Investor Tax Credits and Entrepreneurship:

Evidence from U.S. States

MATTHEW DENES, SABRINA T. HOWELL, FILIPPO MEZZANOTTI,

XINXIN WANG, and TING XU∗

ABSTRACT

Angel investor tax credits are used globally to spur high-growth entrepreneurship. Exploiting

their staggered implementation in 31 U.S. states, we find that they increase angel investment

yet have no significant impact on entrepreneurial activity. Two mechanisms explain these

results: crowding out of alternative financing and low sensitivity of professional investors to

tax credits. With a large-scale survey and a stylized model, we show that low responsiveness

among professional angels may reflect the fat-tailed return distributions that characterize

high-growth startups. The results contrast with evidence that direct subsidies to firms have

positive effects, raising concerns about promoting entrepreneurship with investor subsidies.

JEL Classification: E24, G24, H71, L26

Keywords: entrepreneurship, investor tax credit, angel financing, government subsidy
∗
Matthew Denes is at the Carnegie Mellon University Tepper School of Business. Sabrina T. Howell is at the New York
University Stern School of Business. Filippo Mezzanotti is at the Northwestern University Kellogg School of Management.
Xinxin Wang is at the UCLA Anderson School of Management. Ting Xu is at the University of Toronto Rotman School of
Management. This paper subsumes two prior working papers: Denes, Wang, and Xu (2019) and Howell and Mezzanotti (2019).
We thank Jim Albertus, Tania Babina, Laurent Bach, Shai Bernstein, Greg Brown, Annamaria Conti, Jesse Davis, Ryan
Decker, Mike Ewens, Joan Farre-Mensa, Paolo Fulghieri, Andra Ghent, Juanita Gonzalez-Uribe, Will Gornall, Apoorv Gupta,
Arpit Gupta, Jorge Guzman, Thomas Hellmann, Yael Hochberg, Edith Hotchkiss, Yunzhi Hu, Jessica Jeffers, Simone Lenzu,
Josh Lerner, Song Ma, David Matsa, Arnobio Morelix, Holger Mueller, Ramana Nanda, Daniel Paravisini, Andrea Passalacqua,
David Robinson, Pian Shu, Morten Sorenson, Denis Sosyura, Chester Spatt, Luke Stein, Kairong Xiao, Emmanuel Yimfor,
Linghang Zeng, Eric Zwick, and seminar participants at the 3rd Junior Entrepreneurial Finance and Innovation Workshop, 8th
HEC Paris Workshop on Entrepreneurship, AFA, ASU Sonoran Winter Finance Conference, Barcelona GSE Summer Forum,
Bocconi, Carnegie Mellon University, Duke-UNC Innovation and Entrepreneurship Research Symposium, Finance in the Cloud
II, Jackson Hole Finance Group Conference, Kellogg, Kenan Institute Frontiers of Entrepreneurship Conference, MFA, Mid-
Atlantic Research Conference in Finance, Northeastern Finance Conference, NBER Conference on Business Taxation in a
Federal System, NTU, NYU Stern Corporate Governance Seminar, Red Rock Finance Conference, Southern California PE
Conference, Texas A&M, UCLA, UNC Entrepreneurship Working Group, UVA Darden, WFA, and Young Scholar in Finance
Consortium for helpful comments. We thank Abhishek Bhardwaj, Grant Goehring, Michael Gropper, Sunwoo Hwang, Nick
McMonigle, Danye Wang, and Jun Wong for excellent research assistance. Howell’s research on this project was funded by the
Kauffman Foundation and the UNC Kenan Institute. We have read the Journal of Finance’s disclosure policy and have no
conflicts of interest to disclose.
Correspondence: Sabrina T. Howell, New York University, Stern School of Business, 44 West Fourth Street, New York, NY
10012, email: [email protected]

1
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fostering high-growth entrepreneurship is crucial for long-term economic success. As a

result, governments around the world deploy tools such as grants, loan guarantees, prize

competitions, and tax subsidies. This paper studies a popular policy that has been adopted

by more than 14 countries around the world and by the majority of U.S. states: angel

investor tax credits.1 These programs offer personal income tax credits equal to a certain

percentage of the investment, regardless of the investment outcome. While this tax policy

has attracted much attention and debate, we know little about its effects on investors and

startups.2

Tax subsidies targeting angel investors have several attractive features. First, there is no

need for the government to “pick winners,” which requires policymakers to be informed about

firm quality and could lead to regulatory capture (Lerner (2009)). Tax credits retain market

incentives, leaving investors with skin in the game. Second, the administrative burden of

tax subsidies is relatively low. Third, angel investor tax credits are a more precise tool than

broad cuts to capital gains taxes (Poterba (1989)). However, while tax credit programs offer

attractive flexibility, there is no guarantee that investors will respond by increasing financing

in the startups that policymakers target.

To assess the effect of angel tax credits, we exploit their staggered introductions and

terminations from 1988 to 2018 across 31 states in the U.S. In our baseline analysis, we use a

differences-in-differences framework at the state-year level to identify the effect of tax credits.

We show that state-level economic, political, fiscal, and entrepreneurial factors do not predict
1
Angels are wealthy individuals who invest in early-stage startups in exchange for equity or convertible debt.

Other countries with angel tax credits include Australia, Brazil, Canada, China, England, France, Germany, Ireland,

Italy, Japan, Portugal, Singapore, Spain, and Sweden. In the U.S., angel tax credits represent significant portions

of state entrepreneurship budgets, and we calculate that they support up to $13.2 billion of angel investment. On

average, investors use 88% of available funding.

2
See “The Problem with Tax Credits for Angel Investors,” Bloomberg Businessweek, 8/20/2010; “Should Angel

Investors Get Tax Credits to Invest in Small Businesses?,” Wall Street Journal, 3/18/2012; “Angel Investment Tax

Credit Pricey but Has Defenders,” Minnesota Star Tribune, 10/31/2015.

2
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
the implementation of angel tax credits, which suggests that program timing is unrelated to

local economic conditions. We evaluate the impact of angel tax credit programs using data

on angel activity from Crunchbase, VentureXpert, VentureSource, Form D filings, and

AngelList. For a subset of states, we also employ data from state governments on the

identity of firms and investors that benefit from these tax credit programs.

We find that angel tax credits increase the number of angel investments by approximately

18% and the number of individual angel investors by 32%. This effect is amplified when

programs impose fewer restrictions and when the supply of alternative startup capital is

more limited. However, additional investment flows to older firms, to firms with lower

employment growth, and to fewer serial entrepreneurs. Average ex-ante growth

characteristics of angel-backed firms in the state also deteriorate after the implementation of

angel tax credits. This may be expected if relaxing financial constraints reduces the quality

of firms financed at the margin (Evans and Jovanovic (1989)), and does not imply that the

investments are not privately or socially valuable. Nonetheless, the declines raise concerns

about the ability of angel tax credits to reach high-growth startups and have a significant

impact on the local economy.

We next test whether angel tax credits achieve the programs’ objectives—as stated in

legislation—of high-tech firm entry and job creation using data from the U.S. Census

Business Dynamics Statistics. Across many approaches, we consistently find null effects that

are statistically insignificant and have economically small confidence intervals. To address

the concern that angel tax credits reallocate capital within a state, we show that there are no

effects either in regions with the most angel investments or those with limited early-stage

capital. Null effects persist across other outcome variables, including LinkedIn-based firm

entry and job creation, Delaware-incorporated firms, and patenting activity.

To assess whether the null results reflect a lack of statistical power, we conduct a power

analysis to determine the smallest effect that could be statistically rejected, which is referred

to as the minimum detectable effect. We find that the minimum detectable effects are small

3
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
both in absolute terms and relative to a range of plausible expected effects of angel tax credits

(i.e., priors), calculated under assumptions about how the increase in angel investments may

translate to new firm creation. For example, the estimated effect on the count of young, high-

tech firms in our preferred model is -0.3%, compared to a minimum detectable effect at 80%

power of 1.9% and a corresponding prior of 3.3%. These null effects are informative. Abadie

(2020) notes that when a policy is expected to be effective and there is sufficient power, null

effects are potentially more informative than significant effects.

We also examine whether the null effects could be due to small program scale. We find

no effect at the firm level when we compare firms backed by subsidized investors with firms

that were certified but failed to have an investor receive a tax credit. Further, we continue to

find null results for states with large programs or when we use a dollarized treatment variable.

This indicates that the null effects do not reflect small program scale.

The null real effects on state-level firm entry and job creation contrast with the positive

effects documented in the literature for other tax credits (e.g., Cummins et al. (1994), Hall

and Van Reenen (2000), Dechezleprêtre et al. (2020), Zwick and Mahon (2017), Arefeva et al.

(2020), Edwards and Todtenhaupt (2020), Freedman, Neumark, and Khanna (2021)). These

papers study programs that either directly target the operating firm rather than the financial

intermediary, or target investment in firms with relatively predictable cash flows. Conversely,

angel tax credit programs target financial intermediaries and projects with fat-tailed return

distributions. These differences lead us to two mechanisms that together can help explain why

angel tax credits increase investment yet have no real effects.

The first mechanism is that additional angel investments partially crowd out investment

that would have occurred in the absence of the programs. Several pieces of evidence support

this channel. First, increased angel investment appears to displace other types of early-stage

financing. We show that, following angel tax credits, nonangel early-stage investment decreases

while total early-stage investment does not change. Second, investments that would have likely

occurred regardless of angel tax credits appear to be relabeled as “angel.” Relabeling might be

4
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
more prevalent among insiders who face negligible coordination frictions when investing in their

own firms and may invest for nonfinancial reasons, particularly because tax credit programs

do not restrict how firms use subsidized capital. We find that 35% of beneficiary companies

have at least one investor who is also a company executive or a family member of an executive.

Comparatively, only 8% of angel-backed firms on AngelList had at least one insider investor.

Beyond insiders, investors in general may relabel deals that would have happened regardless

of the policy as angel investment to receive angel tax credits. We examine Securities and

Exchange Commission (SEC) Form D filings, which deals often bypass (Ewens and Malenko

(2020)) but help to demonstrate a legal equity round in order to obtain tax credits, and show

that these filings are more likely for firms with subsidized investors compared to matched

nonbeneficiary firms. Last, we find that firms with subsidized investors do not perform better

than certified firms that failed to have investors receive a tax credit, consistent with crowding

out.

The second mechanism emerges from the type of investors who respond to angel tax credits.

