination for Formerly 2017-susb.html
Incarcerated People,” Academy of Management Proceedings,
Vol. 2020, No. 1, July 29, 2020. U.S. Small Business Administration, “Business Loan Program
Temporary Changes; Paycheck Protection Program—Additional
Lageson, Sarah, Elizabeth Webster, and Juan Sandoval, Eligibility Revisions to First Interim Final Rule,” SBA-2020-0039,
“Digitizing and Disclosing Personal Data: The Proliferation of 2020. As of May 18, 2021:
State Criminal Records on the Internet,” Law and Social Inquiry, https://www.sba.gov/sites/default/files/2020-06/PPP%20IFR%20
2021, pp. 1–31. 1%204.0%20%286.23.2020%20330pm%29%20SIGNED%20
6.24.20.pdf
McElhattan, David, “Punitive Ambiguity: State-Level Criminal
Record Data Quality in the Era of Widespread Background U.S. Small Business Administration, Office of Inspector General,
Screening,” Punishment & Society, February 18, 2021. Inspection of SBA’s Implementation of the Paycheck Protection
Program, Washington, D.C, Report Number 21-07, January 14,
Minnesota Public Criminal History, “Public Criminal History 2021. As of May 14, 2021:
Search FAQ’s,” webpage, undated-a. As of May 18, 2021: https://www.sba.gov/sites/default/files/2021-01/SBA%20OIG%20
https://chs.state.mn.us/Home/Faq Report-21-07.pdf
———, “Search Public Criminal History,” webpage, undated-b. As The White House, “Fact Sheet: Biden-Harris Administration
of May 18, 2021: Increases Lending to Small Businesses in Need, Announces
https://chs.state.mn.us/Search/ChsSearch Changes to PPP to Further Promote Equitable Access to Relief,”
North American Industry Classification System Association, Washington, D.C., press release, February 22, 2021. As of May 14,
“Management of Companies and Enterprises,” webpage, 2021:
undated. As of May 17, 2021: https://www.whitehouse.gov/briefing-room/statements-
https://www.naics.com/naics-code-description/?code=55 releases/2021/02/22/fact-sheet-biden-harris-administration-
increases-lending-to-small-businesses-in-need-announces-
North Carolina Department of Public Safety, “Downloads,” changes-to-ppp-to-further-promote-equitable-access-to-relief
webpage, undated-a. As of May 18, 2021:
https://webapps.doc.state.nc.us/opi/downloads.do?method=view
———, “Offender Search,” webpage, undated-b. As of May 21,
2021:
https://webapps.doc.state.nc.us/opi/offendersearch.
do?method=view
18
Acknowledgments
The authors would like to thank Tim Pinkerton of DatabaseUSA for his patience through hundreds of iterations
and refinements of their data request. The authors would also like to thank Amy Solomon and Jocelyn Fontaine
at the Arnold Foundation and Michael Mueller-Smith at the University of Michigan.
At the RAND Corporation, we would like to thank Sarah Lageson and Marek Posard for their careful
review of this report and John Pane for his careful review of our code.
19
C O R P O R AT I O N
The RAND Corporation is a research
organization that develops solutions to
public policy challenges to help make
communities throughout the world
safer and more secure, healthier and
more prosperous. RAND is nonprofit,
nonpartisan, and committed to the
About This Report public interest.
The Paycheck Protection Program (PPP) has provided funds for payroll, rent,
Research Integrity
mortgage, and utilities to businesses with fewer than 500 employees to help ease
Our mission to help improve policy and
the economic effects of the coronavirus disease 2019 pandemic. The PPP did not decisionmaking through research and
initially allow participation by companies owned by individuals with a criminal analysis is enabled through our core
values of quality and objectivity and our
background. This definition of criminal background included individuals with unwavering commitment to the highest
felony convictions in the past five years. level of integrity and ethical behavior. To
help ensure our research and analysis
Following pushback from policymakers and advocacy groups and a lawsuit are rigorous, objective, and nonpartisan,
we subject our research publications to
by the American Civil Liberties Union, the Trump administration limited the a robust and exacting quality-assurance
five-year restrictions to certain felonies; for all other felonies, the restrictions process; avoid both the appearance and
reality of financial and other conflicts of
were limited to a one-year window after conviction for those not incarcerated. interest through staff training, project
In February 2021, the Biden administration further relaxed the PPP restrictions, screening, and a policy of mandatory
disclosure; and pursue transparency
removing the one-year window.
in our research engagements
through our commitment to the open
In this report, RAND researchers estimate the number of small business owners publication of our research findings and
who have criminal history records. We also estimate how many small businesses recommendations, disclosure of the
source of funding of published research,
and employees were potentially affected by the initial PPP restrictions and how
and policies to ensure intellectual
these numbers changed under the revised 2021 restrictions. We use information independence. For more information, visit
from a consumer and background check company for some of our estimates, www.rand.org/about/principles.
which collects and organizes data in a wide variety of domains and which links RAND’s publications do not necessarily
information from individual criminal history records with information about reflect the opinions of its research clients
and sponsors. is a registered
company ownership. To the best of our knowledge, this is the first time that trademark.
commercially collected and indexed criminal history data have been used to
Limited Print and Electronic
generate national estimates of the prevalence of criminal history records in the Distribution Rights
United States. This document and trademark(s)
contained herein are protected by law.
This report was produced with the generous support of Arnold Ventures. The This representation of RAND intellectual
views expressed herein are the authors’ alone and do not represent the opinions property is provided for noncommercial
use only. Unauthorized posting of
or positions of Arnold Ventures. This report should be of interest to criminal this publication online is prohibited.
and civil justice researchers and to readers interested in the evolution of the use Permission is given to duplicate this
document for personal use only, as
of data algorithms in research. long as it is unaltered and complete.
Permission is required from RAND to
reproduce, or reuse in another form,
RAND Justice Policy Program any of our research documents for
commercial use. For information on
RAND Social and Economic Well-Being is a division of the RAND Corporation
reprint and linking permissions, please
that seeks to actively improve the health and social and economic well-being visit www.rand.org/pubs/permissions.
of populations and communities throughout the world. This research was
For more information on this publication,
conducted in the Justice Policy Program within RAND Social and Economic visit www.rand.org/t/RRA1295-1.
Well-Being. The program focuses on such topics as access to justice, policing,
© 2021 RAND Corporation
corrections, drug policy, and court system reform, as well as other policy con-
cerns pertaining to public safety and criminal and civil justice. For more infor-
mation, email justicepolicy@rand.org. www.rand.org
RR-A1295-1
File and source
- File
- RAND_RRA1295-1_ppp-criminal-record-eligibility.pdf
- Size
- 237,346 bytes
- SHA-256
- 2faa541229d7d5328e9db250d5dfe4f3e82db5d63cb03ba8fe9009ebe09f25d6
- Original
- www.rand.org