Entrepreneurship and the Gig Economy:

Agency decision

Ask Donna

What actually matters in this document.

Text

Evidence from U.S. Tax Returns∗

Matthew Denes†

Spyridon Lagaras‡

Margarita Tsoutsoura§

December 11, 2024

Abstract

Platform intermediation of goods and services has considerably transformed the U.S.

economy. We use administrative data on U.S. tax returns to study the role of the gig

economy on entrepreneurship. We find that gig workers are more likely to become

entrepreneurs, particularly those who are lower income, younger, and benefit from

flexibility. We track all newly created firms and show that gig workers start firms in

similar industries as their gig experience, which are less likely to survive and demonstrate higher performance. Overall, our findings suggest on-the-job learning promotes

entrepreneurial entry and shifts the types of firms started by entrepreneurs.

JEL Classification: G30, J21, J22, J24, L26

Keywords: Entrepreneurship, Gig Economy, Learning, Experimentation

†

Carnegie Mellon University. Email: denesm@andrew.cmu.edu.

University of Illinois Urbana-Champaign. Email: lagaras2@illinois.edu.

Washington University at St. Louis, CEPR, ECGI, and NBER. Email: tsoutsoura@wustl.edu.

We thank Paul Beaumont, Greg Buchak, Tony Cookson, Mike Ewens, Slava Fos, Yael Hochberg, Johan Hombert, Edith Hotchkiss, Katrin Hussinger, Emilie Jackson, Ankit Kalda, William Kerr, Cameron

LaPoint, Tong Liu, David Robinson, David Sovich, Rick Townsend, Melanie Wallskog, Sheng-Jun Xu, Ting

Xu, and Hanyi (Livia) Yi and conference participants at CEPR Endless Summer, CMU-Pitt-Penn State,

Corporate Finance Day, Craig Holden Memorial Finance, ECB-CEPR Labour Market, FIRS, FOM, FSU

SunTrust Beach, Labor and Finance Group, LBS Summer Finance, MFA, Munich Summer Institute, NBER

Entrepreneurship, NFA, Northeastern University, RCFS Winter, Red Rock, and WFA, and seminar participants at the Bank of Greece, Boston College, Drexel University, Georgia Tech, North Carolina State

University, University of Illinois at Urbana-Champaign, University of Maryland, University of Michigan,

University of Oregon, University of Oxford, University of Rochester, and U.S. Department of Treasury for

helpful comments. We gratefully acknowledge funding support from the Annie E. Casey Foundation, the

Block Center for Technology and Society, the Ewing Marion Kauffman Foundation, and the W.E. Upjohn

Institute for Employment Research. This research is conducted through the Joint Statistical Research Program of the Statistics of Income Division of the IRS. Terry Cheng, Caitlin Hartley, Casey Li, Maxwell Sacher,

Jacob Triplett, and Collin Zoeller provided excellent research assistance. We thank many people at the IRS

for their assistance with and support of our research, including Anne Herlache, Alicia Miller, and Tomas

Wind. The views and opinions presented in this paper reflect those of the authors and do not necessarily

reflect the views or the official position of the Internal Revenue Service. All results have been reviewed to

ensure that no confidential information is disclosed.

‡

Introduction

Labor markets play a central role in an individual’s decision to become an entrepreneur

(Hombert et al. (2020), Gottlieb, Townsend, and Xu (2022), and Hacamo and Kleiner (2022)).

Relying on recent technological advancements, the gig economy has disrupted labor markets

and reshaped income opportunities for many individuals. There has been a correspondingly

large take-up, with nearly 10 million people participating in the U.S. gig economy over the

past decade. Characterized by relatively low entry costs and flexibility, the gig economy

could reduce uncertainty and encourage experimentation, which are vital components of entrepreneurship (Manso (2011) and Kerr, Nanda, and Rhodes-Kropf (2014)). In this paper,

we use novel data linking gig workers to their newly created firms and provide the first

evidence about how the gig economy alters the profiles of new entrepreneurial ventures.

Uncertainty is an inherent element of entrepreneurship. The gig economy might influence the risks faced by potential entrepreneurs in several ways. First, opportunities in the

gig economy could mirror the experiences of an entrepreneur, allowing individuals to learn

about entrepreneurship and accumulate industry-specific experience. Second, the costs of

experimentation impact innovative activities (Ewens, Nanda, and Rhodes-Kropf (2018)).

The gig economy may reduce these costs and provide an additional source of startup capital.

Third, it lowers downside risk by providing entrepreneurs with the ability to smooth income.

By encouraging learning and supporting risk taking by prospective founders, the gig economy may facilitate the creation of firms that eventually grow larger and disproportionately

contribute to economic growth.

To study the interaction between the gig economy and entry into entrepreneurship, we

use administrative data from federal tax returns filed with the Internal Revenue Service

(IRS) on the universe of firms and individuals in the United States. For every year in our

sample from 2012 to 2021, the micro-level information allows us to observe income received

in the gig economy for each individual linked with entrepreneurial entry.1 We also track firms

at founding and their subsequent performance over time, in addition to the characteristics of

entrepreneurs. Using these novel data, we seek to understand how the gig economy influences

the types of new firms started by gig workers and offer insights about mechanisms spurring

new firm creation.

In our first set of analyses, we study the effect of the gig economy on entry into entrepreneurship. We compare the likelihood of starting a new firm for all individuals in the

U.S. aged 25 to 65 who previously received income from the gig economy relative to those

who have not participated in it. We find that gig workers are 1.0 percentage points more

likely to create a new firm. This estimate holds across a variety of specifications, where the

strictest model absorbs time-varying local economic conditions and saturates the model with

granular fixed effects for individual characteristics to account for differences in income and

age, in addition to other controls. We also separately examine entrepreneurs who start a

firm for the first time and show that they account for about three-quarters of the effect of

the gig economy on entrepreneurial entry.

Though we include a rich set of individual-level controls, there could be a potential

concern about omitted variables that might be correlated both with working in the gig

economy and entry into entrepreneurship. To strengthen our evidence, we leverage plausibly

exogenous variation in the availability of the gig economy both geographically and over

time following the methodology of Jackson (2022). This approach relies on the staggered

availability of gig work across counties over time. Using this empirical design, we find that

gig workers remain 1.0 percentage points more likely to start a new firm, which is also

statistically quite similar. We continue to use this methodology in the following analyses.

We also evaluate the potential influence of omitted variables following the approach of Oster

(2019). The Oster test indicates that the bias-adjusted coefficients are quite close to the

We observe an individual’s income from the gig income using information returns filed by those firms

operating in the gig economy. We measure firm creation in tax returns using the most common type of firms,

which are sole proprietorships. Sections 2.2 and 3.1 explain how we construct data on gig income and firms,

respectively. We exclude firms that are mechanically created due to tax reporting requirements.

baseline estimates. The stability of the estimated coefficients across specifications and the

findings from the Oster test suggest that it is unlikely an omitted variable drives the result.

We continue our individual-level analyses by evaluating the characteristics of gig workers

who respond by creating new firms. We investigate the role of capital constraints, lifecycle

considerations, and flexibility in spurring entrepreneurial entry. We find that individuals

who participate in the gig economy and have relatively lower incomes are more likely to

start new firms. We measure income using adjusted gross income, income distribution in a

particular county-year, and whether an individual received an income-related tax credit. We

also show that the propensity to become an entrepreneur is higher for younger gig workers.

Further, the gig economy is expected to particularly benefit individuals who value flexibility.

To proxy for individuals who might benefit from flexibility, we examine all individuals with

dependents as well as single filers with dependents. We find that both groups are significantly

more likely to become entrepreneurs when participating in the gig economy. Taken together,

the first set of results links gig workers with entrepreneurial entry, highlighting shifts in the

profile of responsive individuals.

Next, we study the universe of firms created in the United States from 2012 to 2021.

For each firm, we determine whether an entrepreneur participated in the gig economy before

the firm was created (“gig founder”) or if the individual did not receive gig income prior

to starting the firm (“non-gig founder”). This distinction allows us to evaluate how firms

started by entrepreneurs with previous experience in the gig economy differ from other firms.

To our knowledge, this is the first paper that links individuals deriving income from the gig

economy with the firms that they create.

We start our firm analyses by asking two related research questions. First, how does

the industry composition of firms created by gig founders compare to non-gig founders?

For all firms in our sample, we determine the share of newly created firms in a particular

industry separately for gig- and non-gig-founded firms. We find that gig workers create

a higher share of firms in personal services, trade, and transportation. We corroborate

the industry composition of gig-founded firms by exploring the transition of gig workers into

entrepreneurship based on their experience in the gig economy. In this approach, we calculate

the proportion of gig workers who start a firm in a particular industry based on the type of

gig firm from which they received income. We show that gig founders often transition from

the gig economy into similar industries. Second, what is the role of prior work experience

for entrepreneurs? Using the universe of earnings for salaried employees and independent

contractors, we find that entrepreneurs are significantly more likely to start a firm in an

industry where they have prior work experience. We separately examine the importance of

experience in the gig economy and show that it remains elevated for gig workers. Combined,

this evidence suggests that learning might be an important mechanism in the gig economy

for facilitating entry into entrepreneurship.

In the next set of analyses, we investigate firm-level characteristics at founding and

subsequent performance. We find that gig-founded firms are significantly larger, both in

terms of revenues and number of employees, relative to firms created by individuals not

participating in the gig economy. These estimates hold when we include county-year fixed

effects to absorb time-varying local differences and industry fixed effects to compare firms

created in the same industries.

We now turn to evaluating firm performance. For each firm in our sample, we track

whether the firm survives to a particular year and, conditional on surviving to this period, we

determine its performance using profitability and employment. We construct these measures

for years one to three after founding because many firms are recently created. We find that

the likelihood of survival for gig-founded firms is 2.6 to 3.3 percentage points lower relative

to the probability of survival for non-gig founded firms, which is a 3.8% to 7.3% decrease

relative to the respective sample mean. These estimates continue to include county-year

and industry fixed effects. When we examine measures of performance, we show that the

profitability of gig-founded firms is about 39.4% to 46.9% higher relative to firms with a

founder who did not participate in the gig economy. Additionally, gig-founded firms are

both more likely to have employees and to operate firms with a relatively high number of

employees, defined as having at least five employees.

Overall, the firm-level evidence offers insights into the role of the gig economy on entrepreneurship through experimentation and learning. When gig workers establish a new

firm, they might bear more risk by starting larger firms. Consistent with experimentation,

these firms appear to be riskier as they survive for shorter periods of time, yet they realize

higher performance. Shorter survival, on average, is also consistent with gig founders learning about the prospects of their firms more quickly and shutting down less promising firms

sooner. These findings indicate that gig workers might learn on-the-job and relate to recent literature highlighting the importance of experimentation in entrepreneurial endeavors

(Manso (2011) and Kerr, Nanda, and Rhodes-Kropf (2014)).

We also examine the employment decisions and capital structure of firms started by

gig workers. The administrative data on U.S. tax returns allow us to track the number of

workers at a firm and whether employees are salaried workers or independent contractors.

For every firm in our sample, we measure the extensive margin based on whether a firm

has any independent contractors and the intensive margin using the number of independent

contractors employed. Similar to our previous analyses, we track employment for one to

three years after founding. On the extensive margin, we find that gig-founded firms are 9.5%

to 18.2% more likely to use independent contractors relative to firms with a non-gig founder.

We also evaluate the intensive margin and show economically large increases in the number

of independent contractors hired by gig-founded firms. These findings further highlight that

gig workers may learn by transferring their knowledge and experience from the gig economy

to the new firms that they create.

In the final set of firm-level analyses, we explore the capital structure of newly created

firms. We use reported information on a firm’s interest expense to measure whether a firm

has debt by a particular year after founding. We find that the likelihood of having debt

is significantly higher for gig-founded firms. In economic terms, there is a 10.2% to 17.9%

increase in the probability of a gig-founded firm having debt. This suggests that the gig

economy might enable gig workers, particularly those who are capital constrained, to access

capital markets.

We conclude our analyses by evaluating how gig founders fare after starting a new

firm. Prior papers highlight that entrepreneurs may start firms for non-pecuniary reasons

(Moskowitz and Vissing-Jørgensen (2002)), though there is substantial real option value

embedded in the decision to create firms (Manso (2016)). In this analysis, we evaluate the

effect of being a gig worker on a founder’s subsequent income. We track both whether a

founder’s income increases in the years after founding and if a founder’s rank in the income

distribution changes. We find that gig founders have higher subsequent income compared to

founders who did not participate in the gig economy. We also show that their rank in the

income distribution is more likely to rise. This highlights that gig founders are better off

than other entrepreneurs in terms of income.

In sum, we use administrative data on U.S. tax returns to study gig workers and the

firms that they create. Gig income appears to facilitate entry into entrepreneurship and

might allow gig workers to learn about becoming entrepreneurs. The outcomes at newly

created firms started by gig founders suggest that gig workers experiment with new ideas.

Our findings are related to the growing literature on the factors impacting entrepreneurial

entry by demonstrating the crucial interactions between labor markets and entrepreneurship.

Furthermore, this work provides tax administration with a foundational understanding of

the potential service needs of newly created firms and how these needs may differ based on

firm characteristics.

Our paper broadly contributes to the literature on entrepreneurship. Newly created

firms support economic growth and spur new jobs (Haltiwanger, Jarmin, and Miranda

(2013)). Accordingly, there has been a large focus on factors influencing entrepreneurial

entry. Capital constraints often limit new firm creation (e.g., Evans and Jovanovic (1989),

Hurst and Lusardi (2004), Cagetti and De Nardi (2006), and Bianchi and Bobba (2013)). A

related stream of papers shows that cash windfalls often lead to more entrepreneurs (Bellon

et al. (2021) and Cespedes, Huang, and Parra (2023)). Regulations reduce entrepreneurship

(Klapper, Laeven, and Rajan (2006)), while banking deregulation spurs creative destruction

(Kerr and Nanda (2009)). Entrepreneurship, using similar data as in this paper, increases

with higher credit limits and credit scores (Herkenhoff, Phillips, and Cohen-Cole (2021)).

Government programs frequently target entrepreneurial activity, though they can be unsuccessful (Denes et al. (2023)). A connected set of papers argues that experimentation is a key

ingredient of entrepreneurship (Manso (2011) and Kerr, Nanda, and Rhodes-Kropf (2014)),

and that lower costs increase innovation (Ewens, Nanda, and Rhodes-Kropf (2018)). Additionally, the option value of entrepreneurship plays an important role in experimentation

(Manso (2016) and Catherine (2022)). We link gig workers with their newly created ventures to highlight the importance of labor markets in supporting entrepreneurship through

learning and how this might promote experimentation.

Our paper is most closely related to Barrios, Hochberg, and Yi (2022) and Mao et al.

(2023), which study aggregate measures of entrepreneurship following the entry of a gig firm.2

Barrios, Hochberg, and Yi (2022) show that the launch of ride-hailing services is associated

with an increase in new business registrations and internet searches about entrepreneurship.

Mao et al. (2023) find that the introduction of short-term rentals is positively related to

new firm creation. We corroborate their findings and differ in several important dimensions.

First, we use granular administrative data from U.S. tax returns to track individuals participating in the gig economy and the firms that they start, overcoming prior limitations

to connecting gig income with entrepreneurial entry. Second, and related, this allows us to

provide suggestive evidence on mechanisms that could support the creation of new firms,

which might include learning by gig workers who could experiment by starting riskier firms.

Third, we study the universe of firms and individuals using the gig economy in the U.S.,

Burtch, Carnahan, and Greenwood (2018) is an earlier paper about a ride-hailing service from 2012

to 2015, and explores its relationship with self-employment based on crowdfunding activity in Kickstarter

and the Current Population Survey. They report a negative association between these measures and the

availability of a ride-hailing service.

capturing a wide range of gig firms and corresponding economic activity.

More generally, we also add to the growing literature on the gig economy. A relatively

early group of papers documented the size of the gig economy using survey and administrative

data (Abraham et al. (2021), Collins et al. (2019), and Lim et al. (2019)). A series of recent

papers examine how unemployed individuals use the gig economy (Jackson (2022) and Fos

et al. (2024)). Flexibility is often valuable for gig workers (Hall and Krueger (2018) and

Chen et al. (2019)). Additionally, access to financing impacts participation at ride-hailing

gig firms (Buchak (2024)).