We start by showing that investors receiving angel tax credits are primarily younger, more

local, and less experienced than the average angel investor. The composition of investors also

shifts following the introduction of these programs, with a surge of in-state and inexperienced

investors and little entry of professional, arms-length angels. We conduct a survey of angel

investors to understand why nonprofessional investors are much more responsive to angel tax

credits and receive 1,411 responses. The survey asks angel investors about the importance of

nine factors relevant to evaluating early-stage startups. We find that 51% of respondents rate

angel tax credits as not at all important (the lowest of five options), which increases to 71%

among the most experienced investors. This contrasts with all other factors, which receive

much higher importance. For example, 97% of investors rate the management team as very

or extremely important. When prompted to explain why credits are unimportant, 57% report

that it is because they invest based on whether the startup has the potential to be a home

run. In the words of one respondent, “I’m more focused on the big win than offsetting a loss.”

5
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
To understand why professional investors are less responsive than nonprofessional investors

to tax credits, we build a stylized model by studying the return distributions of early-stage

investments. We assume that more professional investors are more likely to access potentially

high-growth startups whose returns tend to have a fatter right tail. We show that while

angel tax credits increase the probability of investment, this effect declines as the right tail of

the return distribution grows fatter. In particular, professional investors are less sensitive to

investor tax credits because the marginal benefit of the subsidy—which is a fixed percentage

of the investment—decreases as the expected return increases. This suggests that the return

distribution of potentially high-growth firms may limit the effectiveness of angel tax credits.

The stylized model and survey shed new light on how early-stage investors make decisions

(Bernstein, Korteweg, and Laws (2017), Ewens and Townsend (2020)).

Taken together, these results suggest that U.S. state angel tax credits fail to reach the

investor-startup pairs intended by policymakers and can explain why angel tax credits do not

produce significant real effects despite sizable program scale. The crowding-out mechanism

highlights that the increase in angel investment does not appear to translate into an increase

in early-stage capital. The investor heterogeneity mechanism suggests that nonprofessional

investors enter following the introduction of programs and support relatively low-growth and

mature firms, limiting the effect on aggregate firm entry and job creation. The impact of

investor subsidies may therefore depend crucially on the type of investors responding to the

policy (Lee and Persson (2016)).

This paper contributes to the literature on early-stage financing (Robb and Robinson

(2012), Kerr, Lerner, and Schoar (2014a), Hellmann and Thiele (2015), Hochberg, Serrano,

and Ziedonis (2018), Lerner et al. (2018), Xu (2019), Davis, Morse, and Wang (2020)). In

related work, Gonzalez-Uribe and Paravisini (2019) and Lindsey and Stein (2020) look

specifically at policies targeting angel investment. Our findings highlight the importance of

investor heterogeneity. Inexperienced investors or insiders use tax credits for reasons besides

the intended purpose of additional investment in high-growth startups, which is thought to

6
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
be a challenge facing entrepreneurship policy (Acs et al. (2016), Lerner (2020)). To our

knowledge, we are the first to analyze this issue systematically.

More broadly, we contribute to the literature on investment incentives. There is substantial

evidence that related policies have positive effects, including capital gains tax relief, accelerated

investment depreciation, R&D tax credits, and corporate tax cuts (Cummins et al. (1994), Hall

and Van Reenen (2000), Ivković, Poterba, and Weisbenner (2005), Dai et al. (2008), Zwick and

Mahon (2017), Curtis and Decker (2018), Arefeva et al. (2020), Dechezleprêtre et al. (2020),

Edwards and Todtenhaupt (2020)). R&D grant programs have a positive effect on high-tech

startups (Lach (2002), Bronzini and Iachini (2014), Howell (2017), Howell and Brown (2019)).

Accelerators and new-venture competitions are also useful for startups and benefit from public

funds (McKenzie (2017), Cohen et al. (2019), Fehder and Hochberg (2019), Howell (2020)).3

Especially relevant to our setting is an evaluation of the California Competes Tax Credit

(CCTC) by Freedman et al. (2021), which provides businesses with tax credits to incentivize

job creation. They find large local multipliers from each subsidized job. In contrast to these

studies, we present evidence of crowding out of alternative financing.

The above programs are diverse, yet—in addition to being effective—they have a key

feature distinguishing them from angel tax credits: rather than targeting investors or financial

intermediaries, they target firms directly. In contrast, the literature on government-backed

venture capital, where the investor rather than the firm is subsidized, is more mixed (Cumming

and MacIntosh (2006), Brander, Du, and Hellmann (2015), Denes (2019)). Despite being

attractive to policymakers, the flexibility of tax incentives for investors may also limit their

impact. There may be a trade-off between program flexibility and effective targeting, consistent

with evidence from public economics that informational and transaction costs to accessing

government programs can deter the individuals who the programs wish to target (Bhargava
3
Yagan (2015) is one of very few papers to document a null effect of a tax policy that aims to promote business

investment. He finds that the 2003 dividend tax cut had no impact on firm investment or employee compensation,

although it did increase dividend payouts.

7
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
and Manoli (2015), Deshpande and Li (2019), Chetty and Finkelstein (2020)).

The paper is organized as follows. Section I provides an overview of angel investor tax

credits. Section II details the data. Section III explains the identification strategy and studies

the effects of angel investor tax credits on angel investment and real outcomes. Section IV

presents evidence of two mechanisms that explain why angel investment increases yet there is

no impact on entrepreneurial activity, and Section V concludes.

I. Angel Investor Tax Credits

Over the last three decades, 31 states in the U.S. have introduced and passed legislation

to provide accredited angel investors with tax credits. Figure 1, Panel A, provides a map

of states with angel tax credit programs, which we abbreviate as “ATC” hereafter. Blue

shading indicates the tax credit percentage, with darker shades representing larger tax credits.

The figure shows that ATCs are prevalent across the U.S. The extent of these programs is

particularly notable since they do not occur in the seven states with no income tax (shaded in

grey). Panel B shows the introduction and termination of these programs. The earliest was

Maine’s Seed Capital Tax Credit Program, introduced in 1988. A steady progression of states

launched programs over the following three decades. Colorado, Maryland, Minnesota, North

Dakota, and Ohio passed more than one version of an ATC. Although the pace of adoption

has increased in recent years, the geography is dispersed, and program duration varies from

just one year to three decades.

[Figure 1 here]

ATCs are economically meaningful. The mean ratio of program expenditures to total angel

investment is 23%.4 Based on an average tax credit percentage of 34%, these tax credits
4
The mean ratio of program expenditures to seed venture capital is 105%, and the mean ratio of program

expenditures to the Small Business Administration’s (SBA) 7(a) loan program is 14.3%.

8
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
support up to $13.2 billion in angel investment. Furthermore, while the programs are typically

small relative to overall state budgets, they often represent a significant portion of funding

allocated to supporting entrepreneurship or small businesses.5 Finally, investors often use

ATCs, with an average 88% of funding allocated by state legislatures distributed as tax credits.

Tax credits are available to accredited investors and their pass-through entities.6 They

require both the firm and the investor to be certified by the state ex-ante as eligible for the

credit. The investor may apply only after the deal is complete. This requires substantial

coordination between the firm and the investor over a period that is typically several months.

State-level ATCs reduce the state income tax of an investor. For example, suppose that an

investor earns $250,000 in a given year and invests $20,000 in a local startup. If the state tax

rate is 5% on all income, then the investor pays annual state taxes of $12,500. Assuming that

the state introduced an ATC of 35%, the investor can reduce her state taxes by $7,000, which

is a decrease of 56% relative to her annual state taxes.7 Unlike capital gains tax credits that

require positive returns, ATCs are not contingent on the startup’s outcome. Therefore, ATCs

are a fixed subsidy to investors after making an investment.

Policymakers state that they implement ATCs to increase local economic activity,

particularly high-tech firm entry and job creation. For example, Wisconsin notes that “the

Qualified New Business Venture (QNBV) Program helps companies create high-paying,

high-skill jobs throughout Wisconsin.” The Louisiana program goals are “To encourage third

parties to invest in early-stage wealth-creating businesses in the state; to expand the

5
For example, funding for ATC programs in Ohio, Minnesota, and Wisconsin are 19%, 58%, and 86% of annual

state funding for high-tech jobs or small businesses, respectively.

6
We refer to accredited angel investors as angels throughout the paper. An accredited investor is defined as

a person who earned income of more than $200,000 ($300,000 with a spouse) or has a net worth over $1 million.

Since July 2010, net worth excludes home equity (Lindsey and Stein (2020)). The tax implications might differ

for accredited investors compared to pass-through entities. Angel investor tax credits are more likely provided to

individuals because most programs include investment caps.

7
The tax credit available to a particular investor will depend on her state tax liability. Some programs allow

transferable and refundable tax credits, which enable out-of-state investors to benefit from tax credits as well.

9
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
economy of the state by enlarging its base of wealth-creating businesses; and to enlarge the

number of quality jobs available.” The stated goal of Maine’s ATC program is “to spur

venture capital investment in Maine startups and ultimately create more jobs in the state.”8

Since most programs cite spurring new investment and job creation as their goals, the

analysis in subsequent sections focuses on financing outcomes, firm entry, and employment.

Table I provides summary statistics on the ATCs. The maximum share of an investment

that can be deducted from an investor’s tax liability is defined as T ax credit %. The mean

(median) tax credit percentage is 34% (33%). Programs often have eligibility criteria for both

beneficiary companies and investors. They frequently do not allow investors to request cash

in lieu of the credit if they do not have local state income tax liability (72%) or to transfer

the credit (72%). Other restrictions include firm age caps (31% of programs), employment

caps (39%), revenue caps (47%), assets caps (22%), and minimum investment holding periods

(50%). Most programs target the high-tech sector, which guides our empirical design. While

many programs do not allow participation by owners and their families (61%), the majority of

states permit full-time employees, executives, and officers to receive tax credits. Tax credits

reduce income tax liability for the current year, but most programs have a carry-forward

provision (89%). Table IA.I provides comprehensive details for all programs.

[Table I here]

We examine whether economic, political, fiscal, and entrepreneurial factors explain the

introduction of ATCs. Consistent with our identification strategy, we find that these factors

do not significantly predict the introduction of ATC programs. The lack of predictability is

consistent with the presence of considerable frictions in the passage of these programs. Section

III of the Internet Appendix provides additional details about the predictive regression and
8
See Wisconsin Economic Development Corporation 2013 Qualified New Business Venture Program Report,

Louisiana legislation (http://www.legis.la.gov/Legis/Law.aspx?d=321880), and “Startup investors camp out for

Maine tax credit” (https://www.pressherald.com/2019/01/02/startup-investors-camp-out-for-maine-tax-credit).

10
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
examples of frictions in the implementation of ATCs.9

II. Data

This section discusses the data we use on angel deals and investors (Section II.A), state-level

outcomes (Section II.B), and program applicants and beneficiaries (Section II.C).

A. Angel Deals and Investors

Angel investments are difficult to systematically observe in the U.S. because, to our

knowledge, there are no comprehensive data sets about them. Much of what is known about

the size of the angel market relies on survey estimates (Shane (2009)). To overcome this

challenge, we combine data from Crunchbase, Thomson Reuters VentureXpert, and Dow

Jones VentureSource, which we refer to collectively as CVV hereafter, and Form D filings

available from the SEC.10 Form D is a notice of an exempt offering of securities under

Regulation D allowing startups to raise capital from accredited investors without registering

their securities (Ewens and Farre-Mensa (2020)).11 To identify angel rounds, we drop all

financial issuers and focus on the first Form D filing that is not a venture capital (VC)
9
The Internet Appendix is available in the online version of this article on the Journal of Finance website.
10
Crunchbase tracks startup financings using crowdsourcing and news aggregation. VentureXpert and

VentureSource are commercial databases for investments in startups and mainly capture firms that eventually

received venture capital financing. We identify angel investments from these two databases based on round type

and investor type. In Crunchbase, we include round types “pre-seed,” “seed,” “convertible note,” “angel,” or “equity

crowdfunding,” and investor types “angel,” “micro,” “accelerator,” or “incubator.” In VentureXpert, we keep rounds

when the investment firm or fund type is identified as “individual,” “angel,” or “angel group.” In VentureSource, we

incorporate round types identified as “seed,” “pre-seed,” “crowd,” “angel,” or “accelerator.”

11
Offerings under Regulation D preempt state securities law. Before March 2008, Form D filings were paper based.

We use a FOIA request to obtain nonelectronic Form D records from 1992 to 2008.

11
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
round.12 We then disambiguate and eliminate duplicates.13

This process generates 206,885 angel investments from 1988 to 2018. While not all angel

investments trigger a Form D filing or appear in the databases described above, our data set

represents one of the most comprehensive sources of angel deals available. Table II shows that

for the full sample there are on average 133.5 angel investments in a state-year.

[Table II here]

To observe the characteristics of firms receiving angel investments, we match these data

to the National Establishment Time-Series (NETS) database using firm name, address, and

founding year. We only use actual, nonimputed employment and employment growth in the

year before angel investment (Crane and Decker (2020)).14 For firms in the CVV sample, we

observe entrepreneurs’ prior founder experience at the time of investment, which proxies for

startup growth potential (Hsu (2007), Lafontaine and Shaw (2016)). Since tax credit programs

primarily target high-tech sectors, we use detailed information on industries to focus on angel

investments in sectors specifically targeted by the policy.15 In our baseline analysis, we collapse

the data to state-year panels of angel investment volume and average deal characteristics in

high-tech sectors. Summary statistics for this sample are under “Financing Outcomes” in
12
Specifically, we drop all financial issuers and pooled investment funds. Further, we match all first rounds in

Form D with VC rounds in CVV based on firm name, location, and round date within three months of each other.

We discard rounds that are identified as VC rounds.

13
We use the order of VentureXpert, VentureSource, Crunchbase, and Form D filings. We find similar results

using different orderings to disambiguate our data.

14
The NETS-matched sample period is 1993 to 2016. We start the sample in 1993 because Form D data are

incomplete in 1992. Additionally, we require up to two years of pre-investment data from NETS to measure ex-ante

growth characteristics. Given that NETS covers 1990 to 2014, our sample ends in 2016. We do not use sales from

NETS because 90% of the sales data are imputed.

15
Following the programs’ most common eligibility restrictions, we define high-tech as the following NAICS codes

corresponding to information technology, healthcare, and renewable energy: 221110-221120, 3254, 3340-3349, 3353,

3391, 4234, 5112, 5161, 5171-5174, 5179, 5181, 5182, 5414-5417, and 6200-6239. When these NAICS codes are not

available, we map them into comparable industry classifications.

12
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table II. Our investment analysis shows that the main results are similar in the full sample

and the NETS-matched sample, and then focuses on the NETS-matched sample to study

heterogeneity based on the firm characteristics that it provides.

Finally, we collect data from AngelList to study the effect of ATCs on investor composition.

While AngelList is largely self-reported, it represents the most comprehensive data available

about the identities and locations of investors for angel investments. The drawback of AngelList

is that its coverage increases in more recent years. Summary statistics on this sample are

reported at the bottom of Table II.

B. State-Level Real Outcomes

The main goal of ATC programs is to enable new business creation and the jobs supported

by these new businesses. To evaluate whether these programs achieve their stated objectives,

we use data from the Census Business Dynamics Statistics (BDS). We construct measures of

high-tech firm entry and job creation. Specifically, we use the count of new high-tech firms

aged zero to five and jobs created at those firms.16 Since the BDS provides only coarse sector-

specific data for these state-level variables, we restrict the main analysis to the sectors most

aligned with the policy targets of NAICS 51 (Information) and 54 (Professional, Scientific, and

Technical Services), but show robustness to including additional sectors as well as to restricting

to these two sectors when studying angel investment. Table II presents summary statistics for

our main real outcomes, and Section II of the Internet Appendix provides detailed definitions

of all variables.

We employ several supplementary data sets in robustness tests. First, we use two

alternative measures of startup entry. The first is the number of new potentially high-growth
16
Using ages zero to five permits the programs to affect growth at young firms in addition to new entrants. In

Table IA.XXI, we use only age zero firms and find stronger results. We use establishments, which are the unit of

measurement in BDS, but we refer to them as “firms” because essentially all firms in our data have one establishment.

13
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
firms, measured as the number of Delaware-incorporated firms registered in the state.17 This

measure was developed by the Startup Cartography project (Fazio, Guzman, and Stern

(2019), Andrews et al. (2020)), which documents that registering as a Delaware corporation

is the single strongest predictor of a growth outcome (successful acquisition or initial public

offering (IPO)). Second, we gather data at the state-year level on new high-tech startups

from 2000 to 2019. The data are provided by Steppingblocks and based on LinkedIn.

Steppingblocks defines a startup as a firm that appears in LinkedIn for the first time in a

given year and begins with no more than 20 employees.18 We also examine innovation using

patent applications from the United States Patent and Trade Office (USPTO) and the

number of successful startup exits based on CVV data.

C. Tax Credit Microdata

We obtain data on startups receiving subsidized investment (“beneficiary companies”) for

12 states from public records or privately from state officials. Among these, we also received

identities of tax credit recipient investors for seven states. We gather data on these investors

from LinkedIn. For 10 states, we also observe companies that were certified to receive

subsidized investment, but for which no investor was awarded a tax credit. We refer to these

firms as “failed applicants.” The sample period for these data is 2005 to 2018. The data are

complete for a given program-year, although we do not observe all years for all programs.

Table IA.III, Panel A, shows the number of unique companies by state. In total, there are

1,823 beneficiary companies and 1,404 failed applicants. To obtain outcomes for the

beneficiary companies and failed applicants, we match them to two data sets. First, we

match 1,227 firms to financing data. Second, we match 1,350 startups to Steppingblocks
17
We are grateful to Jorge Guzman for providing an updated and expanded version of the data.
18
To confirm that a company is a startup, Steppingblocks checks that the company had no employees at any

time prior to the year 2000 (back to 1990). High-tech is defined as a subset of their industry classification. A list is

provided in Table IA.II.

14
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
LinkedIn data. Steppingblocks provides an employment panel based on comprehensive

LinkedIn profiles.

III. Effects of Angel Investor Tax Credits

In this section, we first explain the estimation approach for evaluating state-level effects of

ATCs (Section III.A). We then discuss results of this analysis on angel investment (Sections

III.B and III.C). Effects on real outcomes are presented in Section III.D.

A. Identification Strategy

Our empirical approach is a differences-in-differences design that exploits the staggered

introduction and expiration of 36 ATC programs in 31 states. Specifically, we estimate the

following specification:

Yst = αs + αt + β · 1(AT Cst ) + γ ′ · Xs,t−1 + εst , (1)

where 1(AT Cst ) is an indicator variable equal to one if state s has an ATC program in year

t. The dependent variable is angel investments or a real outcome. The vector Xs,t−1 contains

state-year controls.19 The specification also includes state (αs ) and time (αt ) fixed effects.

Standard errors are clustered by state (Bertrand, Duflo, and Mullainathan (2004)). The

coefficient of interest, β, captures the marginal effect of ATCs on angel investment and real

outcomes. For robustness, we exploit variation in the size of tax credits across programs by

replacing 1(AT Cst ) in equation (1) with a continuous variable, T ax credit %st , which equals

the maximum tax credit percentage available in a state-year with an ATC program, and zero
19
In particular, we include the following state-year controls, which are lagged by one year: gross state product

(GSP) growth, log income per capita, log population, maximum state personal income tax rate, and log number of

young (0 to 5 years old) high-tech establishments. We find similar results without these controls (see Section III.C).

15
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
otherwise.

A key identifying assumption for our empirical design is that, in the absence of ATCs, there

would be parallel trends in states with these programs relative to those without them. To test

for parallel trends and study the immediacy of any effects, we estimate the following dynamic

differences-in-differences specification:

3
Yst = αs +αt +δ · 1(AT Cs,≤t−4 ) +β · 1(AT Cs,t+n )+θ · 1(AT Cs,≥t+4 )+γ ′ ·Xs,t−1 +εst , (2)
X
′

n=−3

where 1(AT Cs,t+n ) are indicator variables for each year around the tax credit introduction.
The year before the start of an angel tax credit is normalized to zero. We group years that

are more than four years before or after the policy change (1(AT Cs,≤t−4 ) and 1(AT Cs,≥t+4 )).