Our findings are also related to work at the intersection of finance and labor. One

strand of this literature studies how financial distress affects an individual’s decision to start

a firm (Babina (2020) and Hacamo and Kleiner (2022)). Further, labor-related regulation can

influence entrepreneurial entry through increases in different types of employment insurance

(Hombert et al. (2020) and Gottlieb, Townsend, and Xu (2022)). We focus on the effect of

the gig economy, which is a major disruption to labor markets, and its role in newly created

firms by gig workers.

Gig Economy

In this section, we describe the gig economy in the United States and how we measure it

using U.S. tax returns. Section 2.1 provides information on the platforms in the gig economy

that we study in our analyses. Section 2.2 details how income from gig work is observed for

all individuals in the United States. Section 2.3 provides summary statistics on economic

activity in the gig economy.

2.1

Gig Economy in the United States

The gig economy typically refers to short-term income opportunities. While this part of the

economy existed long before the 2000s, this paper specifically studies recent technological

advancements that substantially expanded the size of this market. Mobile devices allow

platforms to match customers with workers across numerous goods and services. This has

disrupted many industries and altered the income opportunities for an increasingly large

number of individuals. Throughout the paper, we refer to the “gig economy” to capture

these recent changes, including the related platforms (“gig firms”) and individuals receiving

income from these arrangements (“gig workers”).

In the United States, many gig firms were founded in the first decade of the 2000s.

However, the corresponding economic activity started to rise in the following decade. In

Section 2.3, we show that more than a million individuals received income from gig firms in

2015. We focus on gig firms that broadly fall into four categories. First, we include platforms

offering transportation services to customers. Second, we incorporate gig firms that allow

individuals to monetize their assets. Third, we add marketplaces providing opportunities

for individuals to sell goods. Fourth, we include gig firms where individuals can provide

short-term services that generally require specific skills.

As the gig economy continues to grow, so has the debate surrounding its benefits and

costs. In this paper, we study whether the gig economy enables gig workers to become

entrepreneurs and evaluate the firms that they create. If gig workers are better off by

pursuing entrepreneurial entry, which we examine in Section 6, then new firm creation could

be considered a benefit, along with the flexibility afforded by the gig economy. Yet it is also

important to highlight costs that might be borne by gig workers. These include the loss of

worker protections because gig workers are independent contractors (Ravenelle (2019)) and

negative spillovers, such as driving fatalities (Barrios, Hochberg, and Yi (2023)).

2.2

Measuring the Gig Economy

We use administrative data from U.S. tax returns to measure participation in the gig economy. This allows us to directly observe income received by individuals in the U.S. from

firms operating in the gig economy. This section describes how we construct an individual’s

income from the gig economy for each year. We primarily use information returns provided

to the IRS by gig firms supplemented with data from tax returns of individuals.

We manually compile a list of gig firms in the U.S. We begin with a list developed by

previous researchers using tax returns to study different aspects of the gig economy (Collins

et al. (2019)), which contains about 50 gig firms. We conduct extensive internet searches to

expand the coverage of the gig economy. We classify each gig firm into one of the following

categories: leasing, selling, services, and transportation. Due to confidentiality reasons, we

cannot identify specific firms in the underlying data. The combined lists include a total of

174 gig firms. Appendix Figure A1 shows the distribution of gig firms in our sample. About

half of gig firms are in the services sector. Almost 50 firms are operating in the transportation

sector. The remaining gig firms are equally in the leasing and selling sectors.

We observe an individual’s participation in the gig economy using information returns

provided to the IRS by gig firms. Specifically, we use Forms 1099-MISC, 1099-NEC, and

1099-K to measure income received by an individual from gig firms.3 Though tax returns

can be filed jointly, information returns identify the exact individual who received income

from a gig firm. We construct a dataset of gig income using the list of 174 gig firms matched

to the universe of Forms 1099-MISC, 1099-NEC, and 1099-K. We augment these data by

using individual tax returns. Gig workers generally file a Schedule C as part of Form 1040

to report income derived from the gig economy. This schedule includes a description of the

activity related to its filing. If this description includes the name of a gig firm in our list,

we add it to the dataset on gig income. The combined dataset using Forms 1099-MISC,

1099-NEC, and 1099-K, in addition to Schedule C, allows us to track the gig income that is

received by an individual over time.

While U.S. tax returns offer novel insights into individuals participating in the gig

There are thresholds for reporting information using these forms. Gig firms are required to report when

individuals receive at least $600 in non-employee compensation. Form 1099-NEC replaced Form 1099-MISC

for reporting non-employee compensation in 2020. Form 1099-K is used by certain gig firms classified as

third-party networks and has higher thresholds. It is based on transactions and required when the total

income from these transactions is higher than $20,000 and there are more than 200 transactions.

economy, a caveat should be mentioned about the data. We primarily rely on information

returns provided to the IRS by gig firms to observe gig income. The requirement to provide

this information is generally based on the amount of income or the number of transactions

paid to an individual. Accordingly, we do not observe gig income below these thresholds

unless gig firms voluntarily provide the information to the IRS. The vast majority of gig

firms only provide information returns to the IRS if required. However, it is also important

to note that we rely on information returns provided by gig firms to the IRS, rather than

individuals reporting gig income on Form 1040.4

[Insert Figure 1 Here]

Figure 1 maps participation in the gig economy from 2012 to 2021. For each state,

we determine the number of individuals who participated in the gig economy during the

sample period relative to the state’s labor force in 2021. Darker blue shading represents a

larger share of participation in the gig economy. The map highlights that there has been

substantial participation in the gig economy during the past decade. At the top quartile of

states, 9% to 20% of individuals have received income from the gig economy. These states

are largely represented by coastal states and Illinois. In sum, there has been striking and

broad participation in the gig economy throughout the United States.

2.3

Summary Statistics on the Gig Economy

This section provides summary statistics on the gig economy in the United States. We use

administrative data from U.S. tax returns described in Section 2.2. The sample period goes

from 2012 until 2021. We start in 2012 since gig activity prior to this year is minimal. We

end in 2021 since this is the last year when data are currently available.

[Insert Figure 2 Here]

There can be underreporting or no reporting of gig income by a tax filer on Form 1040. Consequently,

it is important to use Forms 1099-MISC, 1099-NEC, and 1099-K to capture a substantial share of activity

in the gig economy.

In Figure 2, Panel A shows the cumulative number of individuals who have worked in

the gig economy by a particular year. We focus on the cumulative number to capture the

extent of participation in the gig economy over the past decade. We find that the number of

U.S. individuals who received income in the gig economy has markedly increased, starting

at less than one million individuals in the first few years to about 10 million in the last year

of the sample. Panel B provides the total amount of income in billions of dollars received

by gig workers in a particular year. We adjust this to real terms by converting to dollars

in 2012. We show that gig income follows a similar trajectory, growing from less than $10

billion at the beginning of the sample period to almost $120 billion in 2021. Overall, this

figure demonstrates the substantial rise of the gig economy in the United States.

[Insert Table 1 Here]

Table 1 provides summary statistics on the U.S. gig economy. Panel A tabulates the

number of individuals participating in the gig economy each year and shows information on

their income. We find that the number of gig workers exceeds one million individuals in

2015 and mostly rises in subsequent years.5 Mean gig income is winsorized at the 1% level in

each tail to reduce the influence of outliers and converted to dollars in 2012. It ranges from

$8,000 to $20,000 over the sample period. Due to confidentiality reasons, all income values

reported in the paper are rounded to thousands. We also show that gig income received

by individuals is often substantial, with about a quarter of gig workers earning more than

$10,000 and at least 10% obtaining more than $20,000 in every year of the sample.

Panel B shows the characteristics of individuals in the first year they received gig income. Adjusted Gross Income (AGI) is a gig worker’s adjusted gross income converted to

dollars in 2012 and W -2 Income is a gig worker’s W-2 income converted to dollars in 2012.

Receives EIT C is an indicator variable equaling one if a gig worker receives an Earned

The number of gig workers declines in the middle of the sample since some gig firms no longer submitted

information returns that they were not required to file with the IRS. This also explains why mean gig income

is relatively lower in 2015 and 2016.

Income Tax Credit. The means of these variables suggest that gig workers are often lowerincome individuals. The average age of gig workers is 39 and the majority of gig workers are

male. Gig workers are usually single and just under half have dependents.

Last, we examine differences in the number of gig workers by type of gig firm. Appendix

Figure A2 plots U.S. participation in the gig economy from 2012 to 2021 based on gig firms

classified as transportation relative to non-transportation, which includes leasing, selling, and

services. Panel A shows the cumulative number of individuals working in the gig economy and

Panel B provides the total amount of income in billions of dollars received by gig workers

in a particular year, which is converted to dollars in 2012. The lighter gray bars show

participation in the gig economy for gig firms categorized as transportation. The darker red

bars indicate participation in the gig economy for gig firms categorized as non-transportation.

While the trends are similar for both types of gig firms, the number of gig workers and gig

income have grown at a higher rate for transportation gig firms.

Data from U.S. Tax Returns

This section details how we use U.S. tax returns to study entrepreneurship. These data

allow us to overcome several challenges with linking new firm creation to participation in

the gig economy, which is described in Section 2. First, comprehensive data on every new

firm in the economy is generally not provided in publicly available datasets. Second, it can

be difficult to observe the performance of new firms at and following their creation. Third,

the characteristics of founders, their prior labor income, and employees at their firms are

usually unavailable. Federal tax returns represent a new and largely unexplored approach

to studying entrepreneurship in the U.S. economy. Section 3.1 explains how we measure

entrepreneurship. Section 3.2 provides information on additional data incorporated into the

analyses and summary statistics.

3.1

Measuring Entrepreneurship

Our paper seeks to understand the role of the gig economy on entry into entrepreneurship.

We use the universe of U.S. tax returns to determine when individuals start new firms. To

measure entrepreneurial activity across a wide swath of potential entrepreneurs, we focus on

sole proprietorships. This is motivated by several considerations. First, sole proprietorships

are the most common type of firm in U.S. tax returns. Individuals participating in the gig

economy are potentially more likely to form a sole proprietorship when starting a new firm

relative to other firm types, which include partnerships and corporations. Second, focusing

on one firm type allows us to construct standardized measures of firm outcomes. Third, we

observe ownership of sole proprietorships and these types of firms are wholly owned by one

individual.

In U.S. tax returns, sole proprietorships file Schedule C, which is part of a household’s

Form 1040. This schedule identifies the specific entrepreneur owning and operating the firm

within a household. To construct a dataset on firms for our analyses, we start with the

universe of Schedule C filings, which are available from 1997 to 2021. While the sample

for our analyses is from 2012 to 2021, using data back to 1997 allows us to identify when

individuals are first-time entrepreneurs. We restrict our attention to Schedule C filings that

include an employer identification number (EIN) to focus on firms that are separate entities.6

If a particular tax return is amended, we use the most recent filing available.7

An important aspect to consider when using Schedule C filings to measure entrepreneurship is alternative reasons why taxpayers might file this schedule. In the context of this paper,

individuals participating in the gig economy are generally required to file a Schedule C to

report gig income, which does not represent entrepreneurial activity. As mentioned in the

The main requirement for firms to have an EIN is they file employment returns or have a qualified

retirement plan. Additional details about requirements for having an EIN are available on the IRS website

at: https://www.irs.gov/instructions/i1040sc.

We also apply the following filters to construct the dataset. First, we only use Schedule C filings with

valid zip codes. Second, we remove filings where the same EIN appears on a Schedule C for a different Form

1040 in the same year. Third, we drop filings where the EIN is the same as a social security number (SSN)

or the SSN is used as an EIN.

preceding paragraph, we only use Schedule C filings with an EIN. This restriction will remove gig workers who file a Schedule C solely to report gig income and do not have an EIN.

We incorporate two additional steps to eliminate Schedule C filings solely used to report gig

income. First, using the data on gig income described in Section 2.2, we drop firms where

the reported income is within a narrow band of the gig income. Specifically, we remove those

Schedule C filings where the gross receipts or sales is within $100 of the gig income received

by an individual in a particular year. Second, we remove Schedule C filings where the firm

name matches the name of a gig firm. Overall, we implement several approaches to remove

filings that stem from tax reporting requirements.

Next, we compare our measure of entrepreneurship to related papers using similar data.

Our approach for defining entrepreneurship is closely related to Herkenhoff, Phillips, and

Cohen-Cole (2021), who use Schedule C filings available through the U.S. Census Bureau

to measure entrepreneurial activity. This definition of entrepreneurship focuses on unincorporated entrepreneurs. Their paper matches data at the individual-year level to a credit

bureau and requires firms to exist in Census datasets, reporting a sample average of 0.4%

for firm ownership with employees. Also related to our paper, Bellon et al. (2021) use data

at the individual-year level from a credit bureau to measure self-employment, which captures unincorporated entrepreneurs, and business ownership, which includes incorporated

entrepreneurs. In their paper, the sample mean for self-employed is 2.0%. Table 3 shows

that 0.7% of our sample are founders, which is also at the individual-year level. Our sample mean for founders is below Bellon et al. (2021) since we require a Schedule C filing to

have an EIN and remove these filings that stem from reporting gig income. Our average

is above Herkenhoff, Phillips, and Cohen-Cole (2021) because we do not impose that firms

have employees, or that they match to credit bureau data or Census datasets.

[Insert Figure 3 Here]

Figure 3 maps the geography of entrepreneurship in the United States from 2012 to

2021, which is the sample period for our analyses. We determine the number of new firms

created in a state for a particular year relative to the total number of new firms created in

the U.S. in a particular year. For each state, we average the share of new firms created in

the state across years. Then, we determine the quartile ranking across states. Darker blue

shading indicates a larger share of firms created in a particular state. New firm creation is

highest in California, the East Coast, Florida, Illinois, and Texas. These patterns are broadly

similar to aggregate patterns of entrepreneurship (Andrews et al. (2022)), supporting our

measure of entrepreneurship using U.S. tax returns.

[Insert Figure 4 Here]

Figure 4 shows a map of the relationship between the gig economy and entrepreneurship

across the United States. For each state, we determine the correlation between the yearly

count of new firms created and the number of individuals participating in the gig economy

during the previous year. Then, we determine the ranking across states. Darker blue shading

indicates a higher positive correlation between new firm creation and participation in the

gig economy in a particular state. The correlation is highest mainly in New York, Texas,

and the Southeast, indicating areas of the United States where both the gig economy and

entrepreneurship are comparatively larger.

3.2

Other Data and Summary Statistics

We incorporate additional data from U.S. tax returns for individuals and entrepreneurs. We

use adjusted gross income, filing status, dependents, and Earned Income Tax Credit (EITC)

from Form 1040. For adjusted gross income, we also construct income percentiles by countyyear. Filing status indicates whether a taxpayer is a single or joint filer in a particular year.

We add data on age and gender using information from the Social Security Administration.

We use data from Schedule C to construct firm outcomes at founding and subsequent

performance. Since we focus on sole proprietorships with an EIN, we use this unique iden16

tifier to track firms over time. We construct variables to measure firm survival, revenues,

employment, and profitability. We determine if a firm hires an independent contractor using

the universe of Forms 1099-MISC, 1099-NEC, and 1099-K in a particular year. We do not

focus specifically on gig firms since we cannot directly observe in tax returns if a firm hires

a gig worker through a gig firm. We use interest expense to determine if a firm has debt.

[Insert Table 2 Here]

Table 2 provides summary statistics for variables used in our analyses. Panel A shows

variables for individual analyses, which includes the universe of individuals in the U.S. from

2012 to 2021 aged 25 to 65 in the year of filing a tax return. We form this sample by

splitting tax returns filed by a household with more than one person to separately include

an observation for the primary filer and for the spouse. These variables include nearly 1.3

billion individual-years. About 1.4% of the sample are gig workers prior to year t. Almost

0.7% of individuals start a new firm in a particular year. Nearly 20% of the sample receives

an Earned Income Tax Credit and half of individuals have dependents. Panel B includes

variables for the cross-section of entrepreneurs. About 3.5% of founders are previously gig

workers. Panel C has the variables for firm outcomes. More than half of firms do not survive

three years after founding. Approximately 10% of firms use independent contractors in years

one, two, and three following their opening.

Entrepreneurship and the Gig Economy

In this section, we use administrative data on the population of U.S. tax filers to study

the role of the gig economy in facilitating entry into entrepreneurship. Section 4.1 provides

univariate evidence on the connection between gig work and entrepreneurial activity. Section 4.2 evaluates the relationshop between prior work in the gig economy and entry into

entrepreneurship. Section 4.3 examines the characteristics of individuals working in the gig

economy who create new firms.