B. Effect of Angel Tax Credits on Angel Investments

We begin by studying the effect of ATC programs on the number of angel investments

in Table III, Panel A, using equation (1). We estimate this equation using the unrestricted

sample (columns (1) and (2)) and the NETS-matched sample (columns (3) and (4)), which we

use in the subsequent angel investment analysis since it allows us to more precisely identify

targeted firms and observe firm characteristics.20 Across both samples, we show that angel tax

credit programs (1(AT Cst )) increase angel investments by 17.8% to 19.0% (columns (1) and

(3)).21 We also find that a 10-percentage-point rise in the tax credit percentage (T ax credit %st )

increases the number of angel investments by 3.5% to 5.5% (columns (2) and (4)). The dynamic

differences-in-differences estimates using equation (2) are reported in Figure 2, Panel A. The

positive effect is immediate and there are no pre-trends, consistent with the parallel trends
20
The unrestricted sample period is from 1988 to 2018. The NETS-matched sample is restricted to a shorter

period, from 1993 to 2016, due to NETS coverage.

21
When the outcome is a natural logarithm, we report the exponentiated coefficient minus one in the text. The

tables contain the raw coefficients.

16
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
assumption. In sum, these estimates indicate that ATCs lead to an economically significant

increase in angel activity.

[Table III here]

[Figure 2 here]

We confirm this result using AngelList data, which include investor identities. In Table

IA.IV, Panel A, we find that ATCs significantly increase the number of angel investments,

the number of angel-backed firms, and the number of unique angel investors by 32%, 27%,

and 32%, respectively (columns (1), (3), and (5)). The interpretations of these estimates are

similar using the tax credit percentage. These results imply that the programs induce entry

of new angel investors, rather than more deals among existing investors.

Next, we evaluate heterogeneity in Table IA.IV, Panel B. We first examine program

flexibility and expect a larger effect for more flexible programs. We define Flex to measure

the presence and strictness of the 17 restrictions in Table I.22 We find that a

one-standard-deviation increase in program flexibility leads to an additional 12.9% increase

in the quantity of angel investments (column (1)). When we use the tax credit percentage as

the treatment, we find similar and significant results (column (3)). These results support a

causal interpretation of our main findings and highlight the importance of program design.23

We next study heterogeneity in local VC availability. We construct V C supplyst as the total

VC amount (excluding angel and seed rounds identified in our main sample) divided by the

number of young firms (ages 0 to 5 years) in a state-year. We find that ATCs have a weaker
22
For each nonbinary restriction, we rank programs from least to most strict and assign the highest rank to

programs without this restriction. These rank values are normalized to the unit interval. We also construct indicator

variables for programs that do not exclude insider investors and for each of the nonrefundable, nontransferable, and

no-carry-forward restrictions. To form the Flex index, we sum these 17 variables and then standardize the index by

subtracting its mean and dividing by its standard deviation prior to interacting it with our treatment variables.
23
We also examined individual program restrictions, such as firm size, and did not find significant heterogeneity

in these requirements.

17
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
effect on angel investment volume in states with an ample supply of VC (columns (2) and

(4)). This result is consistent with angel financing and VC being substitutes (Ersahin,

Huang, and Khanna (2021), Hellmann, Schure, and Vo (2021)) and with ATC programs

being particularly effective when firms face more limited options in raising early-stage

capital. It is also consistent with the idea that ATCs may not facilitate investment in

potentially high-growth firms, which are more likely to have access to VC.

To explore this question directly, we examine the type of firms receiving additional angel

financing, focusing on measures of growth potential. We split angel investments flowing to

firms with different ex-ante characteristics around the median. In Table III, Panel B, we

show that ATC programs have an insignificant effect on the amount of capital allocated to

high-employment firms, but significantly increase the capital invested in low-employment firms

(columns (1) and (2)). The results are similar when we look at employment growth (columns

(3) and (4)). An important determinant of startup success is founders’ prior entrepreneurship

experience (Hsu (2007), Lafontaine and Shaw (2016)). We find that ATCs primarily flow to

firms founded by fewer serial entrepreneurs (columns (5) and (6)). Last, we show that ATCs

direct marginal investments mainly to older firms with above-median age at the time of angel

financing, while having no significant impact on investments in nascent firms (columns (7)

and (8)). We confirm these results by showing that the average angel-backed firm has lower

growth characteristics and fewer serial entrepreneurs after a state implements ATCs (Table

IA.IV, Panel C).

It is possible that the average decline in ex-ante growth characteristics reflects higher risk

tolerance or willingness to experiment among investors (Manso (2011), Kerr, Nanda, and

Rhodes-Kropf (2014b)). The results on age are inconsistent with this interpretation because

marginal investments did not shift to younger firms. To further assess experimentation, we

compare the distributions of angel-backed firms’ ex-ante growth characteristics in state-years

with an ATC program to state-years without an ATC program, conditional on eventually

having a program. Figure IA.1 shows that, consistent with our regression estimates, the

18
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
distribution of angel-backed firms shifts to the left towards lower growth characteristics and

exit outcomes. This shift occurs across the distribution without a change in the dispersion

of the tails. Therefore, higher risk tolerance or experimentation are unlikely to explain our

findings.

ATCs might be intended by policymakers to support firms in rural areas with relatively

lower ex-ante growth characteristics. To explore whether effects differ by geography, we

separate each state’s angel investments into those that fund firms in top Metropolitan

Statistical Areas (MSAs)—defined as having at least 90% of the state’s angel deals—and

those that fund firms outside of these hub regions.24 Table IA.IV, Panel D, shows that the

effect of ATCs on angel investments in top MSAs is similar to our baseline results (columns

(1) and (3)) and there is no effect outside of top MSAs (columns (2) and (4)). This suggests

that ATCs primarily support investment in areas that already have substantial angel activity

and do not reallocate angel deals to nonhub locations.

Overall, ATCs lead to more angel investment, with additional financing going to firms with

relatively low growth potential. This result has two important implications. First, the decline

in high-growth investments supports our empirical design. One potential concern about our

identification is that states introduce ATCs in response to a boom in local demand. Since we

find that marginal investments flow to lower-potential firms, our results are more consistent

with ATC programs shifting the supply of angel financing, rather than reflecting changes in

demand. Second, our results suggest that the increase in angel activity does not reflect funding

of new startups with high-growth potential, and is concentrated in regions that already have

substantial angel activity. This finding raises questions about whether ATCs meaningfully

impact the local entrepreneurial ecosystem, a topic that we examine further in Section III.D

below.

24
We measure a state’s angel investment in the year before ATCs were implemented for treated states and in 2005

for control states. The results are not sensitive to alternatively using two or three years before implementation.

19
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
C. Robustness of Effect on Angel Investments

We conduct several robustness tests of the effect of ATCs on angel investments. First, we

test whether the staggered nature of our differences-in-differences context biases the results by

employing both the Callaway and Sant’Anna (2021) and Sun and Abraham (2021) estimators.25

Table IA.V shows that the results are robust to using these estimators, with the magnitudes

for the NETS-matched sample (columns (2) and (4)) nearly the same as the baseline result. In

some specifications, the coefficients become slightly less precise, which is expected given that

these estimates are identified using less data.

We impose sample restrictions in Table IA.VI, Panel A. First, we limit our sample to

2001 to 2016, when our data have better coverage of angel investments. The effect on angel

investment volume in this period is similar to the main sample (column (1)). Second, we

separately estimate our results for the CVV sample (column (2)) and the Form D sample

(column (3)). We again find similar estimates.26 Third, we show that the finding is robust

to dropping angel investments from VentureXpert and VentureSource, which tend to capture

angel-backed firms that eventually received institutional capital (column (4)). Fourth, we show

that the result is similar when we exclude California and Massachusetts, the largest innovation

hubs, from the sample (column (5)). Last, we estimate our results using the same sectors

available in the BDS data for the real effects analysis. Table IA.VI, Panel B, shows that

the effects using only NAICS 51 and 54 sectors are again similar in terms of magnitude and

significance.

We employ alternative specifications in Table IA.VII. The estimates are similar without

controls (Panel A). The results are also not driven by states switching from zero to positive

investments (Panel B, columns (1) and (2)) and are robust to focusing on state-years with
25
These two papers propose alternative estimation methods to address the bias that may arise for two-way fixed

effects regression when there are treatment effect heterogeneity and dynamic treatment effects.
26
This addresses the concern that the Form D data might capture some investments by other types of investors

or that tax credits may induce some investors to file a Form D (see Section IV.A.2 below for a discussion of this

possibility).

20
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
positive investments (Panel B, columns (3) and (4)).27 The results continue to hold when we

scale the number of angel investments by the number of young firms in a state-year (Panel C,

columns (1) and (2)), and when we transform the number of angel investments using the inverse

hyperbolic sine (IHS) function, which unlike the log-transform, is defined for zero (Panel C,

columns (3) and (4)). We also show that our results are robust to using dollarized treatment

variables that incorporate program size, specifically the log of a state’s aggregate annual tax

credit cap (Panel D, columns (1) and (2)) and the log of maximum supported investment,

which is defined as the annual tax credit cap divided by the tax credit percentage (Panel D,

columns (3) and (4)).

Table IA.VII, Panel E, evaluates the effect of ATCs on angel deal size. We find that ATCs

increase the average angel round amount by 23.5% to 25.1%. However, two caveats should be

noted for these estimates. First, many angel deals do not report round amount. Second, the

round amount can include both investment by angels and co-investment by VCs in the same

round.

D. Angel Tax Credits and Real Effects

States introduce ATCs primarily to stimulate the local economy and entrepreneurial

ecosystem. This section evaluates whether ATCs achieved these real effects. After estimating

the main effects (Section D.1), we interpret the results by deriving a prior for the expected

effect and calculating the statistical power of our empirical models (Section D.2).

Additionally, we evaluate the role of program scale (Section D.3) and discuss robustness tests

(Section D.4).

27
In our sample, only 9.7% of state-years have no angel investments.

21
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
D.1. Effect of Angel Tax Credits on Real Outcomes

Since the stated goal of ATC programs is mainly to spur new firms and jobs (see Section

I), we estimate their effects on firm entry and job creation. We use data from the Census BDS

to measure the count of young (0 to 5 years old) high-tech firms and new jobs created by these

firms.28 We construct these variables for top MSAs within a state that account for at least 90%

of angel investment (“top MSAs”) or at the state level. The motivation for the former approach

is that within each state there are innovation centers where both angel investment overall and

beneficiary firms (i.e., firms supported by investors receiving tax credits) are concentrated.

Indeed, the top MSAs contain more than 80% of beneficiary firms and, as shown in Section

III.B, the effect of ATCs on investments is concentrated in these areas. Focusing on these

areas can improve precision in detecting real effects.

Table IV presents the estimates for the effect of ATCs on real outcomes from 1988 to

2018 using equation (1). Panels A and B show the results for firm entry and job creation,

respectively. In each case, the outcome is log-transformed.29 For each outcome, we report

results for counts (columns (1) and (2)) and rates (columns (3) and (4)).30 We use rates

because this measure adjusts for differences in the size of the entrepreneurial ecosystem across

states, and therefore may improve the precision of our tests. Columns with odd numbers are

based on top MSAs and those with even numbers are statewide.

[Table IV here]

Across all models in Table IV, we consistently find small estimates that are not significantly
28
The BDS data only allow us to measure the entry of establishments, rather than firms. However, 99% of

high-tech firms 0 to 5 years old are single-establishment firms in our data.

29
Since the outcomes are never zero, we do not add one before taking the log. The log makes effect sizes more

comparable across outcomes, which is particularly useful in the power analysis in Section D.2. For interpretability,

we also scale Tax Credit % in this section by the average tax credit in state-years that have programs. This average

is 35.5%.
30
Firm entry rates are calculated as establishment entry divided by the average establishments in t and t − 1

(Decker et al. (2020)). We construct job creation rates similarly.

22
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
different from zero. The confidence intervals are also economically small. For example, the

estimated effect on the count of young, high-tech firms in column (1) of Panel A is -2.0% and

the upper bound of the 95% confidence interval is 2.3%. The null effects are not driven by ATCs

reversing a pre-existing negative trend in entrepreneurial activity. The dynamic differences-in-

differences, reported in Figure 2, Panels B to E, show no pre-trends.31 The estimates remain

statistically and economically insignificant for several years following the introduction of ATCs.

These near-zero estimates and small confidence intervals could indicate null effects of ATCs

on real outcomes. Alternatively, they could reflect insufficient statistical power or programs

being too small to generate measurable effects. In the following two sections, we consider these

possibilities.

D.2. Interpretation: Statistical Power

In this section, we assess whether our tests have sufficient power to detect real effects.32 We

conduct a power analysis that provides the smallest effect that could be rejected by our tests

with reasonable certainty, which we refer to as the minimum detectable effect (MDE). The

MDE is useful in two ways. First, it provides an upper bound on the true effect of angel tax

credits, as any effect larger than the MDE should likely be detected by our tests and yield a

significant result. Second, readers or policymakers can compare the MDE with their expected

effect of ATCs based on their assumptions, which we refer to as the prior. For reference, at

the end of this section, we provide calculations of priors for the effect on real outcomes using

a range of plausible assumptions.

To calculate the MDE, we follow Black et al. (2019) in using a simulation method that

calculates how often our empirical model can detect a statistically significant effect of ATCs

on outcome Y when we induce an effect size M in the simulated data. For each effect size M ,
31
Figure IA.2 provides the plots for top MSAs.
32
Statistical power is the probability of rejecting the null hypothesis of no effect when the null is false (i.e., one

minus the probability of Type II error).

23
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
we generate 1,000 random sets of ATC programs in our data and impose a treatment effect of

M on the outcome. The power at M is the fraction of the 1,000 simulations with a positive,

statistically significant effect of the policy. Following convention, we define significance as a

p-value of less than 0.1 and show robustness to a 0.05 threshold. Finally, we identify the MDE

as the effect size that we can reject with 80% power. This power threshold is conservative and

in line with conventions in the field experiment literature.33 A more detailed explanation of

the MDE calculation is in Section IV of the Internet Appendix.

Figure 3 plots the estimated power at a wide range of effect sizes for each of the real

outcomes at the state level, providing a transparent assessment of the power of our results.34

This analysis confirms that our empirical models can detect relatively small changes in a state’s

entrepreneurial activity. For example, if a 3% effect on young, high-tech firm entry exists at

the state level, we should be able to detect it almost 100% of the time (Figure 3, Panel A).

Even for a prior of only 1.9%, we would still detect an effect at the power threshold of 80%.

More generally, these figures can be used to independently assess the ability of our tests to

detect any level of expected effect of ATCs.

[Figure 3 here]

The bottom of Table IV reports the MDEs at 80% power for our main outcomes. The

upper bound of the 95% confidence intervals for the estimates on firm entry at the MSA and

state levels are 2.3% (column (1)) and 1.5% (column (2)), respectively, which are beneath the

MDEs at 80% power. This pattern generally holds for rates (columns (3) and (4) of Panel

A) and for job creation measured using both counts and rates (Panel B). We also show that
33
Abadie (2020) highlights that when the power is above 50%, statistically insignificant effects can be more

informative than significant effects. Shapiro, Hitsch, and Tuchman (2021) assess the statistical power of their analyses

using the 50% threshold. The field experiment literature typically uses 80% as a threshold for high-powered analysis

(Chow, Wang, and Shao (2007), Sakpal (2010), Mumford (2012), Black et al. (2019), Isakov, Lo, and Montazerhodjat

(2019)).
34
Figure IA.3 repeats the plots for top MSAs. In each panel, the vertical line denotes 80% power.

24
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
our ability to detect an effect is even larger when we consider the joint power across multiple

outcomes (see Section IV of the Internet Appendix).35

To facilitate assessment of the power, we calculate priors for the expected real effects of

ATCs given their effect on angel investments, which we compare with the MDEs. While the

priors rely on assumptions about how additional angel deals translate to new firms and job

creation, they are nonetheless useful as a benchmark. For the effect on new firm count, we

construct the prior as the number of new angel-backed firms induced by ATCs as a share of

all young, high-tech firms. Since we include only a firm’s first deal in our analysis of angel

investments, we assume that the estimated effect on angel investments of 18% corresponds to

an equal number of new firms, or a one-for-one pass-through of new angel deals. We follow

a similar approach to construct the priors for rates and job creation. Table IA.XI reports

the main priors along with alternatives that relax various assumptions. A comprehensive

explanation of the prior construction is in Section V of the Internet Appendix.

Comparing the baseline prior effects in the first row of Table IA.XI with MDEs at 80%

power, we find that, for all specifications, the priors are larger than the MDEs. For example,

the prior for the count of young, high-tech firms is 5.9% in top MSAs and 3.3% statewide,

while the corresponding MDEs at 80% power are 4% and 1.9%, respectively (columns (1)

and (2) of Table IV, Panel A). As mentioned above, these priors are calculated based on

particular assumptions. We relax these assumptions in the other rows of Table IA.XI and find

qualitatively similar results, with most having power above or close to 80% at the prior.36

In sum, given the estimated increase in angel investments, our tests have sufficient power to

detect the real effects of ATCs.

35
The results are similar using a 5% significance level (Table IA.VIII), without controls (Table IA.IV), or using

the continuous treatment T ax credit %st (Table IA.X).

36
In Section V of the Internet Appendix, we discuss several implicit assumptions that could lead us to

underestimate these priors.

25
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
D.3. Interpretation: Program Sizing

This section examines whether null real effects reflect small programs. We start by studying

program heterogeneity by size in case larger programs have a significant real effect. Table

IA.XII restricts the sample of treated states to those with an above-median annual budget.37

Table IA.XIII exploits variation in the program budgets by using the annual tax credit cap in

a state-year or the maximum aggregate investment supported by the credit (i.e., annual tax

credit cap divided by tax credit percentage) as alternative treatment variables. In both tables,

we continue to find statistically and economically insignificant real effects.

We next evaluate the effect of ATCs on startups by comparing firms financed by subsidized

investors (“beneficiary companies”) to firms that were certified but failed to have an investor

receive a tax credit (“failed applicants”). This approach allows us to detect an effect at the

firm level, irrespective of the aggregate size of these programs. Failed applicants represent a

useful comparison group because they are in the same state and were interested in the tax

credit. However, failed applicants are likely to be of relatively lower quality because they either

failed to raise angel financing or applied after the state ran out of funding for the tax credits. If

there is bias in comparing these groups, it should be in the direction of beneficiary companies

performing better. Table IA.III, Panel B, provides summary statistics on beneficiary companies

and failed applicants.

We estimate the following equation:

Yi,t+k = αjt + αst + β · 1(T ax Creditit ) + θYi,t + εi,t+k , (3)

where the dependent variable Yi,t+k is the outcome for startup i in year t + k. Year t is the

year that the startup either first had an investor receive a tax credit or applied for an investor

to receive a tax credit for the first time. We define 1(T ax Creditit ) as an indicator variable

equal to one if startup i had an investor receive a tax credit in year t. The specification
37
These large programs have an average annual budget of $13.7 million and can support up to $40.3 million of

angel financing per year based on the average tax credit percentage.

26
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
includes sector-year (αjt ) and state-year (αst ) fixed effects. Standard errors are clustered by

state-year.38

Table V reports estimates of equation (3). We find that receiving subsidized angel

investment does not impact raising venture capital within two years of t (column (1)) or the

probability of a successful exit based on an IPO or acquisition (column (2)). We also

examine measures of firm-level employment using LinkedIn data from Steppingblocks. We

construct indicators for the firm having at least 25 employees (columns (3) and (4)) and

employment greater than the 75th percentile within the sample (columns (5) and (6))

measured in the second and third years after the tax credit. We find no differences in future

employment between beneficiary firms and failed applicants. Table IA.XIV shows that this

result is robust to using a matching estimator that compares beneficiary companies to similar

control firms in nearby states without tax credit programs. In Table IA.XV, we also find

similar results using NETS rather than LinkedIn. Overall, tax credits did not affect recipient

firms, which is consistent with the aggregate results and suggests that program size does not

explain the null real effects.

[Table V here]

D.4. Robustness of Effect on Real Outcomes

We conduct a wide range of robustness tests. First, we employ the Callaway and Sant’Anna

(2021) and Sun and Abraham (2021) estimators in Table IA.XVI. As for angel investments,

the results are robust, with no evidence of positive effects. For one outcome at one level of

aggregation (top MSAs), one coefficient is significant and negative. This would be expected

by chance under a null effect given the number of models.

Second, we use the continuous treatment variable, T ax credit %st , in Table IA.X and again
38
We cluster by state-year because there are limited clusters by state. The results are similar when we use other

approaches, such as robust standard errors.

27
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
find no effect of ATCs on firm entry and job creation. In Table IA.XVII, Panel A, we show

that the results are similar using levels rather than logs. Next, we assess whether ATCs

produce real effects in areas that do not typically foster entrepreneurial activity. We focus

on those regions outside of top MSAs (“nontop MSAs”) and examine the impact of ATCs on

firm entry and job creation. Table IA.XVIII shows that we continue to find no statistically

and economically significant effects in these regions. This finding suggests that ATCs do not

increase real activity outside of the top MSAs.