4.1

Univariate Evidence

We start by providing univariate evidence about the connection between gig work and entrepreneurial activity. A novel aspect of the tax administrative data is that we are able to

identify and follow over time the population of individuals receiving gig income from a wide

swath of gig firms. Additionally, we can observe when they start new firms. This allows us

to link when individuals participate in the gig economy with new firms that they create over

time.

We present univariate evidence on entrepreneurial activity and individual characteristics

for different samples of our data. This analysis is at the individual-year level and the sample

period is 2012 to 2021. U.S. P opulation includes all individuals filing taxes in the U.S. aged

25 to 65 in a particular year. Gig W orkers contains those individuals in the U.S. P opulation

who received gig income in a particular year. We also separate those founders receiving gig

income before starting a firm (Gig F ounders) from those who do not receive gig income

prior to creating a firm (N on-gig F ounders). We explore the following variables for these

samples. F ounder is an indicator variable equaling one if an individual starts a new firm in

a particular year. Adjusted Gross Income is the adjusted gross income of an individual as

reported on Form 1040 in a particular year. Low Income is an indicator variable equaling one

if an individual’s adjusted gross income is in the bottom tercile in a particular county-year.

Receives EIT C is an indicator variable equaling one if an individual receives an Earned

Income Tax Credit in a particular year. Single with Dependents is an indicator variable

equaling one if an individual’s filing status on Form 1040 in a particular year is single and the

individual has dependents attached to Form 1040. Age is an individual’s age in a particular

year.

[Insert Table 3 Here]

Table 3 shows the means for entrepreneurial activity and individual characteristics. We

find that 2.5% of gig workers create new firms, which is significantly higher than the average

in the population of 0.7%. This is an economically large difference, indicating that gig

workers are, on average, 2.6 times more likely to start new firms. Gig workers have lower

income and often are located in the bottom of the income distribution. We also find that

gig workers are more likely to be single and have dependents, suggesting that gig workers

are individuals who appear to have relatively less flexibility. Additionally, gig workers tend

to be younger. We continue by examining differences between founders who are gig workers

relative to those who are not. The average income for gig founders is about half relative to

non-gig founders, which is further reflected in 43% of gig founders having low income. A

significantly higher share of gig founders are single with dependents and gig founders are

younger than non-gig founders.

Overall, based on the the univariate evidence, gig work might support entrepreneurship

for those individuals who are capital constrained and who might require relatively more

flexibility in their work arrangements. We provide regression evidence for the relationship

between the gig economy and entry into entrepreneurship and which individuals respond in

the following sections.

4.2

Entry into Entrepreneurship

We next study the role of participating in the gig economy on entry into entrepreneurship. On

the one hand, the gig economy may ease barriers faced by an individual looking to start a new

firm. Alternatively, the gig economy could substitute for entrepreneurial activity. Previous

work has shown ride-hailing services (Barrios, Hochberg, and Yi (2022)) and monetizing an

asset for short-term rentals (Mao et al. (2023)) are associated with increases in new firm

creation at the aggregate level. Using the population of U.S. tax filers from 2012 to 2021 for

a wide range of gig firms, we compare the likelihood of starting a new firm for individuals who

previously received income from the gig economy relative to those who did not participate

in the gig economy. A distinct feature of our data is that we directly estimate the impact of

the gig economy on the propensity to enter into entrepreneurship at the individual level.

In Table 4, we estimate the following specification at the individual-year level in columns

(1) to (4):

Yi(c),t = αc×t + β · Gig Workeri,bef ore t + γ · Xi,t−1 + εi(c),t ,

(1)

where Yi(c),t measures the incidence of new firm creation by individual i located in county c

in year t. Gig W orkeri,bef ore t is an indicator variable equaling one if individual i received

gig income prior to year t. Section 2.2 describes how we identify individuals who received

gig income using a manually compiled list of gig firms and primarily Forms 1099-MISC,

1099-NEC, and 1099-K. The vector X i,t−1 includes log adjusted gross income, gender, filing

status, whether an individual has any dependents, and log age. The specification includes

county-year fixed effects (αc×t ) to absorb unobserved time-varying local differences, except

in column (1) which includes county and year fixed effects. In our strictest specification,

we include granular fixed effects for each category of our control variables. The inclusion of

granular fixed effects allows us to compare entrepreneurial entry for individuals who differ in

terms of participation in the gig economy and are the same across observable characteristics.

Standard errors are clustered at the county level. The coefficient of interest is β, which

estimates the marginal effect of participating in the gig economy on entrepreneurial entry.

[Insert Table 4 Here]

In Panel A, we measure new firm creation as F ounder, which is an indicator variable

equaling one if an individual starts a new firm in a particular year.8 Based on column (1), we

find that individuals participating in the gig economy are 1.0 percentage points more likely

to create a new firm. The effect is statistically significant and economically substantial,

representing more than a doubling in the propensity of starting a new firm relative to the

sample mean of 0.7%. We include county fixed effects to absorb time-invariant heterogeneity

across locations and year fixed effects to account for macroeconomic time trends. Using the

A potential concern is that gig workers use Schedule C filings to report gig income, which might be

interpreted as new firm creation. Accordingly, we carefully exclude all Schedule C filings used for this reason

as described in Section 3.1.

population of U.S. individuals allows us to include increasingly stricter fixed effects to assess

the robustness of our estimates. In column (2), we interact county and year for the fixed

effects to absorb time-varying local economic activity. The estimate for the effect of gig

work on entrepreneurial entry remains statistically and economically unchanged. Next, we

use individual characteristics to account for differences in the composition of individuals who

participate in the gig economy. In column (3), we find that the result is quite similar when we

include the controls. In column (4), we augment our model by saturating it with granular

fixed effects for each individual characteristic. Specifically, this model includes 100 fixed

effects for each income percentile, which is determined using an individual’s adjusted gross

income in a particular county-year. It also has 41 fixed effects for each age from 25 to 65.

Further, it includes indicator variables for gender, filing status, and having any dependents.

The results remain unchanged.

Although the stricter specification in column (4) accounts for differences between the

observed characteristics of individuals participating in the gig economy compared to those

in the rest of the economy, there could be a potential concern about omitted variables. A

common unobserved factor might be correlated both with working in the gig economy and

entry into entrepreneurship. To strengthen our evidence about the relationship between

the gig economy and entrepreneurship, we leverage plausibly exogenous variation in the

availability of gig work both geographically and over time following the approach of Jackson

(2022). In particular, gig firms rolled out their platforms sequentially across different regions

at different periods.9 This approach defines a county as treated starting in the first year when

at least 30 individuals located in the county receive gig income.

In column (5), we estimate the following specification:

Yi(c),t = αc×t + β · Gig Worker Staggeredi,bef ore t + γ · Xi,t−1 + εi(c),t ,

(2)

This approach is related to papers studying the impact of the gig economy by focusing on a particular gig

firm (Burtch, Carnahan, and Greenwood (2018), Barrios, Hochberg, and Yi (2022), and Mao et al. (2023)).

Due to confidentiality reasons, we cannot study the availability of gig work through a specific gig firm.

where Yi(c),t continues to measure the incidence of new firm creation by individual i located

in county c in year t, similar to equation (1). Gig W orker Staggeredi,bef ore t is an indicator

variable equaling one if an individual received gig income prior to year t and if there are at

least 30 individuals located in the county receiving gig income prior to year t. Note that

the inclusion of county-year fixed effects absorbs the level effect of gig work availability.

For this specification, we define X as a vector of granular fixed effects for each category

of our control variables, following column (4) in Table 4. In column (5), we show that the

estimated coefficient using the methodology of Jackson (2022) remains quite similar and that

gig workers are 1.0 percentage points more likely to start a new firm. The stability of the

estimated coefficient across the specifications suggests that it is unlikely an omitted variable

drives the result.

To further evaluate the potential influence of omitted variables, we use the approach of

Oster (2019) to formally test for this bias. For the Oster test, we focus on the strictest specification using controls in column (4) and the staggered availability of gig work in column (5).

The test calculates the potential influence of omitted variables by comparing the coefficient

and R2 for a particular specification with the same regression excluding all controls. The

adjusted coefficient (βadj ) is defined by Oster (2019) as:

βadj = βc − δ (βu − βc )

Rmax

− Rc2

Rc2 − Ru2

(3)

where βc is the coefficient from the specification that includes controls and βu is the coefficient

excluding all controls. The R2 values are also from the specifications with and without

controls. The parameter δ is the level of selection on unobservables relative to observed

controls. The parameter Rmax

is the hypothetical R2 from a regression including the observed

and unobserved controls. Following Mian and Sufi (2014), Hebert (2023), and Hu and Ma

(2024) and the guidance in Oster (2019), we set δ to 1 and Rmax

equal to min(2.2 · Rc2 , 1).

Table A1 reports the results for the Oster test. Panel A provides the inputs and findings

for the regression in column (4) and Panel B shows them for the regression in column (5).

For the strictest specification with controls, we find that the estimate of βadj is 0.749. We also

show that the estimate of βadj is 0.746 for the specification using the staggered availability of

gig work. Both of these estimates are close to the corresponding estimates in Table 4, Panel

A. The identified set is defined by Oster (2019) as the interval from βadj to βc . The interval

for both specifications is relatively narrow from [0.749, 0.981] and [0.746, 0.980] for columns

(4) and (5), respectively. Accordingly, the Oster test rejects the null hypothesis that the

estimated coefficient is zero after accounting for the potential influence of omitted variables.

The last column in each panel estimates the δ such that the βadj is zero. These estimates show

that the null hypothesis continues to be rejected when the level of selection on unobservables

relative to observed controls is up to 422.8% and 417.9% for the specifications in columns (4)

and (5), respectively. This level of explanatory power for the unoservables would be quite

high. Oster (2019) argues that 100%, which implies a δ of 1, is a reasonable value since

researchers focus on the most relevant observed controls (Angrist and Pischke (2010)). An

important caveat is that we cannot completely rule out omitted variables. The approach

in Oster (2019) allows us to test for potential bias under reasonable parameterizations and

quantify the magnitude of omitted variables needed to account for the baseline estimates.

In Panel B, we turn to separately studying the effect of the gig economy on starting a

new firm for the first time by defining F irst-time F ounder as an indicator variable equaling

one if an individual starts a new firm in a particular year and has not previously created

a firm. This allows us to understand the role of the gig economy in facilitating entry into

entrepreneurship for individuals who have not previously created a firm. Column (1) shows

that gig workers are 0.8 percentage points more likely to start a new firm, which includes

county and year fixed effects. Comparing this estimate to the corresponding model in Panel

A, we find that about 74% of the effect of the gig economy on entrepreneurial entry is

explained by individuals creating a firm for the first time. We include increasingly stricter

fixed effects or controls as in Panel A. We continue to find nearly identical estimates in

columns (2) to (4). The results remain similar in column (5), where we exploit the staggered

availability of gig work.

We provide several extensions of the baseline analysis. First, we consider alternative

measurements of an individual’s participation in the gig economy. Note that the variable of

interest, Gig W orkeri,bef ore t , in the baseline specification includes the effect of working in

the gig economy in t-1 and also the effect of working in the gig economy before t-1. In Table

A2 of the Appendix, we define Gig W orker P revious Y ear as an indicator variable equaling

one if an individual received gig income in year t-1. We also construct the analogous variable

for the staggered availability of gig work (Gig W orker Staggered P revious Y ear). Across

the same specifications as in Table 4, we find that participating in the gig economy in the

previous year increases an individual’s likelihood of starting a new firm by 0.9 percentage

points. Notably, this estimate is close to the baseline estimate of a 1.0 percentage point

increase in the propensity to start a new firm. This suggests that the effect of the gig

economy on entrepreneurship is primarily driven by recent gig work.

Next, we examine the extent of prior experience in the gig economy. We construct

Gig W orker N umber of Y ears as the number of years that an individual received gig

income prior to year t. We continue to construct the similar variable for the staggered

availability of gig work (Gig W orker Staggered N umber of Y ears). In Table A2, Panel

B shows that a one year increase in the number of years an individual received gig income

increases the propensity of starting a new firm by 0.5 percentage points. This indicates that

additional years of experience in the gig economy increase the propensity to start a new firm.

In Table A3, we also differentiate between the source of gig income as transportation,

leasing, selling, and services as defined in Section 2.2. Based on the strictest specification

using the staggered availability of gig work in column (5), we find that transportation gig

work increases the likelihood of starting a new firm by 0.8 percentage points, while leasing,

selling, and services gig work amplify the probability of creating a firm by 1.0, 1.3, and, 0.8

percentage points, respectively. These estimates suggest that the effects are somewhat larger

for selling and leasing gig work.

We continue by evaluating if the estimates differ over the sample period. This might

occur because there has been considerable growth in the gig economy from 2012 to 2021, as

shown in Figure 2. In Table A4, we split the sample from 2012 to 2016 in Panel A and from

2017 to 2021 in Panel B. The estimates are broadly similar across the subsamples, suggesting

that the relation between gig work and entrepreneurship has been stable over the sample

period.

Do financial constraints play a role in the effect of the gig economy on entrepreneurship?

We construct a proxy of fixed costs of starting a new firm across industries by measuring a

firm’s assets in its founding year. We use data on assets available for partnerships and corporations (Forms 1065, 1120, and 1120-S). Note that Schedule C does not include information

on a firm’s assets. We calculate the average assets across firms in a sector based on two-digit

North American Industry Classification System (NAICS) codes during our sample period.

We split sectors into high versus low fixed costs based on the median. This approach splits

the outcome variable of F ounder into two parts: firms started in high fixed cost industries

and those formed in low fixed cost industries. In Table A5, Panel A provides the results for

firms started in high fixed cost industries and Panel B shows the estimates for firm created

in low fixed cost industries. There is a 0.7 percentage point increase in the likelihood of

starting a new firm in a high fixed cost industry if an individual previously worked in the

gig economy across each specification, while the estimate is a 0.3 percentage point increase

for the comparable coefficient in a low fixed cost industry for all columns. Comparing the

same column in Panels A and B, we find that the coefficients are statistically significantly

different. The larger estimates in high fixed cost industries are consistent with gig work

mitigating financial constraints in entering entrepreneurship and connect our findings to the

related literature (e.g., Evans and Jovanovic (1989), Hurst and Lusardi (2004), and Robb

and Robinson (2014)).

Taken together, we provide the first micro-level evidence about the relationship between

the gig economy and new firm creation across the entire U.S. population for a large array

of gig firms. The set of results in this section suggest that individuals with prior experience

in the gig economy are substantially more likely to start new firms and new firm creation

is driven by those with no prior entrepreneurial experience. These estimates are consistent

with aggregate evidence provided by Barrios, Hochberg, and Yi (2022) and Mao et al. (2023)

for specific gig firms. They also build on prior literature about levers influencing entry into

entrepreneurship (e.g., Evans and Jovanovic (1989), Kerr and Nanda (2009), Hombert et al.

(2020), and Gottlieb, Townsend, and Xu (2022)).

4.3

Who Responds?

We next examine the characteristics of individuals in the gig economy who respond by creating new firms. Detailed information on individuals available in U.S. tax returns allows

us to explore differential responses based on specific traits. We extend our baseline specification for the staggered availability of gig work by interacting an individual’s particular

characteristic with Gig W orker Staggeredi,bef ore t as follows:

Yi(c),t = αc×t + β1 · Gig Worker Staggeredi,bef ore t · Characteristici,t−1

+ β2 · Gig Worker Staggeredi,bef ore t + β3 · Characteristici,t−1

(4)

+ β4 · Gig Availabilityc,t × Characteristici,t−1 + εi(c),t .

We focus on all newly created firms as the outcome in this section (F ounder). We continue

to define Gig W orker Staggeredi,bef ore t as an indicator variable equaling one if an individual

received gig income prior to year t and if there are at least 30 individuals located in the county

receiving gig income prior to year t. The variable Characteristici,t−1 is the characteristic of

individual i in year t-1. We include county-year fixed effects (αc×t ). The coefficient of

interest is β1 , which captures the marginal effects of having a specific characteristic and

previously participating in the gig economy on entrepreneurial entry. The specification also

includes terms for the direct effects of participation in the gig economy and the specific

individual trait, in addition to the interaction between Gig Availabilityc,t and the individual

characteristic. We also provide the results following equation (1) in Table A6 and show that

the estimates are similar.

We broadly investigate three categories of individual characteristics. First, we evaluate the differential response by individuals with lower income. Since the downside of entrepreneurship is relatively smaller for this group, they might be more willing to experiment

with a new idea (Salgado (2020)). Gig income might also mitigate liquidity constraints faced

by low-income individuals entering into entrepreneurship. Second, we explore heterogeneity

in entrepreneurial entry by age. This characteristic relates to the potential importance of

lifecycle considerations in entrepreneurial choice (Azoulay et al. (2020) and Bernstein et al.