We consider several alternative measures of real outcomes in Table IA.XIX. We construct

similar variables at the state-year level for firm entry and job creation using LinkedIn data (see

Section II.B for details). In Panel A, we find economically small and statistically insignificant

effects of ATCs on new startups (columns (1) and (2)), new high-tech startups (columns (3)

and (4)), employment at new startups (columns (5) and (6)), and employment at new high-

tech startups (columns (7) and (8)). In Panel B, we use data from the Startup Cartography

Project on the number of high-quality startups (columns (1) and (2)) and the number of new

Delaware-incorporated firms (column (3) and (4)), which proxies for high-quality firms. We

also examine successful exits in the form of IPOs and large acquisitions (columns (5) and (6))

and the number of patent applications (columns (7) and (8)). We find no effect of ATCs on

these alternative outcomes and obtain similar results using levels rather than logs in Table

IA.XVII, Panel B.

The null effects on real outcomes persist when we use alternative sectors to define high

tech. In Table IA.XX, we use all two-digit sectors that have any four-digit subsector included

in the angel analysis.39 The coefficients are qualitatively similar to those estimated in our

main analysis, with just one model significant at the 10% level (Panel B, column (6)). This is

less than what we would expect by chance.

39
Panel A uses sectors 31-33, 51, and 54. Panel B uses sectors 22, 31-33, 42, 51, 54, and 62. The sample in Panel

B includes all of utilities (22), wholesale (42), and healthcare (62), and thus is not especially relevant to the angel

policies, but we include these models for completeness.

28
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
In sum, we do not find evidence that ATCs significantly impact state-level entrepreneurial

activity based on a variety of real outcomes relevant to policymakers’ goals of stimulating

high-growth, high-tech new firms.40 It is important to note that this does not rule out the

possibility of any effect; there may be positive effects along dimensions that we cannot measure.

However, null effects are especially informative when the prior is that a policy will be effective,

and they become more informative than a significant effect when there is sufficient power

(Abadie (2020)). A positive prior is reasonable since the literature shows that other tax

credits have positive effects (Cummins et al. (1994), Hall and Van Reenen (2000), Zwick and

Mahon (2017), Arefeva et al. (2020), Dechezleprêtre et al. (2020), Edwards and Todtenhaupt

(2020), Freedman, Neumark, and Khanna (2021)). These papers either study programs directly

targeting the operating firm rather than the financial intermediary, or programs targeting

investment in firms with relatively predictable cash flows. Below we present mechanisms for

our results that follow from two distinctive features of ATC programs, namely that they target

financial intermediaries and projects with fat-tailed return distributions.

IV. Mechanisms

Thus far, we have shown that despite increasing angel investments, ATCs have no

measurable real effects, a finding that does not reflect program size or limited statistical

power. In this section, we present evidence for two mechanisms. First, the increase in angel

investment is driven in part by crowding out, where additional funding displaces funding

from other sources that would have occurred in the absence of the ATCs. We document a

decline in nonangel early-stage investment after ATCs (Section A.1) and relabeling of

investment as “angel” in order to access the ATCs (Section A.2). Second, to the extent that

ATCs do increase investment, they have little impact on the professional, sophisticated
40
As DellaVigna and Linos (2020) discuss, reporting null results reduces publication bias in policy evaluation

towards effective policies.

29
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
angels who typically fund high-growth startups that could generate large benefits for the

local economy. Instead, the increase in angel investment is driven mostly by local,

inexperienced investors without entrepreneurial backgrounds (Section B.1). Based on a

survey of angel investors and a theoretical model, we argue that the nature of returns for

early-stage firms combined with the tax credit being a fixed percentage of investment can

explain the limited response from professional investors (Sections B.2 and B.3).

Taken together, these two channels can explain our main results. The crowding-out channel

suggests that the observed increase in angel investment does not translate entirely to increased

access to financing for firms. The investor heterogeneity channel explains why subsidized firms

are relatively low-growth and mature and therefore are unlikely to significantly drive aggregate

firm entry and job creation.

A. Crowding Out

A.1. Angel Tax Credit and Alternative Finance

Our firm-level analysis (see Section D.3) points in the direction of crowding out. Above,

we show that beneficiary firms (firms with investors who receive a tax credit) do not perform

better than firms with investors that applied but ultimately did not receive a tax credit. This

is consistent with crowding out because it implies that, conditional on applying, receiving

subsidized investment does not alleviate constraints—failed firms raise subsequent VC and

succeed at the same rates as beneficiary firms. This logic follows the practice of identifying

crowding out as occurring when government funds displace private capital, observable when a

subsidy program has no effect on its targeted outcome (Knight (2002), Andreoni and Payne

(2003), Howell (2017), Moretti, Steinwender, and Van Reenen (2019)).

One way crowding out could occur is if ATCs increase angel investment by displacing other

sources of early-stage investment. The tax credits might crowd out sources such as early-stage

VC and accelerator funding, for either supply- or demand-side reasons. On the supply side,

30
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
some investors may participate in both angel (including angel groups) and early-VC rounds,

leading to a substitution between the two if these investors are constrained. There may also

be competition between different early-stage investors in both financing and product markets,

such that an increase in angel investment reduces the returns to other early-stage investors.

This would be consistent with the theories of Inderst and Mueller (2009) and Khanna and

Mathews (2022), as well as the empirical evidence of substitution between angel and VC

investment in Ersahin, Huang, and Khanna (2021) and Hellmann, Schure, and Vo (2021). On

the demand side, a limited supply of projects or a limited size of each project could lead to

inelastic financing demand. Also, entrepreneurs may not want to raise more money than they

need to limit dilutution of their equity due to early-stage investment (Bergemann and Hege

(1998)).

To test for various forms of crowding out in the startup financing market, such as

between different types of investors, between angel-backed and nonangel-backed firms, and

between subsidized and unsubsidized angel-backed firms, we examine all early-stage financing

for young firms. We estimate equation (1) with measures of early-stage financing as the

outcome variables.41 We use dollar measures because we expect crowding out to manifest via

dollar rather than deal substitution since the deal types have dramatically different sizes.

That is, a dollar of angel funding would crowd out a dollar of VC funding. Table VI, Panel

A, reports the results. First, we find an insignificant, slightly negative effect on total

early-stage investment at the state-year level (column (1)).42 We further find there is a

negative effect on nonangel investment (column (2)) and an offsetting positive effect on angel

investment (column (3)). Nonangel investors are commonly early-stage VCs. As a result, the
41
We include all early-stage rounds in CVV and Form D data. Specifically, we define early-stage rounds as

the first two rounds in VentureXpert, round types “1st,” “seed,” “angel,” “crowdfunding,” and “accelerator” in

VentureSource, founding types “pre-seed,” “seed,” “grant,” “angel,” “convertible debt,” “equity crowdfunding,”

“product crowdfunding,” and “series A” in Crunchbase, and the first two rounds of financing in Form D data.
42
The pre-ATC share of angel investments among early-stage investments is substantial, at 41% on average and

34% at the median.

31
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
share of angel investment increases by 7.5 percentage points (column (4)) from a mean of

42%. This suggests that ATCs did not affect aggregate early-stage financing while angels’

share of the total increased, consistent with crowding out.

[Table VI here]

In Panel B, we examine the effect of ATCs on total early-stage financing received at the

firm level. The sample includes all firms receiving early-stage financing, which form the basis

for the state-year panel in Panel A. All columns include state, year, and age fixed effects. The

even columns are augmented with controls. Additionally, the specifications in columns (3) and

(4) are weighted by the inverse of the number of firms in a state to mitigate the influence of

hub states. Across all specifications, we find no effect of ATCs on early-stage financing for a

firm. The effects are statistically and economically small. Overall, these results suggest that

subsidized angel financing may crowd out alternative early-stage financing, limiting the degree

to which the policy increases firms’ overall access to finance.

A.2. Insider Investors and Relabeling

In addition to crowding out across investment stages, crowding out could also occur within

investors via relabeling, where investments that would have occurred regardless of ATCs are

identified as “angel investments” to obtain the subsidy. While relabeling is extremely difficult

to prove, in this section, we narrow our focus to those investors who receive tax credits and

provide evidence consistent with relabeling being important in the data.

We first examine corporate insiders, a special class of investors who are in a particularly

advantageous position to benefit from ATCs. Insiders face relatively low information or

coordination frictions when investing in their own companies and claiming tax credits.

Insiders may invest for tax arbitrage reasons (Slemrod and Yitzhaki (2002), Korinek and

Stiglitz (2009)), potentially even making “investments” that are subsequently paid out as

dividends. They may also relabel pre-existing corporate transactions as “angel investments.”

32
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Angel investment among insiders induced by the ATCs is more likely to represent crowding

out, in the sense that any new capital from insiders would likely have been deployed

regardless within the beneficiary firm. Lee and Persson (2016) also argue that insider

investment in the form of friends-and-family financing is not a perfect substitute for external

formal sources of capital, and is less likely than other sources to lead to firm growth.

We assess the prevalence of insider investors among tax credit recipients. Our data include

628 unique firms and 3,560 investors from five states.43 We identify an investor as an insider

if the person is an executive on a Form D filing, listed as an employee on LinkedIn, or shares

a last name with an executive. Further details are in Section VI of the Internet Appendix.

In Table VII, we find that 35% of firms have at least one investor who is an executive or

family member of an executive. The share is 24% or higher in all states except Kentucky. As a

benchmark, only 8% of startups in AngelList have at least one investor who is also employed at

the company in which they are investing. At the investor level, 14% of subsidized investors are

executives of the invested company or their family members. The corresponding benchmark

in AngelList is only 2%.

[Table VII here]

Beyond insiders, investors more broadly may relabel transactions that would have

happened regardless of the program as “angel investments” in order to receive the tax

credits. Such relabeling could increase the rate of Form D filings because this document can

serve as evidence that a legal equity round occurred, which is needed to access the tax

credit.44 Relabeled investments would appear in our sample as an angel investment when
43
These states are Ohio, New Jersey, Maryland, New Mexico, and Kentucky. They are reasonably representative

of states that employ ATCs, including some high-tech clusters (New Jersey and Maryland), rural areas (Kentucky and

New Mexico), and the Rust Belt (Ohio). Some states explicitly permit the investor to be employed at the company

(Table IA.I). Ohio, New Jersey, Kentucky, and Maryland do not exclude executives, but do exclude owners with a

pre-investment ownership stake above a certain threshold, ranging from 5% for Ohio to 80% for New Jersey. New

Mexico excludes executives but has no limits for owners, families, or employees.
44
While a Form D is often theoretically required to exempt an equity investment from SEC registration,

33
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
they might not have otherwise. To explore this, we compare the Form D filing rate across

beneficiary firms and matched nonbeneficiary firms that also received angel financing. We

focus on Form Ds filed within three years of the tax credit because some states have a

minimum holding period. We match each beneficiary firm with up to five similar control

firms from nearby states without ATCs through a nearest-neighbor matching procedure.45

Table VII, Panel B, reports the results. Rows 2 to 5 show that beneficiary firms and control

firms have similar ex-ante characteristics, indicating a proper matching procedure. However,

the likelihood of a beneficiary firm filing a Form D is 64.4%, while the chance of filing for

control firms is only 32%. This difference is both statistically and economically significant.