(2022)). Third, gig workers generally can decide when to participate in the gig economy.

This flexibility might be particularly valuable for time-constrained gig workers, especially

compared to other income opportunities such as salaried employment. We explore the importance of flexibility in facilitating entrepreneurial entry for individuals who potentially

face relatively higher time constraints. By entering the gig economy, these individuals might

be more likely to experiment through entrepreneurial activities.

[Insert Table 5 Here]

Table 5 presents the results. In column (1), we start with log adjusted gross income

(Log AGI) as the characteristic and find that gig workers with lower income are more likely

to start new firms. Similarly, columns (2) and (3) show that gig workers who are in the

bottom of the income distribution or claim income-related tax credits, respectively, are more

likely to create new firms. These estimates are about 4.5% of the overall effect reported in

Table 4. We also continue to report throughout the estimates in this table that gig workers

are more likely to become entrepreneurs, which is consistent with Section 4.2.

We continue by evaluating the role of lifecycle considerations. In column (4), we find

that the propensity to become entrepreneurs is significantly higher for relatively younger

gig workers. Last, we explore the role of flexibility using two measures. In column (5),

we proxy for individuals with relatively higher time constraints as those with dependents

(Has Dependents) and show that gig workers with dependents are 0.1 percentage points

more likely to start new firms. In column (6), we focus on gig workers with dependents

and whose filing status is single to measure individuals facing particularly elevated time

constraints. The likelihood of becoming an entrepreneur for gig workers who are single with

dependents increases by 0.2 percentage points. In sum, our results highlight the role of

individual heterogeneity in the responsiveness to entrepreneurial entry. Gig workers who are

lower income, are relatively younger, and who might benefit from flexibility are substantially

more likely to create new firms.

Firms Created by Gig Workers

This section studies newly created firms linked to individuals receiving income from the gig

economy over time. Using all firms in the U.S. during the sample period, we evaluate firms

at founding and their subsequent performance. In Section 5.1, we examine the industry

composition of firms and its relation to an entrepreneur’s previous work experience. Section

5.2 turns to assessing the characteristics of firms created by gig workers at founding and their

subsequent performance. Section 5.3 explores the employment of independent contractors

and capital structure at firms started by gig workers.

5.1

Industry Composition and Previous Work Experience

This section uses data on the universe of firms in the United States described in Section 3.1

from 2012 to 2021. For each firm, we determine whether an entrepreneur participated in the

gig economy before the firm was started (“gig founder”) or if the individual did not receive

gig income prior to founding (“non-gig founder”). This allows us to precisely link individuals

deriving income from the gig economy with the firms that they create.

We begin our firm-level analyses by evaluating how the industry composition of firms

started by gig founders compares to those created by non-gig founders. We extract information on the industry classification of newly created firms from the two-digit NAICS code

reported in firms’ tax returns. We aggregate each two-digit NAICS sector in parentheses to

nine broad industries using the following classification: Arts and Media (51, 71), Finance and

Real Estate (52, 53, 55), Healthcare (62), Manufacturing (23, 31, 32, 33), Personal Services

(61, 72, 81), Professional Services (54, 56), Resource Extraction (11, 21, 22), Trade (42, 44,

45), and Transportation (48, 49). We exclude firms with no industry reported and those in

sector 92 (Public Administration). For all firms in the sample, we construct the share of

newly created firms in a particular industry for gig founders. We also separately determine

these shares for firms with non-gig founders.

[Insert Figure 5 Here]

Figure 5 provides the share of new firms started in industries split by gig and non-gig

founders. The darker red bars show the proportion of firms in a particular industry for

founders receiving gig income before creating a firm. The lighter gray bars display the share

of firms formed in a particular industry for non-gig founders. Note that we exclude Schedule

C filings that are used to report gig income as described in Section 3.1. We find that a

substantial amount of entrepreneurship is concentrated in personal and professional services

and trade across all firms. There is also large variation in industries based on gig versus nongig founders. Gig founders tend to create more firms in personal services, transportation,

and trade. Further, we show that non-gig founders start a relatively higher share of firms in

healthcare, manufacturing, and professional services.

Next, we examine the transition into entrepreneurship based on the type of gig firm

from which a gig worker received income. For this analysis, we restrict our attention to

entrepreneurs who previously participated in the gig economy and classify an entrepreneur’s

experience in the gig economy based on gig firm type. As in Section 2.2, each gig firm

is classified into four categories based on the tasks intermediated by the specific platform:

leasing, selling, services, and transportation. Then, we estimate the share of individuals

receiving gig income and starting new firms in a particular industry across the different gig

firm categories.

[Insert Table 6 Here]

Panel A of Table 6 reports the transitions from gig firm type to the industry of newly

created firms. We broadly find evidence that entrepreneurs transition from the gig economy into similar industries. For example, 67.2% of entrepreneurs with experience in selling

through the gig economy operate in the trade industry. Likewise, about half of entrepreneurs

participating in the services sector of the gig economy operate in personal and professional

services.

We continue by comparing the likelihood of creating a firm in a sector where an entrepreneur has been previously employed for gig and non-gig founders. Specifically, we

employ the following specification for our sample of entrepreneurs:

Yj(c),t = αc×t + β · Gig Worker Staggeredj,bef ore t + εj(c),t ,

(5)

where Yj(c),t is the outcome variable equaling one if entrepreneur j located in county c starts

a firm in year t in the same sector as any prior employment. Gig W orker Staggeredi,bef ore t

is an indicator variable equaling one if an individual received gig income prior to year t and

if there are at least 30 individuals located in the county receiving gig income prior to year t.

We include county-year fixed effects (αc×t ) to absorb unobserved time-varying county-level

variation. Standard errors are clustered at the county level.

Panel B of Table 6 provides the results. We separately investigate whether this is overlap

in an entrepreneur’s newly created firm with any prior experience in a sector and, specifically,

experience in the gig economy. In column (1), the outcome variable (Repeat Sector) is an

indicator variable equaling one if an entrepreneur starts a new firm in the same sector as

prior work experience based on W-2 and 1099 income. This measure of experience includes

any W-2 and 1099 income received before starting a new firm, both from gig and nongig firms. We find entrepreneurs who previously participated in the gig economy are 4.1

percentage points more likely to create a firm in a sector where they have prior experience.

The effect is statistically significant at the 1% level and economically large, representing an

8.1% increase relative to the sample mean. We also examine the role of the gig economy

by defining Repeat Sector Gig as an indicator variable equaling one if an entrepreneur’s

newly started firm is in the same sector as any prior gig experience. In column (2), we show

that entrepreneurs who previously participated in the gig economy are 1.2 percentage points

more likely to create a firm in the sector where they have prior gig experience. Additionally,

Table A7 provides the estimates for the analogous specification where the variable of interest

is Gig W orker, which we refer to as the controls approach. We show that the estimates are

statistically and economically similar.

To sum up, this section provides evidence suggesting that gig workers often start new

firms in industries that are related to the type of gig firm from which they received income,

as well as in industries related to their overall past work experience. A natural interpretation

of these results is that the gig economy, in addition to past work experience, could act as a

pathway to entrepreneurship by allowing individuals to learn about a specific industry. This

is also consistent with our results on who responds by entering into entrepreneurship after

working in the gig economy: those individuals who are lower income, relatively younger, and

value flexibility are likely potential entrepreneurs with a higher marginal benefit from onthe-job learning. By learning from prior work experiences, entrepreneurs might be able to

alleviate information frictions associated with the uncertainty of entrepreneurship, allowing

them to accumulate industry-specific knowledge and business-related acumen.

5.2

Firms at Founding and Performance

In this section, we investigate firm-level characteristics at founding and subsequent performance. We compare firm outcomes for entrepreneurs who have previously participated in

the gig economy with founders who have not received gig income using the following specification:

Yk(cs),t = αc×t + αs + β · Gig Worker Staggeredk,bef ore t + εk(cs),t ,

(6)

where Yk(cs),t is an outcome for firm k located in county c, operating in industry s, and

founded in year t. We measure outcomes at founding in the year a firm is created. We also examine performance in years one to three after a firm is started. Gig W orker Staggeredi,bef ore t

is an indicator variable equaling one if an individual received gig income prior to year t and

if there are at least 30 individuals located in the county receiving gig income prior to year t.

We include county-year fixed effects (αc×t ) to absorb unobserved time-varying county-level

variation. We use industry (αs ) fixed effects to capture time-invariant industry heterogeneity.

Standard errors are clustered at the county level. The coefficient of interest, β, estimates the

marginal effect of participating in the gig economy by an entrepreneur on a firm outcome.

[Insert Table 7 Here]

In Table 7, we first examine the size of newly created firms at founding for entrepreneurs

with and without prior experience in the gig economy using equation (6). We use two

measures of firm size. First, we use revenues reported on a firm’s Schedule C, which has

nonnegative values. Since it can be zero and is a continuous variable, we define Revenues

as the log of one plus revenues for a firm. Second, we construct Employees as a count of

the number of employees at a firm, which includes both salaried employees and independent

contractors. In column (1), we find that firms started by gig workers have revenues that

are 22.9% higher than those founded by entrepreneurs who have not participated in the gig

economy.10 In column (2), we use a Poisson model since the outcome is a count variable

(Cohn, Liu, and Wardlaw (2022)). We show that the number of employees at gig-founded

firms is 39.3% higher at firms started by gig workers. These results suggest that entrepreneurs

with prior experience in the gig economy tend to create larger firms at founding.

We provide several extensions to assess the robustness of these results. First, we include the following controls in equation (6): log adjusted gross income, gender, filing status,

an indicator for having dependents, and log age. In Table A8, we continue to find that

firms started by gig workers have higher revenue and more employees at founding. We also

examine the differential response based on the founders’ characteristics by estimating the

coefficients for subsamples based on income, flexibility, and age. In Figure A3, Panels A and

B provide the estimates for revenues and employees at founding, respectively. We generally

find that most estimates are similar to the baseline coefficients. Second, Table A9 presents

the estimates using the controls approach where the variable of interest is Gig W orker and

reports similar findings. Last, we condition the sample to include firms with at least one

employee within the first five years of founding. In Table A10, we continue to show that firms

started by gig workers have higher revenue and more employees. Combined, we broadly find

consistent evidence that gig workers form larger firms in their founding year.

Next, we evaluate firm outcomes in the years following creation. We construct the

following firm outcomes in years one to three after a firm is established. To capture the

likelihood of survival, we define Survival as an indicator variable equaling one if a firm is

observed in a particular year after founding. We also construct P rof itability as the inverse

hyperbolic sine of a firm’s gross profits as reported in a firm’s Schedule C for a particular

year after a firm is started.

[Insert Table 8 Here]

Table 8 provides the results for firm performance. Panel A shows the estimates for the

When the outcome is a natural logarithm, we report the exponentiated coefficient minus one in the text.

The tables contain the raw coefficients.

effect of participating in the gig economy on firm survival. Note that the number of observations decreases across specifications since recently created firms do not yet have survival

measures for future years. We find that firms started by gig founders are significantly less

likely to survive in the one to three years following their creation relative to firms formed by

non-gig founders. These effects are present in both the short and long run. Gig-founded firms

are 2.6 percentage points less likely to survive in the first year after founding, which rises to

3.2 percentage points three years following creation. Relative to the respective sample means,

this is a decrease of 3.7% to 7.1% in the likelihood of survival. Since firms enter our sample

in the first year that we observe them in the tax returns data, an alternative interpretation of

these findings is that the initial gig period could be considered a soft launch of the business

and, accordingly, gig-founded firms might not have a lower likelihood of survival compared to

firms started by non-gig founders when including this initial period. Panel B examines firm

profitability for years one to three after founding. Since firm performance is only available

for existing firms, it is necessarily conditional on survival. For firm profitability, we find that

firms started by gig workers have 39.4% to 46.9% higher profits.

We examine the robustness of the estimates for firm performance. In Table A11, we

include controls for income, gender, filing status, dependents, and age, and report similar

estimates. Figure A3 provides the estimates for subsamples based on income, flexibility,

and age. Panels C and D show the estimates for survival and profitability three years

after founding, respectively. We largely show that the estimates are close to the baseline

coefficients. Table A12 provides the results using the controls approach where the variable of

interest is Gig W orker and finds similar estimates. In Table A13, we reestimate the results

for firms with at least one employee within the first five years of founding and report broadly

consistent findings.

[Insert Table 9 Here]

In the last set of analyses for this section, we provide additional evidence about the

performance of gig-founded firms by examining the evolution of employment in newly created

firms. Table 9 presents the estimates. In Panel A, the outcome is an indicator variable

equaling one if a firm has any employees in a particular year (Has Employees). We find

that gig-founded firms are 0.9 to 2.1 percentage points more likely to have employees, which is

a 5.7% to 12.3% increase relative to the respective sample means. Panel B reports estimates

where the outcome is an indicator variable equaling one if a firm has at least five employees

in a particular year after founding (At Least F ive Employees). We show that gig-founded

firms tend to have a significantly higher number of employees. In particular, firms started by

gig workers are 0.5 to 1.2 percentage points more likely to employ at least five individuals,

which is a 7.6% to 16.1% rise compared to the respective sample means. We also provide

the estimates using the controls approach where the variable of interest is Gig W orker in

Table A14 and show that the estimates are similar. These findings indicate that gig-founded

firms are more likely to have employees and operate firms with a relatively high number of

employees in the years following their founding.

Paired with the results in Section 5.1, this section provides novel evidence on the characteristics of firms started by gig workers. When they start new firms, gig workers appear

to bear more risk by creating larger new ventures that are less likely to survive. However,

surviving firms realize higher performance and grow larger over time. Prior literature highlights that experimentation plays a key role in entrepreneurship (Manso (2011) and Kerr,

Nanda, and Rhodes-Kropf (2014)). If the gig economy lowers the costs of entry into entrepreneurship, it could allow founders to experiment and take greater risks. Then, newly

created firms would exhibit more extreme outcomes, including higher exit rates and better

performance at surviving firms. An interpretation of our findings is that firms started by gig

founders display greater experimentation and increased risk taking. Shorter average survival

is also consistent with gig founders learning about the prospects of their firm more quickly

and shutting down sooner.

5.3

Independent Contractors and Capital Structure

We conclude our firm-level analyses by offering additional evidence on the employment decisions and capital structure of firms started by gig workers. The administrative data on

U.S. tax returns allow us to separately track salaried workers, based on receiving a W-2,

and independent contractors, based on information returns provided by firms to the IRS.

For every firm in our sample, we construct an extensive margin measure based on whether

a firm employs independent contractors (Has Contractors). We also define an intensive

margin proxy using the number of independent contractors employed in a particular year

(N umber of Contractors). Similar to our previous analyses, we track employment of independent contractors for one to three years after founding.

[Insert Table 10 Here]

Table 10 presents the results. Panel A shows the extensive margin. We find that

gig-founded firms are significantly more likely to employ independent contractors relative

to firms with a non-gig founder in the one to three years after founding. The increases

are statistically significant at the 1% level and economically substantial. The likelihood of

hiring an independent contractor is 9.5% to 18.2% higher at gig-founded firms relative to

ones with a non-gig founder compared to the respective sample means. Panel B provides

the intensive margin, which is estimated using a Poisson model because the outcome is a

count variable. There is an economically large increase of 25.3% to 44.7% in the number of

independent contractors hired by gig-founded firms. Table A15 presents the results using

the controls approach where the variable of interest is Gig W orker and reports broadly

consistent estimates.

In the final set of firm-level analyses, we evaluate the capital structure of firms started by

gig workers. Participation in the gig economy might allow entrepreneurs to earn additional

income, subsequently increasing liquidity and the likelihood of accessing external credit.

Additionally, the gig economy often relies on the provision of an individual’s physical capital

(Buchak (2024)), which may improve access to external financing for gig founders who can

use this capital as collateral. We use information on a firm’s interest expense as reported

in Schedule C to measure whether a firm has debt in a window after founding. We define

Has Debt as an indicator variable if a firm has debt by a particular year after founding

based on reporting interest expense.

[Insert Table 11 Here]

Table 11 provides the results on capital structure. We find that there is a sizable increase

in access to external credit for gig founders relative to firms started by entrepreneurs who

have not received gig income. In the first year after a firm is started, a gig founder is 1.2

percentage points more likely to use debt. The estimate rises to 3.4 percentage points by

year three. In economic terms, there is a 10.2% to 17.9% increase in the probability of a

gig-founded firm having debt one to three years after creating a new firm relative to the

respective sample means. We also present the estimates in Table A16 using the controls

approach where the variable of interest is Gig W orker and show that the estimates are

similar.

In sum, this section presents results suggesting that firms started by gig workers employ more independent contractors. They are consistent with evidence in Sections 5.1 and

5.2 about the gig economy potentially supporting learning and experimentation by entrepreneurs. To the extent that experience in the gig economy acts as a vehicle for learning,

gig founders might rely relatively more on independent contractors to accelerate firm growth

and more flexibly respond to unexpected shocks. Additionally, if the gig economy encourages

experimentation by entrepreneurs, gig-founded firms could use more independent contractors

to explore the prospects of new entrepreneurial ventures. The findings also indicate that gig

founded firms have more debt, suggesting that the gig economy might enable gig workers to

access capital markets.

Are Gig Founders Better Off ?

A longstanding puzzle in the entrepreneurship literature is why individuals start new firms if

they appear to receive less income and bear more risk (Hamilton (2000) and Moskowitz and

Vissing-Jørgensen (2002)). An explanation is the option to return to a job may encourage

individuals to experiment with entrepreneurship (Manso (2016) and Catherine (2022)). Motivated by this literature, this section evaluates the income of individuals participating in the

gig economy who enter into entrepreneurship. Based on the results in Section 5, the effect on

entrepreneurs’ income is not necessarily clear as survivorship is lower for gig-founded firms,

while performance is generally higher.

We compare changes in income for entrepreneurs who participate in the gig economy

prior to starting a firm relative to those who have not received gig income using the following

specification:

Yj(cs),t = αc + αs + αt + β · Gig Workerj,bef ore t + εj(cs),t ,

(7)

where Yj(cs),t is an outcome that measures income changes for entrepreneur j whose firm is

located in county c, operating in industry s, and founded in year t. Gig W orkerj,bef ore t

is an indicator variable equaling one if entrepreneur j received gig income prior to year t.

We continue to include county (αc ) and industry (αs ) fixed effects to capture time-invariant

heterogeneity in the local economy and industry, respectively. We use founding-year fixed

effects to absorb time trends in firm creation (αt ). Standard errors are clustered at the

county level.

Using the administrative data on U.S. tax returns, we track income dynamics using

adjusted gross income for all entrepreneurs over time, which we can observe regardless of

a firm’s survival. We construct two measures of an entrepreneur’s income. First, we define Change in Income as the difference in the log of an entrepreneur’s adjusted gross

income in a particular year relative to the firm’s founding year. This variable provides

an estimate of the growth in a founder’s income since starting a firm. Second, we form

Increase in Income P ercentile as an indicator variable equaling one if an entrepreneur’s

income percentile in a particular year increases relative to the firm’s founding year. Income

percentiles are based on adjusted gross income within a county-year.

[Insert Table 12 Here]

Table 12 presents the results. In Panel A, we evaluate the change in an entrepreneur’s

income relative to founding. Column (1) shows that gig founders earn 3.2% higher adjusted

gross income in the year after they start a firm relative to entrepreneurs who have not participated in the gig economy. The wedge between gig and non-gig founders is persistent and

gradually grows over time, increasing to 9.5% and 13.1% two and three years after founding,

respectively (columns (2) and (3)). In Panel B, we report the estimates for the change in an

entrepreneur’s income percentile relative to a firm’s founding year. This approach allows us

to compare the ranking of an entrepreneur in the income distribution in a specific geography

and at a particular time. We show that gig founders are more likely to rise in the income

distribution relative to non-gig founders. Column (1) shows that there is a 1.1 percentage

point increase in the likelihood of rising in the income distribution one year after founding,

which grows to 1.9 and 2.0 percentage points two and three years after founding, respectively

(columns (2) and (3)). Economically, this represents a 2.1% to 3.2% increase compared to

the respective sample means. Table A17 provides the estimates using the controls approach

where the variable of interest is Gig W orker, and shows that the results are statistically and

economically similar.

Overall, our findings suggest that gig workers who start new firms are better off in

terms of their income. This is also consistent with the gig economy benefiting workers

through higher future income from opportunities beyond the gig job itself. An important

caveat for these results is that we compare gig founders to entrepreneurs, which might not

be an appropriate reference group for a gig worker’s income trajectory. However, this group

likely has a higher income trajectory relative to many gig workers.

Conclusion

The gig economy has grown considerably in the U.S. economy over the past decade. We use

detailed administrative data on U.S. tax returns for the universe of firms and individuals from

2012 to 2021 to study the effect of the gig economy on entrepreneurial entry. We find that

individuals who previously received income from the gig economy are significantly more likely

to start new firms. We also show that first-time entrepreneurs account for three-quarters of

this effect. We further investigate the role of capital constraints, lifecycle considerations, and

flexibility in spurring entrepreneurial entry by exploring heterogeneity in terms of individual

characteristics. We find that the effect is amplified for individuals with lower income, who

are relatively younger, and who might benefit from flexibility.

We track the universe of firms created in the United States during our sample period

linked to individuals participating in the gig economy. This allows us to study firms at

founding and evaluate their subsequent performance. We find that gig workers generally

start firms in industries similar to the gig firms from which they received income. Entrepreneurs who had participated in the gig economy create larger new ventures at founding.

Following these firms over time, we show that gig-founded firms are less likely to survive.

However, surviving firms realize higher performance and grow larger relative to firms started

by non-gig founders. Overall, an interpretation of these results is that experience in the gig

economy allows individuals to learn about entrepreneurship and experiment through their

newly created firms.

Labor market disruptions can play a role in the profile of entrepreneurial endeavors. As

the gig economy grows, much attention has been paid to its benefits and costs. We provide

evidence that the gig economy can provide a pathway to entrepreneurship and show that gig

workers appear to be better off. Future research can expand our understanding of transitions

between the gig economy and other labor markets. There are also open questions about how

the gig economy might reduce or amplify shocks faced by individuals and firms.

References

Abraham, Katharine G., John C. Haltiwanger, Kristin Sandusky, and James R. Spletzer,

2021, Measuring the Gig Economy: Current Knowledge and Open Issues, Measuring and

Accounting for Innovation in the 21st Century, Chicago: University of Chicago Press,

257–298.

Andrews, R.J., Catherine Fazio, Jorge Guzman, Scott Stern, and Yupeng Liu, 2022, The

Startup Cartography Project: Measuring and Mapping Entrepreneurial Ecosystems, Research Policy, 51(9), 104581.

Angrist, Joshua D and Jörn-Steffen Pischke, 2010, The Credibility Revolution in Empirical

Economics: How Better Research Design Is Taking the Con out of Econometrics, Journal

of Economic Perspectives, 24(2), 3–30.

Azoulay, Pierre, Benjamin F. Jones, J. Daniel Kim, and Javier Miranda, 2020, Age and

High-Growth Entrepreneurship, American Economic Review: Insights, 2(1), 65–82.

Babina, Tania, 2020, Destructive Creation at Work: How Financial Distress Spurs Entrepreneurship, Review of Financial Studies, 33(9), 4061–4101.

Barrios, John M., Yael V. Hochberg, and Hanyi Yi, 2022, Launching with a Parachute: The

Gig Economy and New Business Formation, Journal of Financial Economics, 144(1),

22–43.

Barrios, John M., Yael V. Hochberg, and Hanyi Yi, 2023, The Cost of Convenience: Ridehailing and Traffic Fatalities, Journal of Operations Management, 69(5), 823–855.

Bellon, Aymeric, J. Anthony Cookson, Erik P. Gilje, and Rawley Z. Heimer, 2021, Personal

Wealth, Self-Employment, and Business Ownership, Review of Financial Studies, 34(8),

3935–3975.

Bernstein, Shai, Emanuele Colonnelli, Davide Malacrino, and Tim McQuade, 2022, Who

Creates New Firms When Local Opportunities Arise?, Journal of Financial Economics,

143(1), 107–130.

Bianchi, Milo and Matteo Bobba, 2013, Liquidity, Risk, and Occupational Choices, Review

of Economic Studies, 80(2), 491–511.

Buchak, Greg, 2024, Financing the Gig Economy, Journal of Finance, 79(1), 219–256.

Burtch, Gordon, Seth Carnahan, and Brad N. Greenwood, 2018, Can You Gig It? An Empirical Examination of the Gig Economy and Entrepreneurial Activity, Management Science,

64(12), 5497–5520.

Cagetti, Marco and Mariacristina De Nardi, 2006, Entrepreneurship, Frictions, and Wealth,

Journal of Political Economy, 114(5), 835–870.

Catherine, Sylvain, 2022, Keeping Otions Open: What Motivates Entrepreneurs?, Journal

of Financial Economics, 144(1), 1–21.

Cespedes, Jacelly, Xing Huang, and Carlos Parra, 2023, More Money, More Options? The

Effect of Cash Windfalls on Entrepreneurial Activities in Small Businesses, Working

Paper.

Chen, M. Keith, Peter E. Rossi, Judith A. Chevalier, and Emily Oehlsen, 2019, The Value

of Flexible Work: Evidence from Uber Drivers, Journal of Political Economy, 127(6),

2735–2794.

Cohn, Jonathan B., Zack Liu, and Malcolm I. Wardlaw, 2022, Count (and Count-Like) Data

in Finance, Journal of Financial Economics, 146(2), 529–551.

Collins, Brett, Andrew Garin, Emilie Jackson, Dmitri Koustas, and Mark Payne, 2019, Is Gig

Work Replacing Traditional Employment? Evidence from Two Decades of Tax Returns,

Working Paper.

Denes, Matthew, Sabrina T. Howell, Filippo Mezzanotti, Xinxin Wang, and Ting Xu, 2023,

Investor Tax Credits and Entrepreneurship: Evidence from U.S. States, Journal of Finance, 78(5), 2621–2671.

Evans, David S. and Boyan Jovanovic, 1989, An Estimated Model of Entrepreneurial Choice

under Liquidity Constraints, Journal of Political Economy, 97(4), 808–827.

Ewens, Michael, Ramana Nanda, and Matthew Rhodes-Kropf, 2018, Cost of Experimentation and the Evolution of Venture Capital, Journal of Financial Economics, 128(3),

422–442.

Fos, Vyacheslav, Naser Hamdi, Ankit Kalda, and Jordan Nickerson, 2024, Gig-Labor: Trading

Safety Nets for Steering Wheels, Journal of Financial Economics, Forthcoming.

Gottlieb, Joshua D., Richard R. Townsend, and Ting Xu, 2022, Does Career Risk Deter

Potential Entrepreneurs?, Review of Financial Studies, 35(9), 3973–4015.

Hacamo, Isaac and Kristof Kleiner, 2022, Forced Entrepreneurs, Journal of Finance, 77(1),

49–83.

Hall, Jonathan V. and Alan B. Krueger, 2018, An Analysis of the Labor Market for Uber’s

Driver-Partners in the United States, ILR Review, 71(3), 705–732.

Haltiwanger, John, Ron S. Jarmin, and Javier Miranda, 2013, Who Creates Jobs? Small

versus Large versus Young, Review of Economics and Statistics, 95(2), 347–361.

Hamilton, Barton H., 2000, Does Entrepreneurship Pay? An Empirical Analysis of the Returns to Self-Employment, Journal of Political Economy, 108(3), 604–631.

Hebert, Camille, 2023, Gender Stereotypes and Entrepreneur Financing, Working Paper.

Herkenhoff, Kyle, Gordon M. Phillips, and Ethan Cohen-Cole, 2021, The Impact of Consumer Credit Access on Self-employment and Entrepreneurship, Journal of Financial

Economics, 141(1), 345–371.

Hombert, Johan, Antoinette Schoar, David Sraer, and David Thesmar, 2020, Can Unemployment Insurance Spur Entrepreneurial Activity? Evidence from France, Journal of

Finance, 75(3), 1247–1285.

Hu, Allen and Song Ma, 2024, Persuading Investors: A Video-Based Study, Journal of Finance, Forthcoming.

Hurst, Erik and Annamaria Lusardi, 2004, Liquidity Constraints, Household Wealth, and

Entrepreneurship, Journal of Political Economy, 112(2), 319–347.

Jackson, Emilie, 2022, Availability of the Gig Economy and Long Run Labor Supply Effects

for the Unemployed, Working Paper.

Kerr, William R. and Ramana Nanda, 2009, Democratizing Entry: Banking Deregulations,

Financing Constraints, and Entrepreneurship, Journal of Financial Economics, 94(1),

124–149.

Kerr, William R., Ramana Nanda, and Matthew Rhodes-Kropf, 2014, Entrepreneurship as

Experimentation, Journal of Economic Perspectives, 28(3), 25–48.

Klapper, Leora, Luc Laeven, and Raghuram Rajan, 2006, Entry Regulation as a Barrier to

Entrepreneurship, Journal of Financial Economics, 82(3), 591–629.

Lim, Katherine, Alicia Miller, Max Risch, and Eleanor Wilking, 2019, Independent Contractors in the US: New Trends from 15 Years of Administrative Tax Data, Working

Paper.

Manso, Gustavo, 2011, Motivating Innovation, Journal of Finance, 66(5), 1823–1860.

Manso, Gustavo, 2016, Experimentation and the Returns to Entrepreneurship, Review of

Financial Studies, 29(9), 2319–2340.

Mao, Yifei, Xuan Tian, Jiajie Xu, and Kailei Ye, 2023, Sharing Economy and Entrepreneurship, Working Paper.

Mian, Atif and Amir Sufi, 2014, What Explains the 2007–2009 Drop in Employment?, Econometrica, 82(6), 2197–2223.

Moskowitz, Tobias J. and Annette Vissing-Jørgensen, 2002, The Returns to Entrepreneurial

Investment: A Private Equity Premium Puzzle?, American Economic Review, 92(4), 745–

778.

Oster, Emily, 2019, Unobservable Selection and Coefficient Stability: Theory and Evidence,

Journal of Business & Economic Statistics, 37(2), 187–204.

Ravenelle, Alexandrea J., 2019, Hustle and Gig: Struggling and Surviving in the Sharing

Economy, University of California Press.

Robb, Alicia M. and David T. Robinson, 2014, The Capital Structure Decisions of New

Firms, Review of Financial Studies, 27(1), 153–179.

Salgado, Sergio, 2020, Technical Change and Entrepreneurship, Working Paper.

Figure 1: Gig Economy in the United States

This figure provides a map of the gig economy in the United States from 2012 to 2021. For each state, we construct the

number of individuals who participated in the gig economy during the sample period relative to the state’s labor force in

2021. Then, we determine the quartile ranking across states. Darker blue shading represents a larger share of individuals

participated in the gig economy for a particular state. Section 2.2 describes the data on the gig economy using federal tax

returns.

Percent

(8.9,19.6]

(5.4,8.9]

(3.5,5.4]

[1.5,3.5]

Figure 2: Gig Economy from 2012 to 2021

This figure plots U.S. participation in the gig economy from 2012 to 2021. Panel A shows the

cumulative number of individuals who have worked in the gig economy by a particular year.

Panel B provides the total amount of income in billions of dollars received by gig workers in

a particular year, which is converted to dollars in 2012.

Number of Gig Workers (millions)

Panel A: Cumulative Number of Gig Workers

2012

2014

2016

Year

2018

2020

Gig Income ($ billions)

100

120

Panel B: Gig Income ($ 2012)

2012

2014

2016

Year

2018

2020

Figure 3: Entrepreneurship in the United States

This figure provides a map of new firms created in the United States from 2012 to 2021. We determine the number of new

firms created in a state for a particular year relative to the total number of new firms created in the U.S. in a particular year.

For each state, we average the share of new firms created in the state across years. Then, we determine the quartile ranking

across states. Darker blue shading indicates a larger share of firms created in a particular state. Section 3.1 describes how

new firm creation is measured using federal tax returns.

Average %

(2.5,10.2]

(1.3,2.5]

(0.4,1.3]

[0.2,0.4]

Figure 4: Relationship between the Gig Economy and Entrepreneurship

This figure shows the relationship between the gig economy and new firm creation in the United States from 2012 to 2021.

For each state, we determine the correlation between the yearly count of new firms created and the number of individuals

participating in the gig economy during the previous year. Then, we determine the ranking across states. Darker blue

shading indicates a higher positive correlation between participation in the gig economy and new firm creation in a particular

state. Section 2.2 describes the data on the gig economy and Section 3.1 explains how new firm creation is measured using

federal tax returns.