Consistent with relabeling, beneficiary firms are significantly more likely to file a Form D

than control firms, whose investors are not required to submit proof of a legal equity round.

Finally, we expect insider and nonprofessional investors to be more responsive to increased

incentives to file a Form D because they are more likely to engage in informal transactions

and may not have other financing documentation such as stock purchase and equity rights

agreements. Consistent with this view, we find that the gap in Form D filing rates between

beneficiary and control firms is much higher when the deal contains insider investors. In Table

IA.XXII, we split the sample by whether a firm has an insider investor. We find that treated

firms are 53 percentage points more likely to file Form Ds than control firms when insider
many startups do not file, often to avoid the accompanying disclosure. Ewens and Malenko (2020) show

that for more than 20% of VC-backed startups, no Form D is ever filed. Details are in Disappearing Form

D (https://techcrunch.com/2018/11/07/the-disappearing-form-d). While there could be penalties for failing

to file a Form D, they appear to be rarely enforced. Additionally, U.S. courts and the SEC have ruled

that failing to file a Form D does not cause a startup to lose its security exemption status (SEC Rules

(https://www.sec.gov/divisions/corpfin/guidance/securitiesactrules-interps.htm)). The effect of ATCs on angel

investments is similar when we restrict to deals only from CVV (Table IA.VI, Panel A).
45
We restrict control firms to be located in a different state but belong to the same Census division and the same

industry, have similar age, and have a similar amount of previous financing relative to the year of the treatment firm’s

first tax credit using nearest-neighbor matching.

34
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
investors are present, while this difference is only 30 percentage points when no investors are

insiders. This result is consistent with the marginal benefit of filing being much higher for

insiders than for professional investors when they need to qualify for tax credits.

In sum, additional angel investment following ATCs appears to reflect in part relabeling,

where informal transactions that would have happened regardless are formalized as “angel”

deals via Form D filings. This form of crowding out can help reconcile the increase in angel

investment with the null real effects. However, it likely does not explain the entire increase

in angel investment. For example, it does not explain the increase in investment amount per

deal as shown in Table IA.VII, Panel E, because it concerns the extensive-margin decision of

whether to report an investment. Additionally, Table IA.XXIII shows that the angel results

are similar in states that exclude insiders from receiving tax credits. Nevertheless, together

with the other sources of crowding out, this direct form helps explain why we would see large

increases in reported angel investment with no commensurate effects on economic activity.

B. How Investors Make Decisions

This section explores who responds to angel tax credits and then seeks to explain why. The

success of ATCs might depend on which investors take up the subsidy. A commonly cited goal

of ATCs is to attract professional angel investors who would otherwise not invest in local firms.

If the response is instead concentrated among nonprofessional investors, the effectiveness of

these programs may be limited.

B.1. Investor Heterogeneity in Tax Credit Use and Responsiveness

We first examine heterogeneity among ATC recipients and then assess how ATCs affect

investor composition. For seven states, we obtain data on the identities of subsidized

investors and connect them with LinkedIn information on investor characteristics. Table VIII

reports the statistics for the 5,637 individuals who received tax credits, which excludes a

35
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
small number of fund recipients. We find that 87% of the subsidized investors are male and

95% are white, consistent with the findings in Ewens and Townsend (2020) that angel

investors are overwhelmingly white males.46 The average age is 42 years, which is younger

than the average age of 58 among angel investors in Huang et al. (2017). Subsidized investors

also appear to be relatively nonprofessional. Just 0.7% identify on LinkedIn as professional

investors and only 6.2% have prior entrepreneurial experience. In contrast, Huang et al.

(2017) find that 55% of angels have entrepreneurial experience, and these investors tend to

finance more companies, take a more active role in their portfolio companies, and earn higher

returns. The majority of tax credit recipients in our data are corporate executives (82%).

The next-largest groups are doctors (7.3%) and lawyers (4.1%).

[Table VIII here]

Most subsidized investors are located in the same state as the tax credit program (79%).

This is partly by design as many programs restrict investors to be in-state, which may limit

the ability of programs to attract sophisticated investors. In-state investors are less likely to

come from entrepreneurial hubs, because the major hubs of California and Massachusetts do

not have tax credit programs. Overall, we find that the average angel investor who receives

tax credits is younger, more local, and less entrepreneurial than the typical angel investor.

To quantify the relative importance of different types of investors in explaining the increase

in angel investment, we use AngelList data to examine the effect of ATCs on the composition

of investors. In particular, we consider the following four characteristics of nonprofessional

investors: in-state, less than five years of investing experience, no prior successful exit, and no

prior founder experience. These measures are consistent with Huang et al. (2017), who find

that professional angels tend to have prior entrepreneurial experience and are active in making

investments. In Table IA.XXIV, we verify that these measures of nonprofessional investors are
46
We coded ethnicity or race using pictures. We also coded individuals as Hispanic who our web researchers

identified as “white” but who had names among the top 20 Hispanic names in the U.S. (see Name List

(https://names.mongabay.com/data/hispanic.html)).

36
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
negatively correlated with better startup exit outcomes.

Table IX, Panel A, reports the estimates of equation (1) using investment-level data.47 The

dependent variables are indicators for the investor in a deal having a particular characteristic.

Observations are weighted by the inverse of the number of deals in a state, which gives each

state an equal weight and accounts for the overrepresentation of hub states. In column (1), we

find that ATCs increase the likelihood of being an in-state investor by 7.5 percentage points.

This is a 15% increase relative to the sample mean in Table II. The probabilities of a deal

having investors with limited experienced, no successful exit, and no founder experience also

increase by 4.1, 7.3, and 6.9 percentage points, respectively (columns (2) to (4)). In Panel B,

we examine whether the shift to nonprofessional investors reflects variation in investor entry,

rather than reallocation across deals. Here, the dependent variables are the log number of

investors making investments in a given state-year who are in a particular category. ATCs

increase in-state angel investors by 33% and, to a lesser extent, out-of-state investors by 21%

(columns (1) and (2)). They increase inexperienced investors by 32%, but have a small and

insignificant effect on experienced investors (columns (3) and (4)). We observe a similar pattern

for exit and founder experience (columns (5) to (8)).

[Table IX here]

Overall, local, inexperienced angel investors drive the increase in angel investments

described in Section III.B, while professional, arms-length angels are relatively unresponsive

to the tax incentive. ATCs affect not only the investment decisions of existing investors, but

also who is investing, leading to a larger share of nonprofessional investors. This shift helps

explain why marginal investments flow to lower-growth firms. If nonprofessional investors

have less access to high-quality deals or lower screening ability, they may invest in projects
47
The sample starts in 2003, when AngelList data began to have reasonable coverage. We find similar results when

we restrict the sample to start in 2010 to mitigate a potential concern about backfilled data. We use investment-level,

rather than investor-level, data because investor characteristics are defined relative to the location and timing of a

particular deal.

37
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
that have a limited impact on firm and local economic growth, helping explain the null real

effects. Nonprofessional investors may also be more likely to invest for nonpecuniary reasons

(Huang et al. (2017)) or may have close connections with the firm, making them better

positioned to utilize ATCs to minimize their tax obligations.

B.2. Survey of Angel Investors

To understand how different investors make decisions, we conduct a large-scale survey

of investors. The objective of the survey is threefold. First, it validates whether and how

ATCs affect investment decisions in practice. Second, it explores how these effects differ

across professional and nonprofessional investors. Lastly, it sheds light on why professional

investors do not respond to ATCs. We contribute to the literature using surveys to study

management practices (Bloom and Van Reenen (2010)), institutional investors (McCahery,

Sautner, and Starks (2016)), venture capitalists (Gompers et al. (2020)), and private equity

investors (Gompers, Kaplan, and Mukharlyamov (2016), Bernstein, Lerner, and Mezzanotti

(2019)). To the best of our knowledge, this survey is the first to elicit novel information about

investment approaches among a wide swathe of angel investors.

We develop the sample of investors to survey from two sources described in Section II: state-

provided lists of ATC recipients and all investors on AngelList as of early 2020 who made at

least one investment. We sent each investor an email containing a personalized survey link.

This email and the complete survey are reproduced in Section VII of the Internet Appendix.

In total, we emailed just over 12,000 individuals and obtained 1,411 responses, of which 1,384

are complete, representing a response rate of 11.6%, which is in line with other recent investor

surveys.48 Among respondents, about 11% are from the state ATC recipient data and the
48
We obtained approval from the NYU IRB for this survey. Twenty-seven responses are either incomplete or

cannot be matched back to our investor data due to response from a different email address. Our response rate is

in line with previous literature conducting other large-scale surveys. Gompers et al. (2020) survey VC investors and

obtain a response rate of 8.3%, Bernstein, Lerner, and Mezzanotti (2019) obtain a response rate of 10.3% from private

38
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
remainder are from AngelList. Details on respondents and selection are in Table IA.XXV.49

The survey yields four central insights. First, investors report that they do not consider

ATCs to be important when evaluating investments. Figure 4, Panel A, provides responses

about the importance of nine factors (randomly sorted for each investor). ATCs are not at

all important for 51% of respondents, and are very or extremely important for only 7%. This

contrasts starkly with the other eight factors. For example, 97% rate the management team

as very or extremely important, and 0% rate the team as not at all important, consistent with

Bernstein, Korteweg, and Laws (2017). Only 2% rate valuation and gut reaction as not at all

important, while over 50% rate these factors as very or extremely important.