Correlation

(0.92,0.99]

(0.75,0.92]

(0.00,0.75]

(-0.60,0.00]

[-0.94,-0.60]

Figure 5: Industry Composition of New Firms

tra

rin

sio

tio

vic

ice

Share of Newly-Created Firms in Industry

This figure provides the industry composition of new firms created from 2012 to 2021. The industries are based on groupings

of two-digit NAICS codes as defined in Section 5.1. The darker red bars show the share of firms created in a particular

industry by founders receiving gig income before starting a firm. The lighter gray bars display the share of firms started in

a particular industry by founders who did not receive gig income before starting a firm. The shares sum to one for firms

started by gig founders and also sum to one for firms created by non-gig founders. Section 2.2 describes the data on the

gig economy and Section 3.1 explains how new firm creation is measured using federal tax returns.

Gig Founder

Non-Gig Founder

Table 1: Gig Economy in the United States

The table provides summary statistics on the gig economy in the United States from 2012

to 2021. Panel A tabulates the number of individuals participating in the gig economy

each year. Gig income is the total income received by a worker from gig firms defined

in Section 2.2. All measures of gig income in this panel are converted to dollars in

2012. Panel B shows the characteristics of individuals in the first year they received

gig income. Adjusted Gross Income is a gig worker’s adjusted gross income converted

to dollars in 2012. W -2 Income is a gig worker’s W-2 income converted to dollars in

2012. Receives EIT C is an indicator variable equaling one if a gig worker received an

Earned Income Tax Credit. Age is a gig worker’s age. Single is a gig worker’s filing

status. Has Dependents is an indicator variable equaling one if a gig worker has any

dependents based on Form 1040. F emale is an indicator variable equaling one if a gig

worker is female. Gig income, adjusted gross income, and W-2 income are rounded due

to confidentiality reasons. Appendix A provides additional details on variable definitions.

Panel A: Gig Work from 2012 to 2021

Year

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

Number of Mean Gig Income

Gig Workers

(in $2012)

37,572

148,348

450,646

1,155,501

2,125,347

1,488,755

1,401,706

1,962,557

3,111,025

4,959,749

Gig Income

> $10,000

Gig Income

> $20,000

Standard Deviation

of Gig Income

0.218

0.280

0.237

0.178

0.199

0.303

0.369

0.410

0.251

0.255

0.124

0.164

0.139

0.107

0.116

0.236

0.305

0.333

0.176

0.181

22,000

23,000

21,000

20,000

19,000

26,000

31,000

30,000

10,000

12,000

10,000

8,000

14,000

18,000

20,000

14,000

Panel B: Characteristics of Gig Workers

AGI (in $2012)

W2 Income (in $2012)

Receives EITC

Age

Single

Has Dependents

Female

Number of

Gig Workers

Mean

Median

Standard

Deviation

9,840,231

38,000

28,000

0.440

39.351

0.682

0.443

0.348

24,000

18,000

0.000

1.000

0.000

852,000

129,000

0.496

10.532

0.465

0.496

0.476

Table 2: Summary Statistics

This table provides summary statistics for variables used in our analyses. Panel A shows

variables for individual analyses, which are conducted at the individual-year level. Panel

B includes variables for the cross-section of entrepreneurs. Panel C has firm outcomes.

The sample period is 2012 to 2021. All variables are defined in Appendix A.

Panel A: Individual-Year Variables

Variable

Number of

Observations

Mean

Median

Standard

Deviation

Gig Worker

Gig Worker Staggered

Founder

First-time Founder

Log AGI

Low Income

Receives EITC

Has Dependents

Single with Dependents

Log Age

1,265,172,170

0.014

0.007

0.005

10.758

0.325

0.197

0.505

0.143

3.750

0.000

10.968

0.000

1.000

0.000

3.784

0.119

0.118

0.081

0.073

1.647

0.469

0.398

0.400

0.350

0.273

Panel B: Entrepreneur Variables

Variable

Number of

Observations

Mean

Median

Standard

Deviation

9,805,806

7,988,688

6,681,572

5,575,492

7,988,688

6,681,572

5,575,492

0.035

0.075

0.153

0.211

0.550

0.604

0.635

0.000

0.095

0.168

0.223

1.000

0.183

1.942

2.120

2.208

0.497

0.489

0.481

Gig Worker

Gig Worker Staggered

Change in Income in One Year

Change in Income in Two Years

Change in Income in Three Years

Increase in Income Percentile in One Year

Increase in Income Percentile in Two Years

Increase in Income Percentile in Three Years

Table 2 (continued)

Panel C: Firm Variables

Variable

Gig Worker

Gig Worker Staggered

Repeat Sector

Repeat Sector Gig

Log Revenue at Founding

Number of Employees at Founding

Survival One Year After Founding

Survival Two Years After Founding

Survival Three Years After Founding

Profitability in One Year

Profitability in Two Years

Profitability in Three Years

Log Revenue in One Year

Log Revenue in Two Years

Log Revenue in Three Years

Has Employees in One Year

Has Employees in Two Years

Has Employees in Three Years

At Least Five Employees in One Year

At Least Five Employees in Two Years

At Least Five Employees in Three Years

Has Contractors in One Year

Has Contractors in Two Years

Has Contractors in Three Years

Number of Contractors in One Year

Number of Contractors in Two Years

Number of Contractors in Three Years

Debt in One Year

Debt in Two Years

Debt in Three Years

Number of

Observations

Mean

Median

Standard

Deviation

9,910,508

8,694,929

7,487,511

6,387,234

5,458,610

3,625,469

2,557,390

5,458,610

3,625,469

2,557,390

5,458,610

3,625,469

2,557,390

5,458,610

3,625,469

2,557,390

5,458,610

3,625,469

2,557,390

5,458,610

3,625,469

2,557,390

5,458,610

3,625,469

2,557,390

0.035

0.504

0.000

8.006

2.636

0.694

0.550

0.452

8.294

8.517

8.700

8.854

9.105

0.154

0.164

0.167

0.069

0.073

0.074

0.094

0.099

0.589

0.584

0.591

0.121

0.163

0.190

0.000

1.000

0.000

8.700

0.000

1.000

0.000

9.999

10.204

10.309

9.547

9.741

9.798

0.000

0.184

0.183

0.500

0.021

3.297

81.475

0.461

0.498

4.977

4.854

4.745

3.241

3.266

3.249

0.361

0.370

0.373

0.253

0.260

0.262

0.292

0.298

0.299

9.092

7.149

13.951

0.327

0.369

0.392

Table 3: Entrepreneurship and Gig Work: Univariate Evidence

This table provides univariate evidence on entrepreneurship and gig work. The sample is

at the individual-year level from 2012 to 2021. U.S. P opulation includes all individuals

in the U.S. aged 25 to 65 in the year of filing taxes. Gig W orkers are individuals in

the U.S. P opulation who received gig income in a particular year. Gig F ounders are

individuals in the U.S. P opulation who have received gig income prior to starting a new

firm. Non-gig founders are individuals in the U.S. P opulation who have not received gig

income prior to starting a new firm. F ounder is an indicator variable equaling one if

an individual starts any new firms in a particular year. Adjusted Gross Income is the

adjusted gross income of an individual as reported on Form 1040 in a particular year and

rounded due to confidentiality reasons. Low Income is an indicator variable equaling one

if an individual’s adjusted gross income is in the bottom tercile in a particular countyyear. Receives EIT C is an indicator variable equaling one if an individual received any

Earned Income Tax Credit in a particular year. Single with Dependents is an indicator

variable equaling one if an individual’s filing status on Form 1040 in a particular year

is single and the individual has any dependents based on Form 1040. Age is the age an

individual turns in a particular year. Appendix A provides additional details on variable

definitions. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively, for t-tests

of differences between the U.S. P opulation who are not gig workers and Gig W orkers,

and N on-gig F ounders and Gig F ounders.

Sample

Founder

Adjusted Gross Income

Low Income

Receives EITC

Single with Dependents

Age

Total Observations

U.S.

Population

Gig

Workers

Non-gig

Founders

Gig

Founders

0.007

96,000

0.325

0.197

0.143

0.025***

45,000***

0.571***

0.393***

0.211***

39***

1.000

104,000

0.316

0.214

0.129

1.000

59,000***

0.430***

0.313***

0.139***

38***

1,265,172,170

10,406,449

8,167,949

262,739

Table 4: Entry into Entrepreneurship and the Gig Economy

This table studies the effect of the gig economy on entry into entrepreneurship. Panel A examines all newly created firms

and Panel B focuses on first-time firm creation. F ounder is an indicator variable equaling one if an individual starts any

new firms in a particular year. F irst-time F ounder is an indicator variable equaling one if an individual starts any new

firms in a particular year and has not previously created a firm. Gig W orker is an indicator variable equaling one if an

individual received gig income prior to year t. Gig W orker Staggered is an indicator variable equaling one if an individual

received gig income prior to year t and if there are at least 30 individuals located in the county receiving gig income prior to

year t. The controls are log adjusted gross income, gender, filing status, having any dependents, and log age. The granular

fixed effects are indicators for each category of the controls. The sample includes all U.S. tax filers from 2012 to 2021 aged

25 to 65. The unit of observation is an individual-year. Appendix A provides additional details on variable definitions.

For ease of interpretation, the coefficients and standard errors are multiplied by 100. Standard errors are reported in

parentheses and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Panel A: All Newly Created Firms

Founder

Gig Worker

(1)

(2)

(3)

(4)

1.034***

(0.036)

1.025***

(0.034)

0.984***

(0.033)

0.981***

(0.033)

Gig Worker Staggered

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

0.980***

(0.033)

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Table 4 (continued)

Panel B: First-Time Entrepreneurship

First-time Founder

Gig Worker

(1)

(2)

(3)

(4)

0.760***

(0.024)

0.752***

(0.022)

0.713***

(0.021)

0.712***

(0.021)

Gig Worker Staggered

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

0.712***

(0.022)

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Table 5: Who Responds in the Gig Economy?

This table evaluates the role of characteristics in the effect of the gig economy on entry into entrepreneurship. F ounder is

an indicator variable equaling one if an individual starts any new firms in a particular year. Gig W orker Staggered is an

indicator variable equaling one if an individual received gig income prior to year t and if there are at least 30 individuals

located in the county receiving gig income prior to year t. Gig Availability is an indicator equaling one starting in the

first year when at least 30 individuals located in the county c receive gig income. Log AGI is log adjusted gross income.

Low Income is an indicator variable equaling one if an individual’s adjusted gross income is in the bottom tercile in a

particular county-year. Receives EIT C is an indicator variable equaling one if an individual received any Earned Income

Tax Credit in a particular year. Log Age is the log of an individual’s age in a particular year. Has Dependents is an

indicator variable equaling one if a gig worker has any dependents based on Form 1040. Single with Dependents is an

indicator variable equaling one if an individual’s filing status in a particular year is single and the individual has any

dependents based on Form 1040. Appendix A provides additional details on variable definitions. For ease of interpretation,

the coefficients and standard errors are multiplied by 100. All models include county-year fixed effects. Standard errors

are reported in parentheses and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%,

respectively.

Founder

(1)

(2)

(3)

(4)

(5)

(6)

Gig Availability

× Characteristic

-0.044***

(0.004)

1.463***

(0.072)

-0.002***

(0.000)

-0.013***

(0.001)

0.045***

(0.016)

1.005***

(0.032)

-0.076***

(0.001)

0.054***

(0.006)

0.049*

(0.027)

0.995***

(0.028)

0.034***

(0.002)

0.057***

(0.008)

-0.279***

(0.033)

1.977***

(0.149)

-0.591***

(0.004)

-0.077***

(0.013)

0.066***

(0.018)

0.993***

(0.028)

0.199***

(0.002)

-0.015***

(0.010)

0.219***

(0.028)

0.986***

(0.030)

-0.169***

(0.002)

0.095***

(0.011)

Characteristic

Log AGI

Low Income

Log Age

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Receives

EITC

Yes

0.001

1,265,172,170

Has

Dependents

Yes

0.001

1,265,172,170

Single with

Dependents

Yes

0.001

1,265,172,170

Gig Worker Staggered

× Characteristic

Gig Worker Staggered

Characteristic

County × Year FE

Observations

Yes

0.001

1,265,172,170

Table 6: Previous Work Experience and the Gig Economy

This table examines previous work experience of entrepreneurs and its relation to the gig economy. Panel A shows the

transition from type of gig firm to the industry of newly created firms. Additional details are provided in Section 5.1. Panel

B evaluates the relationship between prior work experience and new firm creation. The sample for this panel is individuals

creating new firms from 2012 to 2021. Repeat Sector is an indicator variable equaling one if an entrepreneur’s newly started

firm is in the same sector as any prior W-2 or 1099 experience. Repeat Sector Gig is an indicator variable equaling one if an

entrepreneur’s newly started firm is in the same sector as any prior gig experience. Gig W orker Staggered is an indicator

variable equaling one if an individual received gig income prior to year t and if there are at least 30 individuals located in

the county receiving gig income prior to year t. Appendix A provides additional details on variable definitions. All models

include county × year fixed effects. For ease of interpretation, the coefficients and standard errors are multiplied by 100.

Standard errors are reported in parentheses and clustered at the county level. *** , ** , and * denote significance at 1%, 5%,

and 10%, respectively.

Panel A: Transitions from Gig Work to Newly Created Firms

Leasing

Selling

Services

Transportation

Arts &

Media

Finance &

Real Estate

Healthcare

12.8

11.2

8.1

6.5

16.3

2.5

5.3

10.0

9.8

2.2

12.1

7.1

Personal

Manufacturing Services

3.7

2.6

8.7

6.5

17.3

7.0

27.1

22.6

Professional Resource

Services

Extraction Trade

22.8

6.8

22.4

17.6

Panel B: Entrepreneurship and Previous Work Experience

Repeat Sector

Repeat Sector Gig

(1)

(2)

Gig Worker Staggered

4.070***

(0.278)

1.195***

(0.038)

County × Year FE

Observations

Yes

0.021

9,910,324

Yes

0.013

9,910,324

0.3

0.2

16.4

67.2

14.8

23.0

Transportation

0.7

0.4

1.2

6.6

Table 7: Firms at Founding

This table studies the role of the gig economy on firms at founding. The sample includes

all firms created by individuals in the year of founding. Gig W orker Staggered is an

indicator variable equaling one if an individual received gig income prior to year t and if

there are at least 30 individuals located in the county receiving gig income prior to year t.

Revenues is the log of one plus revenues for a firm. Employees is a count of the number

of employees at a firm. Appendix A provides additional details on variable definitions.

All models include county × year and industry fixed effects. Industries are defined at

the four-digit NAICS code level. For ease of interpretation, the coefficients and standard

errors are multiplied by 100. Standard errors are reported in parentheses and clustered

at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Revenues

Employees

(1)

(2)

Gig Worker Staggered

20.583***

(1.449)

33.176***

(5.189)

County × Year FE

Industry FE

Observations

Yes

0.049

9,910,312

Yes

0.088

9,899,970

Table 8: Firm Performance

This table evaluates the performance of firms started by entrepreneurs who have received

gig income prior to founding. Panel A examines firm survival and Panel B studies firm

profitability. The sample tracks all firms created by individuals in the three years following founding. Gig W orker Staggered is an indicator variable equaling one if an

individual received gig income prior to year t and if there are at least 30 individuals

located in the county receiving gig income prior to year t. Survival is an indicator variable equaling one if a firm files taxes in a particular year after founding. P rof itability

is the inverse hyperbolic sine of a firm’s gross profits in a particular year. Appendix A

provides additional details on variable definitions. All models include county × year and

industry fixed effects. Industries are defined at the four-digit NAICS code level. For ease

of interpretation, the coefficients and standard errors are multiplied by 100. Standard

errors are reported in parentheses and clustered at the county level. *** , ** , and * denote

significance at 1%, 5%, and 10%, respectively.

Panel A: Firm Survival

Survival After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

-2.565***

(0.134)

-2.942***

(0.178)

-3.201***

(0.200)

County × Year FE

Industry FE

Observations

Yes

0.015

8,694,751

Yes

0.016

7,487,349

Yes

0.016

6,387,091

Panel B: Firm Profitability

Profitability After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

33.225***

(2.159)

38.428***

(2.419)

35.143***

(2.705)

County × Year FE

Industry FE

Observations

Yes

0.072

5,458,294

Yes

0.070

3,625,006

Yes

0.068

2,556,887

Table 9: Employment at Firms

This table examines employment at firms started by entrepreneurs who have participated

in the gig economy prior to founding. Panel A studies total employees and Panel B focuses

on firms with relatively high employment. Gig W orker Staggered is an indicator variable

equaling one if an individual received gig income prior to year t and if there are at least

30 individuals located in the county receiving gig income prior to year t. Has Employees

is an indicator variable equaling one if a firm has any employees in a particular year.

At Least F ive Employees is an indicator variable equaling one if a firm has at least

five employees in a particular year. Appendix A provides additional details on variable

definitions. All models include county × year and industry fixed effects. Industries are

defined at the four-digit NAICS code level. For ease of interpretation, the coefficients

and standard errors are multiplied by 100. Standard errors are reported in parentheses

and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%,

respectively.

Panel A: Employment

Has Employees After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

0.880***

(0.140)

1.204***

(0.171)

2.055***

(0.210)

County × Year FE

Industry FE

Observations

Yes

0.061

5,458,294

Yes

0.064

3,625,006

Yes

0.066

2,556,887

Panel B: High Employment

At Least Five Employees After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

0.523***

(0.094)

0.740***

(0.121)

1.189***

(0.139)

County × Year FE

Industry FE

Observations

Yes

0.056

5,458,294

Yes

0.059

3,625,006

Yes

0.061

2,556,887

Table 10: Employment of Contractors

This table studies the employment of contractors at firms founded by entrepreneurs participating in the gig economy prior to founding. Panel A examines the extensive margin

and Panel B evaluates the intensive margin. Gig W orker Staggered is an indicator

variable equaling one if an individual received gig income prior to year t and if there

are at least 30 individuals located in the county receiving gig income prior to year t.

Has Contractors is an indicator variable if a firm employed any independent contractors

in a particular year. N umber of Contractors is the number of contractors hired by a

firm in a particular year. Appendix A provides additional details on variable definitions.

All models include county × year and industry fixed effects. Industries are defined at

the four-digit NAICS code level. For ease of interpretation, the coefficients and standard

errors are multiplied by 100. Standard errors are reported in parentheses and clustered

at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Panel A: Employment of Contractors

Has Contractors After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

0.892***

(0.121)

1.237***

(0.162)

1.805***

(0.188)

County × Year FE

Industry FE

Observations

Yes

0.027

5,458,294

Yes

0.028

3,625,006

Yes

0.031

2,556,887

Panel B: Number of Contractors

Number of Contractors After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

31.360***

(7.541)

22.522***

(3.455)

36.950***

(10.710)

County × Year FE

Industry FE

Observations

Yes

0.073

5,418,130

Yes

0.074

3,587,734

Yes

0.085

2,520,643

Table 11: Capital Structure

This table evaluates the use of debt at firms started by entrepreneurs who have received

gig income prior to founding. Gig W orker Staggered is an indicator variable equaling one

if an individual received gig income prior to year t and if there are at least 30 individuals

located in the county receiving gig income prior to year t. Has Debt is an indicator

variable if a firm has debt by a particular year after founding based on reporting interest

expense. Appendix A provides additional details on variable definitions. All models

include county × year and industry fixed effects. Industries are defined at the four-digit

NAICS code level. For ease of interpretation, the coefficients and standard errors are

multiplied by 100. Standard errors are reported in parentheses and clustered at the

county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Has Debt After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

1.236***

(0.123)

2.293***

(0.189)

3.409***

(0.270)

County × Year FE

Industry FE

Observations

Yes

0.043

5,458,294

Yes

0.050

3,625,006

Yes

0.055

2,556,887

Table 12: Founder Income

This table examines the role of the gig economy on the income for entrepreneurs. Panel A

evaluates entrepreneurs’ change in income and Panel B focuses on whether entrepreneurs

rise in the income distribution. Gig W orker Staggered is an indicator variable equaling

one if an individual received gig income prior to year t and if there are at least 30

individuals located in the county receiving gig income prior to year t. Change in Income

is the difference in the log of an entrepreneur’s adjusted gross income in a particular

year relative to the firm’s founding year. Increase in Income P ercentile is an indicator

variable equaling one if an entrepreneur’s income percentile in a particular year increases

relative to the firm’s founding year. Income percentiles are based on adjusted gross

income in a county-year. Appendix A provides additional details on variable definitions.

All models include county × year and industry fixed effects. Industries are defined at

the four-digit NAICS code level. For ease of interpretation, the coefficients and standard

errors are multiplied by 100. Standard errors are reported in parentheses and clustered

at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Panel A: Income Change

Change in Income Relative to Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

3.118***

(0.529)

9.105***

(0.738)

12.326***

(1.032)

County × Year FE

Industry FE

Observations

Yes

0.005

8,070,012

Yes

0.006

6,745,290

Yes

0.007

5,626,131

Panel B: Income Distribution

Increase in Income Percentile After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

1.149***

(0.116)

1.861***

(0.142)

2.027***

(0.173)

County × Year FE

Industry FE

Observations

Yes

0.008

8,070,967

Yes

0.010

6,745,833

Yes

0.011

5,626,378

Appendix A

Variable Definitions

This appendix provides variable definitions.

• Adjusted Gross Income is the adjusted gross income of an individual as reported on

Form 1040 in a particular year and rounded due to confidentiality reasons.

• Age is the age an individual turns in a particular year.

• Change in Income is the difference in the log of an entrepreneur’s adjusted gross

income in a particular year relative to the firm’s founding year.

• Employees is a count of the number of employees at a firm.

• F emale is an indicator variable equaling one if is a gig worker is female.

• F irst-time F ounder is an indicator variable equaling one if an individual starts any

new firms in a particular year and has not previously created a firm.

• F ounder is an indicator variable equaling one if an individual starts any new firms in

a particular year.

• Gig W orker is an indicator variable equaling one if an individual received gig income

prior to year t.

• Gig Availability is an indicator variable equaling one if when there are at least 30

individuals located in the county receiving gig income by a particular year.

• Gig W orker N umber of Y ears is the number of years that an individual received gig

income in any year up to and including t-1.

• Gig W orker P revious Y ear is an indicator variable equaling one if an individual

received gig income in year t-1.

• Gig W orker Staggered is an indicator variable equaling one if an individual received

gig income prior to year t and if there are at least 30 individuals located in the county

receiving gig income prior to year t.

• Gig W orker Staggered T ransportation is an indicator variable equaling one if an

individual received gig income in any year prior to year t from a gig firm classified as

transportation and if there are at least 30 individuals located in the county receiving

gig income prior to year t. The variables for leasing, selling, and services are similarly defined for gig income from a gig firm classified as leasing, selling, or services,

respectively.

• Gig W orker T ransportation is an indicator variable equaling one if an individual

received gig income in any year up to and including t-1 from a gig firm classified as

transportation. The variables for leasing, selling, and services are similarly defined for

gig income from a gig firm classified as leasing, selling, or services, respectively.

• Has Contractors is an indicator variable if a firm employed any independent contractors in a particular year.

• Has Debt is an indicator variable if a firm has debt by a particular year after founding

based on reporting interest expense.

• Has Dependents is an indicator variable equaling one if a gig worker has any dependents based on Form 1040.

• Increase in Income P ercentile is an indicator variable equaling one if an entrepreneur’s

income percentile in a particular year increases relative to the firm’s founding year. Income percentiles are based on adjusted gross income in a county-year.

• Low Income is an indicator variable equaling one if an individual’s adjusted gross

income is in the bottom tercile in a particular county-year.

• N umber of Contractors is the number of contractors hired by a firm in a particular

year.

• P rof itability is the inverse hyperbolic sine of a firm’s gross profits in a particular year.

• Receives EIT C is an indicator variable equaling one if an individual received any

Earned Income Tax Credit in a particular year.

• Repeat Sector is an indicator variable equaling one if an entrepreneur’s newly started

firm is in the same sector as any prior W-2 or 1099 experience.

• Repeat Sector Gig is an indicator variable equaling one if an entrepreneur’s newly

started firm is in the same sector as any prior gig experience.

• Revenues is the log of one plus revenues for a firm in a particular year.

• Single is a gig worker’s filing status.

• Single with Dependents is an indicator variable equaling one if an individual’s filing

status on Form 1040 in a particular year is single and the individual has any dependents

based on Form 1040.

• Survival is an indicator variable equaling one if a firm files taxes in a particular year

after founding.

Figure A1: Type of Gig Firms

Number of Gig Firms

This figure provides the distribution of gig firms in the United States from 2012 to 2021. Each gig firm is classified into

one of the following categories: leasing, selling, services, and transportation. The bars show the count of gig firms in a

particular category.

Leasing

Selling

Services

Transportation

Figure A2: Gig Economy: Transportation and Non-Transportation

This figure plots the U.S. participation in the gig economy from 2012 to 2021 based on the type

of gig firm. Panel A shows the cumulative number of individuals working in the gig economy.

Panel B provides the total amount of income in billions of dollars received by gig workers in a

particular year, which is converted to dollars in 2012. The gray bars show participation in the

gig economy for gig firms categorized as transportation. The red bars indicate participation

in the gig economy for gig firms categorized as non-transportation, which includes leasing,

selling, and services as described in Section 2.2 and shown in Figure A1.

Number of Gig Workers (millions)

Panel A: Cumulative Number of Gig Workers

2012

2013

2014

2015

2016

2017

Transportation

2018

2019

2020

2021

Non-transportation

Gig Income ($ billions)

100

120

Panel B: Gig Income ($ 2012)

2012

2013

2014

2015

2016

Transportation

2017

2018

2019

2020

Non-transportation

2021

Figure A3: Mechanisms: Who Responds in the Gig Economy

This figure provides the coefficients and 95% confidence intervals for subsamples based on founders’ characteristics. Panels

A and B plot the estimates for revenues and employees at founding, respectively. Panels C and D show the estimates

for survival and profitability three years after founding, respectively. The characteristics are income, flexibility, and age.

Low (High) Income is based on below (above) bottom tercile income in a particular county-year. M ore T ime Constrained

is defined as individuals who are single with dependents. Less T ime Constrained are individuals who are not single with

dependents. Y oung (Old) is constructed as below (above) median age.

Panel A: Revenues at Founding

Panel B: Employees at Founding

Low Income

High Income

More Time Constrained

Less Time Constrained

Young

Old

Low Income

-.1

Panel C: Survival Three Years After Founding

-.4

Low Income

High Income

More Time Constrained

Less Time Constrained

Young

Old

-.04

-.03

-.02

-.01

-.2

-.1

Panel D: Profitability Three Years After Founding

Low Income

-.05

-.3

.01

-.1

Table A1: Entry into Entrepreneurship and the Gig Economy: Oster Test

This table reports the results for the Oster test. Panel A provides the Oster test inputs and findings for the regression

in Table 4, Panel A, Column (4). Panel B shows the Oster test inputs and results for the specification in Table 4, Panel

A, Column (5). βadj is the adjusted coefficient as defined in Section 4.2. The identified set represents the bounds on the

coefficient incorporating the potential influence of omitted variables. The null hypothesis is that β = 0. The last column

provides the value of δ such that βadj = 0.

Panel A: Oster Test for Table 4, Panel A, Column (4)

Excluding Controls

βu

Ru2

Including Controls

βc

Rc2

βadj

Identified Set

Reject Null?

δ s.t. βadj = 0

1.146

0.981

0.749

[0.749, 0.981]

Yes

4.228

0.000280

0.001910

Panel B: Oster Test for Table 4, Panel A, Column (5)

Excluding Controls

βu

Ru2

Including Controls

βc

Rc2

βadj

Identified Set

Reject Null?

δ s.t. βadj = 0

1.147

0.980

0.746

[0.746, 0.980]

Yes

4.179

0.000277

0.001905

Table A2: Entry into Entrepreneurship and the Gig Economy: Robustness to Measure of Gig Work

This table evaluates robustness of the baseline estimates for the effect of the gig economy on entry into entrepreneurship

using different measures of gig work. F ounder is an indicator variable equaling one if an individual starts any new firms

in a particular year. Gig W orker P revious Y ear is an indicator variable equaling one if an individual received gig income

in year t − 1. Gig W orker Staggered P revious Y ear is an indicator variable equaling one if an individual received gig

income in year t − 1 and if there are at least 30 individuals located in the county receiving gig income prior to year t.

Gig W orker N umber of Y ears is the number of years that an individual received gig income in any year prior to year t.

Gig W orker Staggered N umber of Y ears is the number of years that an individual received gig income in any year prior

to year t and if there are at least 30 individuals located in the county receiving gig income prior to year t. The controls

are log adjusted gross income, gender, filing status, dependents, and log age. The granular fixed effects are indicators

for each category of the controls. The sample includes all U.S. tax filers from 2012 to 2021 aged 25 to 65. The unit of

observation is an individual-year. Appendix A provides additional details on variable definitions. For ease of interpretation,

the coefficients and standard errors are multiplied by 100. Standard errors are reported in parentheses and clustered at the

county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Panel A: Gig Work in Previous Year

Founder

(1)

(2)

(3)

(4)

Gig Worker

Previous Year

Gig Worker Staggered

Previous Year

0.938***

(0.034)

0.934***

(0.033)

0.883***

(0.032)

0.881***

(0.031)

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

Yes

0.001

1,265,172,170

(5)

0.878***

(0.032)

Yes

0.001

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Table A2 (continued)

Panel B: Number of Years of Gig Work

Founder

(1)

(2)

(3)

(4)

Gig Worker

Number of Years

Gig Worker Staggered

Number of Years

0.482***

(0.023)

0.478***

(0.022)

0.463***

(0.021)

0.461***

(0.021)

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

Yes

0.001

1,265,172,170

(5)

0.460***

(0.021)

Yes

0.001

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Table A3: Entry into Entrepreneurship and the Gig Economy: Robustness for Type of Gig Firm

This table provides robustness of the baseline estimates for the effect of the gig economy on entry into entrepreneurship using variation in type of

gig work. F ounder is an indicator variable equaling one if an individual starts any new firms in a particular year. Gig W orker T ransportation

is an indicator variable equaling one if an individual received gig income in any year prior to year t from a gig firm classified as transportation.

Gig W orker Staggered T ransportation is an indicator variable equaling one if an individual received gig income in any year prior to year t from a

gig firm classified as transportation and if there are at least 30 individuals located in the county receiving gig income prior to year t. The variables for

leasing, selling, and services are similarly defined for gig income from a gig firm classified as leasing, selling, or services, respectively. The controls are

log adjusted gross income, gender, filing status, dependents, and log age. The granular fixed effects are indicators for each category of the controls.

The sample includes all U.S. tax filers from 2012 to 2021 aged 25 to 65. The unit of observation is an individual-year. Appendix A provides additional

details on variable definitions. For ease of interpretation, the coefficients and standard errors are multiplied by 100. Standard errors are reported in

parentheses and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Founder

Gig Worker Transportation

Gig Worker Leasing

Gig Worker Selling

Gig Worker Services

(1)

(2)

(3)

(4)

0.875***

(0.039)

1.092***

(0.038)

1.426***

(0.039)

0.942***

(0.019)

0.863***

(0.037)

1.094***

(0.038)

1.430***

(0.039)

0.939***

(0.019)

0.831***

(0.036)

1.108***

(0.036)

1.340***

(0.038)

0.845***

(0.018)

0.833***

(0.035)

1.048***

(0.035)

1.312***

(0.038)

0.835***

(0.017)

Gig Worker Staggered Transportation

0.834***

(0.036)

1.048***

(0.035)

1.323***

(0.040)

0.837***

(0.018)

Gig Worker Staggered Leasing

Gig Worker Staggered Selling

Gig Worker Staggered Services

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Yes

0.002

1,265,172,170

Table A4: Entry into Entrepreneurship and the Gig Economy: Robustness to Early versus Late

This table evaluates robustness of the baseline estimates to splitting the sample period to early versus late. Panel A

presents the estimates from 2012 to 2016 and Panel B shows the results from 2017 to 2021. F ounder is an indicator

variable equaling one if an individual starts any new firms in a particular year. F irst-time F ounder is an indicator variable

equaling one if an individual starts any new firms in a particular year and has not previously created a firm. Gig W orker

is an indicator variable equaling one if an individual received gig income prior to year t. Gig W orker Staggered is an

indicator variable equaling one if an individual received gig income prior to year t and if there are at least 30 individuals

located in the county receiving gig income prior to year t. The controls are log adjusted gross income, gender, filing status,

having any dependents, and log age. The granular fixed effects are indicators for each category of the controls. The sample

includes all U.S. tax filers during the respective sample period aged 25 to 65. The unit of observation is an individual-year.

Appendix A provides additional details on variable definitions. For ease of interpretation, the coefficients and standard

errors are multiplied by 100. Standard errors are reported in parentheses and clustered at the county level. *** , ** , and *

denote significance at 1%, 5%, and 10%, respectively.

Panel A: Early Years (2012−2016)

Founder

Gig Worker

(1)

(2)

(3)

(4)

1.009***

(0.035)

1.010***

(0.034)

0.987***

(0.034)

0.983***

(0.034)

Gig Worker Staggered

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

0.979***

(0.035)

Yes

0.000

556,320,761

Yes

0.000

556,320,761

Yes

0.001

556,320,761

Yes

0.001

556,320,761

Yes

0.001

556,320,761

Table A4 (continued)

Panel B: Late Years (2017−2021)

Founder

Gig Worker

(1)

(2)

(3)

(4)

1.029***

(0.035)

1.026***

(0.035)

0.971***

(0.033)

0.968***

(0.033)

Gig Worker Staggered

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

0.967***

(0.033)

Yes

0.001

708,851,409

Yes

0.001

708,851,409

Yes

0.002

708,851,409

Yes

0.002

708,851,409

Yes

0.002

708,851,409

Table A5: Entry into Entrepreneurship and the Gig Economy: Heterogeneity in Financial Constraints

This table provides heterogeneity by financial constraints for the effect of the gig economy on entry into entrepreneurship.

Panel A examines firms started in industries with high fixed costs and Panel B evaluates firms started in industries with

low fixed costs. F ounder is an indicator variable equaling one if an individual starts any new firms in a particular year.

Gig W orker is an indicator variable equaling one if an individual received gig income prior to year t. Gig W orker Staggered

is an indicator variable equaling one if an individual received gig income prior to year t and if there are at least 30 individuals

located in the county receiving gig income prior to year t. The controls are log adjusted gross income, gender, filing status,

having any dependents, and log age. The granular fixed effects are indicators for each category of the controls. The sample

includes all U.S. tax filers from 2012 to 2021 aged 25 to 65. The unit of observation is an individual-year. Appendix A

provides additional details on variable definitions. For ease of interpretation, the coefficients and standard errors are

multiplied by 100. Standard errors are reported in parentheses and clustered at the county level. *** , ** , and * denote

significance at 1%, 5%, and 10%, respectively.

Panel A: High Fixed Cost

Founder in High Fixed Cost Industry

Gig Worker

(1)

(2)

(3)

(4)

0.713***

(0.027)

0.709***

(0.025)

0.672***

(0.025)

0.670***

(0.025)

Gig Worker Staggered

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

0.712***

(0.027)

Yes

0.000

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.000

1,265,172,170

Table A5 (continued)

Panel B: Low Fixed Cost

Founder in Low Fixed Cost Industry

Gig Worker

(1)

(2)

(3)

(4)

0.299***

(0.010)

0.295***

(0.010)

0.290***

(0.009)

0.289***

(0.009)

Gig Worker Staggered

County FE

Year FE

County × Year FE

Granular FE

Controls

Observations

(5)

0.298***

(0.010)

Yes

0.000

1,265,172,170

Yes

0.000

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Yes

0.000

1,265,172,170

Table A6: Who Responds in the Gig Economy?

This table evaluates the role of characteristics in the effect of the gig economy on entry into entrepreneurship. F ounder is

an indicator variable equaling one if an individual starts any new firms in a particular year. Gig W orker is an indicator

variable equaling one if an individual received gig income in any year prior to year t. Log AGI is log adjusted gross

income. Low Income is an indicator variable equaling one if an individual’s adjusted gross income is in the bottom tercile

in a particular county-year. Receives EIT C is an indicator variable equaling one if an individual received any Earned

Income Tax Credit in a particular year. Log Age is the log of an individual’s age in a particular year. Has Dependents

is an indicator variable equaling one if a gig worker has any dependents based on Form 1040. Single with Dependents is

an indicator variable equaling one if an individual’s filing status in a particular year is single and the individual has any

dependents based on Form 1040. Appendix A provides additional details on variable definitions. For ease of interpretation,

the coefficients and standard errors are multiplied by 100. All models include county-year fixed effects. Standard errors

are reported in parentheses and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%,

respectively.

Founder

Gig Worker × Characteristic

Gig Worker

Characteristic

County × Year FE

Observations

(1)

(2)

(3)

(4)

(5)

(6)

-0.045***

(0.004)

1.484***

(0.072)

-0.013***

(0.001)

0.054***

(0.016)

1.004***

(0.031)

-0.032***

(0.005)

0.060**

(0.028)

0.994***

(0.028)

0.081***

(0.006)

-0.294***

(0.035)

2.036***

(0.155)

-0.655***

(0.010)

0.065***

(0.018)

0.995***

(0.028)

0.186***

(0.009)

0.233***

(0.029)

0.985***

(0.030)

-0.089***

(0.009)

Log AGI

Low Income

Log Age

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Receives

EITC

Yes

0.001

1,265,172,170

Has

Dependents

Yes

0.001

1,265,172,170

Single with

Dependents

Yes

0.001

1,265,172,170

Yes

0.001

1,265,172,170

Table A7: Previous Work Experience and the Gig Economy: Controls Approach

This table provides robustness for the relationship between prior work experience and

new firm creation. The sample is individuals creating new firms from 2012 to 2021.

Repeat Sector is an indicator variable equaling one if an entrepreneur’s newly started

firm is in the same sector as any prior W-2 or 1099 experience. Repeat Sector Gig is

an indicator variable equaling one if an entrepreneur’s newly started firm is in the same

sector as any prior gig experience. Gig W orker is an indicator variable equaling one if an

individual received gig income in any year prior to year t. Appendix A provides additional

details on variable definitions. All models include county × year fixed effects. For ease

of interpretation, the coefficients and standard errors are multiplied by 100. Standard

errors are reported in parentheses and clustered at the county level. *** , ** , and * denote

significance at 1%, 5%, and 10%, respectively.

Repeat Sector

Repeat Sector Gig

(1)

(2)

Gig Worker

4.089***

(0.273)

1.191***

(0.038)

County × Year FE

Observations

Yes

0.021

9,910,324

Yes

0.013

9,910,324

Table A8: Firms at Founding: Controls

This table examines robustness about the role of the gig economy on firms at founding. The sample includes all firms created by individuals in the year of founding.

Gig W orker Staggered is an indicator variable equaling one if an individual received

gig income prior to year t and if there are at least 30 individuals located in the county

receiving gig income prior to year t. Revenues is the log of one plus revenues for a

firm. Employees is a count of the number of employees at a firm. The controls are

log adjusted gross income, gender, filing status, dependents, and log age. Appendix A

provides additional details on variable definitions. All models include county × year and

industry fixed effects. Industries are defined at the four-digit NAICS code level. For ease

of interpretation, the coefficients and standard errors are multiplied by 100. Standard

errors are reported in parentheses and clustered at the county level. *** , ** , and * denote

significance at 1%, 5%, and 10%, respectively.

Revenues

Employees

(1)

(2)

Gig Worker Staggered

13.108***

(1.468)

17.990***

(5.108)

County × Year FE

Industry FE

Controls

Observations

Yes

0.060

9,294,588

Yes

0.113

9,283,567

Table A9: Firms at Founding: Controls Approach

This table examines robustness about the role of the gig economy on firms at founding.

The sample includes all firms created by individuals in the year of founding. Gig W orker

is an indicator variable equaling one if an individual received gig income in any year prior

to year t. Revenues is the log of one plus revenues for a firm. Employees is a count of

the number of employees at a firm. Appendix A provides additional details on variable

definitions. All models include county × year and industry fixed effects. Industries are

defined at the four-digit NAICS code level. For ease of interpretation, the coefficients

and standard errors are multiplied by 100. Standard errors are reported in parentheses

and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%,

respectively.

Revenues

Employees

(1)

(2)

Gig Worker

20.749***

(1.433)

33.266***

(5.135)

County × Year FE

Industry FE

Observations

Yes

0.049

9,910,312

Yes

0.088

9,899,970

Table A10: Firms at Founding: At Least One Employee

This table studies the role of the gig economy on firms at founding. The sample includes

all firms created by individuals in the year of founding and having at least one employee

within the first five years of founding. Gig W orker Staggered is an indicator variable

equaling one if an individual received gig income prior to year t and if there are at least

30 individuals located in the county receiving gig income prior to year t. Revenues is the

log of one plus revenues for a firm. Employees is a count of the number of employees at a

firm. Appendix A provides additional details on variable definitions. All models include

county × year and industry fixed effects. Industries are defined at the four-digit NAICS

code level. For ease of interpretation, the coefficients and standard errors are multiplied

by 100. Standard errors are reported in parentheses and clustered at the county level.

*** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Revenues

Employees

(1)

(2)

Gig Worker Staggered

3.615**

(1.692)

21.648***

(5.138)

County × Year FE

Industry FE

Observations

Yes

0.062

1,798,131

Yes

0.120

1,797,858

Table A11: Firm Performance: Controls

This table evaluates the performance of firms started by entrepreneurs who have received

gig income prior to founding. Panel A examines firm survival and Panel B studies firm

profitability. The sample tracks all firms created by individuals in the three years following founding. Gig W orker Staggered is an indicator variable equaling one if an

individual received gig income prior to year t and if there are at least 30 individuals located in the county receiving gig income prior to year t. Survival is an indicator variable

equaling one if a firm files taxes in a particular year after founding. P rof itability is the

inverse hyperbolic sine of a firm’s gross profits in a particular year. The controls are

log adjusted gross income, gender, filing status, dependents, and log age. Appendix A

provides additional details on variable definitions. All models include county × year and

industry fixed effects. Industries are defined at the four-digit NAICS code level. For ease

of interpretation, the coefficients and standard errors are multiplied by 100. Standard

errors are reported in parentheses and clustered at the county level. *** , ** , and * denote

significance at 1%, 5%, and 10%, respectively.

Panel A: Firm Survival

Survival After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

-1.931***

(0.123)

-2.098***

(0.181)

-2.386***

(0.206)

County × Year FE

Industry FE

Controls

Observations

Yes

0.017

8,140,192

Yes

0.019

6,996,972

Yes

0.020

5,977,986

Panel B: Firm Profitability

Profitability After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

12.799***

(2.247)

16.566***

(2.595)

13.230***

(2.912)

County × Year FE

Industry FE

Controls

Observations

Yes

0.090

5,172,673

Yes

0.089

3,433,107

Yes

0.087

2,421,652

Table A12: Firm Performance: Controls Approach

This table evaluates the performance of firms started by entrepreneurs who have received

gig income prior to founding. Panel A examines firm survival and Panel B studies firm

profitability. The sample tracks all firms created by individuals in the three years following

founding. Gig W orker is an indicator variable equaling one if an individual received gig

income in any year prior to year t. Survival is an indicator variable equaling one if a

firm files taxes in a particular year after founding. P rof itability is the inverse hyperbolic

sine of a firm’s gross profits in a particular year. Appendix A provides additional details

on variable definitions. All models include county × year and industry fixed effects.

Industries are defined at the four-digit NAICS code level. For ease of interpretation, the

coefficients and standard errors are multiplied by 100. Standard errors are reported in

parentheses and clustered at the county level. *** , ** , and * denote significance at 1%,

5%, and 10%, respectively.

Panel A: Firm Survival

Survival After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker

-2.557***

(0.133)

-2.943***

(0.175)

-3.185***

(0.198)

County × Year FE

Industry FE

Observations

Yes

0.015

8,694,751

Yes

0.016

7,487,349

Yes

0.016

6,387,091

Panel B: Firm Profitability

Profitability After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker

33.516***

(2.128)

38.681***

(2.382)

35.796***

(2.659)

County × Year FE

Industry FE

Observations

Yes

0.072

5,458,294

Yes

0.070

3,625,006

Yes

0.068

2,556,887

Table A13: Firm Performance: At Least One Employee

This table evaluates the performance of firms started by entrepreneurs who have received

gig income prior to founding. Panel A examines firm survival and Panel B studies firm

profitability. The sample tracks all firms created by individuals in the three years following founding and having at least one employee within the first five years of founding.

Gig W orker Staggered is an indicator variable equaling one if an individual received

gig income prior to year t and if there are at least 30 individuals located in the county

receiving gig income prior to year t. Survival is an indicator variable equaling one if a

firm files taxes in a particular year after founding. P rof itability is the inverse hyperbolic

sine of a firm’s gross profits in a particular year. Appendix A provides additional details

on variable definitions. All models include county × year and industry fixed effects. Industries are defined at the four-digit NAICS code level. For ease of interpretation, the

coefficients and standard errors are multiplied by 100. Standard errors are reported in

parentheses and clustered at the county level. *** , ** , and * denote significance at 1%,

5%, and 10%, respectively.

Panel A: Firm Survival

Survival After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

-2.629***

(0.274)

-3.370***

(0.317)

-3.636***

(0.370)

County × Year FE

Industry FE

Observations

Yes

0.040

1,670,209

Yes

0.042

1,492,003

Yes

0.044

1,299,151

Panel B: Firm Profitability

Profitability After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker Staggered

8.920***

(2.709)

8.934***

(3.012)

9.055**

(4.237)

County × Year FE

Industry FE

Observations

Yes

0.064

1,164,074

Yes

0.062

859,175

Yes

0.063

637,010

Table A14: Employment at Firms: Controls Approach

This table examines employment at firms started by entrepreneurs who have participated

in the gig economy prior to founding. Panel A studies total employees and Panel B focuses

on firms with relatively high employment. Gig W orker is an indicator variable equaling

one if an individual received gig income in any year prior to year t. Has Employees

is an indicator variable equaling one if a firm has any employees in a particular year.

At Least F ive Employees is an indicator variable equaling one if a firm has at least

five employees in a particular year. Appendix A provides additional details on variable

definitions. All models include county × year and industry fixed effects. Industries are

defined at the four-digit NAICS code level. For ease of interpretation, the coefficients

and standard errors are multiplied by 100. Standard errors are reported in parentheses

and clustered at the county level. *** , ** , and * denote significance at 1%, 5%, and 10%,

respectively.

Panel A: Employment

Has Employees After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker

0.878***

(0.138)

1.217***

(0.169)

2.082***

(0.208)

County × Year FE

Industry FE

Observations

Yes

0.061

5,458,294

Yes

0.064

3,625,006

Yes

0.066

2,556,887

Panel B: High Employment

At Least Five Employees After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker

0.511***

(0.094)

0.730***

(0.120)

1.201***

(0.138)

County × Year FE

Industry FE

Observations

Yes

0.056

5,458,294

Yes

0.059

3,625,006

Yes

0.061

2,556,887

Table A15: Employment of Contractors: Controls Approach

This table studies the employment of contractors at firms founded by entrepreneurs participating in the gig economy prior to founding. Panel A examines the extensive margin

and Panel B evaluates the intensive margin. Gig W orker is an indicator variable equaling

one if an individual received gig income in any year prior to year t. Has Contractors is

an indicator variable if a firm employed any independent contractors in a particular year.

N umber of Contractors is the number of contractors hired by a firm in a particular

year. Appendix A provides additional details on variable definitions. All models include

county × year and industry fixed effects. Industries are defined at the four-digit NAICS

code level. For ease of interpretation, the coefficients and standard errors are multiplied

by 100. Standard errors are reported in parentheses and clustered at the county level.

*** , ** , and * denote significance at 1%, 5%, and 10%, respectively.

Panel A: Employment of Contractors

Has Contractors After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker

0.890***

(0.120)

1.245***

(0.159)

1.818***

(0.185)

County × Year FE

Industry FE

Observations

Yes

0.027

5,458,294

Yes

0.028

3,625,006

Yes

0.031

2,556,887

Panel B: Number of Contractors

Number of Contractors After Founding

One Year

(1)

Two Years

(2)

Three Years

(3)

Gig Worker

31.339***

(7.448)

22.105***

(3.436)

35.830***

(10.583)

County × Year FE

Industry FE

Observations

Yes

0.073

5,418,130

Yes

0.074

3,587,734

Yes

0.085

2,520,643

Table A16: Capital Structure: Controls Approach

This table evaluates the use of debt at firms started by entrepreneurs who have received

gig income prior to founding. Gig W orker is an indicator variable equaling one if an

individual received gig income in any year prior to year t. Has Debt is an indicator

variable if a firm has debt by a particular year after founding based on reporting interest

expense. Append

This text is long and has been trimmed here. Open the source document for the complete record.

This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.