[Figure 4 here]

Second, professional investors find ATCs less useful than other investors and tax credit

recipients, who are relatively less professional (see Section B.1). The top figure of Figure 4,

Panel B, validates the survey by showing that 76% of tax credit recipients view ATCs as at

least slightly important, compared to 49% of all respondents. Among respondents who identify

as professional investors, 64% rate ATCs as not at all important. For investors in the top decile

by number of deals, 71% rate credits as not at all important. We also estimate the relationship

between the importance of ATCs for an investor and the probability that she is a professional

investor. Table X, Panel A, finds a significant negative association between how important

investors rate ATCs and a variety of proxies for investor sophistication and experience (columns
equity investors, Graham and Harvey (2001) obtain a response rate of 8.9% from chief financial officers, and Da Rin

and Phalippou (2017) obtain a response rate of 14.4% from private equity LPs. Our absolute number of responses is

also high relative to other surveys of private equity investors. For example, Gompers, Kaplan, and Mukharlyamov

(2016) survey 79 buyout investors and Gompers et al. (2020) survey 885 VC investors.
49
In Table IA.XXV, Panel C, we find no evidence of selection on key variables related to ATCs, including residing

in a state with an ATC or living in the hub states of California and Massachusetts. However, investors with more

deals are more likely to respond and investors who are company insiders are less likely to respond. In addition,

ATC recipients are less likely to respond. While these relationships are not large in magnitude, they point towards

respondents being somewhat more experienced investors.

39
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
(1) to (3)). For example, being a professional investor reduces ATC importance by 0.38, which

is a 21% decrease relative to the sample mean. This finding provides further evidence that

professional, arms-length angels are relatively unresponsive to the tax incentive.

[Table X here]

Third, we explore why angels do not view ATCs as important. We ask investors who rate

ATCs as unimportant to select one of five options to explain their answer. The majority (57%)

report that ATCs are unimportant because they invest based on whether the startup has the

potential to be a home run (Figure 4, Panel C). We refer to this as the “Home Run” approach,

which characterizes investing in potentially high-growth, early-stage companies. Responses

to the open-ended question are consistent with this view. For example, respondents wrote

that “If the deal is bad a tax credit will not make it good” and “If I believe in the business

model/technology then a tax credit is largely irrelevant. Conversely, if I don’t believe in the

model then the tax credit is also irrelevant.” This approach does not imply that investors leave

money on the table, but rather that ATCs do not change their selection of startups ex-ante.

We formalize why professional investors may follow this investing approach in Section B.3.

In Table X, Panel A, we also see that a focus on financial metrics – the opposite of the “Home

Run” approach – predicts ATC importance (column (4)). In Panel B, we correlate reasons

for ATC unimportance with the investor’s deal volume. More professional investors with

above-median deal volume are more likely to cite the “Home Run” approach and coordination

frictions as reasons for ATCs being unimportant.

Fourth, the survey highlights frictions that could help explain our results, beyond

investment styles. Specifically, administrative costs, coordination frictions with startups, and

lack of information about the ATCs appear to play a role in reducing the use of ATCs among

arms-length, professional investors. Of the investors rating ATCs as unimportant, 11%

report that the reason is coordination costs (Figure 4, Panel C). Coordination costs are likely

to be higher for professional arm-length investors as they typically do not have close ties with

the startups before investing, face a fast-paced deal cycle, or have higher opportunity costs of

40
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
their time. Consistent with this, we find that professional investors are more likely to report

coordination frictions (Table X, Panel B, column (2)).50

In sum, tax credits are not important for our sample of investors, especially for professional

investors, and this unimportance appears to reflect a “Home Run” investing strategy. This

does not imply that investors leave money on the table. For instance, investors using a “Home

Run” approach may take up the tax credit ex-post if the coordination or administrative costs

are not too high, even if the credit does not change their selection of startups ex-ante.

B.3. Stylized Model

Professional investors appear to be less responsive to tax credits than nonprofessional

investors based on the investor heterogeneity and survey results. Further, survey respondents

suggest that a “Home Run” investing approach might explain why professional investors do

not respond to the tax credits. We use a simple model to explore why this might occur. The

model seeks to understand the role of return distributions, although it does not fully

characterize how ATCs affect investment decisions. The full model and proofs are in Section

VIII of the Internet Appendix. A brief summary is presented below.

We study an investor who decides to invest in a startup if and only if the expected return

is higher than a hurdle rate, which captures the opportunity cost of other projects and any

coordination or effort cost. We follow Othman (2019) and Malenko et al. (2020) by assuming

that startup investment returns follow a Pareto distribution, with shape parameter αj . Our

choice of the Pareto distribution is motivated by the well-documented fact that startup returns

exhibit a heavy right tail and extreme skewness (Scherer and Harhoff (2000), Kerr, Nanda,
50
We also ask whether an investor used ATCs and, if not, why. Figure 4, Panel D, shows that 15% do not use ATCs

because of coordination costs, and 60% are unaware the programs exist. Indeed, even among investors whose states

have a program, 19% report that ATCs are not available and 60% do not know about their availability, indicating

information barriers.

41
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
and Rhodes-Kropf (2014b), Ewens, Nanda, and Rhodes-Kropf (2018)).51 We assume that αj

is an investor-specific parameter governing the pool of projects that the investor can access.52

Sophisticated, professional investors have access to projects with higher expected returns

and higher uncertainty, which means a lower αj . A low αj captures the “Home Run” investing

approach. These opportunities might be available to professional investors focusing on early-

stage, high-growth, and high-risk startups with very fat-tailed return distributions. A high αj

characterizes firms with more traditional business models that have lower risk profiles, which

tend to be accessed by nonprofessional investors. The model also allows firms to differ in terms

of observable quality.

In this setting, we study how an investor tax credit affects the ex-ante probability of

investing in a startup and how sensitivity to the tax credit differs across investor types (i.e.,

αj ). Intuitively, the tax credit increases the expected return to the investor, raising the chances

of reaching her hurdle rate. The key insight of the model is that this effect declines as αj

decreases and the right tail of the distribution grows fatter. As αj decreases, the expected

return increases and the marginal benefit of the tax credit decreases, leading to lower sensitivity.

This follows from the fact that the tax credit subsidy does not vary with investment returns.

Instead, it is fixed at the time of investment. For example, the tax credit is the same if it

supports an investment in a new coffee shop or a new high-tech company with high-growth

potential. Given the different return profiles of the two firms, the ATC is less likely to be

pivotal (i.e., change the decision to invest) for investing in the tech company than investing in
51
Hall and Woodward (2010) and Kerr, Nanda, and Rhodes-Kropf (2014b) document that most startups fail

completely while a few generate enormous returns. Malenko et al. (2020) further show that such skewness is much

higher for seed-stage investments than for later-stage ones. Practitioners also embrace the idea that early-stage

startup returns follow a power law (Pareto) distribution (Thiel and Masters (2014) and see Power Law and the

Long Tail (https://avc.com/2015/11/power-law-and-the-long-tail)). Additionally, the Pareto distribution allows us

to capture limited liability facing investors as the distribution is bounded below.

52
We assume that projects have bounded expected returns with αj > 1. We consider the extreme case of αj ≤ 1

in Section VIII of the Internet Appendix.

42
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
the coffee shop.

This result is visualized in Figure 5, which plots the investment probability as a function of

the tax credit rate and shows how the relationship depends on αj . The chances of investment

increase in the tax credit rate, but this relationship is flatter when αj is smaller, indicating

lower sensitivity. As αj converges to one, the slope converges to zero.53 This stylized model

helps us interpret the survey finding that ATCs do not impact the decisions of investors

following a “Home Run” approach. Conditional on access to projects with fat-tailed outcome

distributions, tax credits are not useful at the margin because they represent fixed subsidies.

When investing in more traditional firms with limited upside potential but also limited risk,

the tax credits are more effective. This helps explain the larger sensitivity for nonprofessional

investors documented in both the survey and our investor composition analysis.

[Figure 5 here]

More broadly, the model highlights that fat-tailed return distributions have important

implications for the role of entry prices and thus for the effectiveness of early-stage investor

subsidies. When the potential gains are very high (αj is low), the entry price for early-stage

investments is largely irrelevant for the extensive-margin decision to invest in a startup.54 The

predictions above align well with observations from practitioners such as Charles Birnbaum,

a partner at Bessemer Venture Partners, who noted that “your entry price matters when you

think there’s a ceiling [on the startup’s exit valuation].”55

53
Section VIII of the Internet Appendix provides a numerical example of this relationship based on a calibrated

value of αj .
54
It is important to note that our analysis is positive as opposed to normative. We do not imply that angel

investors should assume that their returns follow the distribution described above, and therefore largely ignore the

entry price. Also, the model does not imply that the tax credit is always an ineffective policy tool; conversely, it may

increase investments in subsistence-type companies. A key feature of the tax credit is that the size of the subsidy

does not scale up with the quality of the company. As we show in Section VIII of the Internet Appendix, other

policies such as capital gains exemptions may work better in this setting.
55
See Birnbaum Podcast (https://podcasts.apple.com/us/podcast/bessemer-venture-partners-charles-birnbaum-

43
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
V. Conclusion

There is substantial government interest in supporting startups, and investor incentives

are a particularly appealing option. As the global angel market rapidly expands, more

jurisdictions are proposing implementing these programs. For example, Senator Christopher

Murphy recently proposed legislation to establish a federal angel investor tax credit in the

U.S.56 Yet there has been no systematic evidence on the effectiveness of these policies.

This paper offers the first analysis of U.S. angel tax credits. We find that angel tax credits

significantly increase state-level angel investment. This increase is connected to a decline in

the ex-ante growth characteristics of marginal startups funded by angels. Yet when we turn

to real outcomes that policymakers focus on, such as new business creation or young firm

employment, we find no significant impacts. The lack of any real effect is not driven by these

programs being too small or limited statistical power. Rather, two mechanisms together help

to explain these seemingly puzzling results. First, investment that increases due to the policy,

generating the positive causal effects that we observe on angel investment, partially crowds

out investment that would have happened in the absence of the policy. Second, the types of

investors who respond tend to be local and nonprofessional, and the additional companies that

they finance tend to be low-growth and relatively old, muting potential effects on firm entry

and job creation.

We next ask why professional investors who tend to fund high-risk, high-growth startups

do not respond to the angel tax credits. A survey documents that investors view tax credits as

unimportant to their investment decisions. The more professional and experienced an investor

is, the higher the chance she will find them unimportant. The survey also suggests that

professional investors find the ATCs unimportant because they take a “Home Run” investment

approach. Using a stylized model, we show that the low sensitivity of professional investors

to the tax credit may stem from the fat-tailed distribution of early-stage investment returns.
fintech/id1042827113?i=1000514179070).
56
See Senate Bill (https://www.congress.gov/bill/114th-congress/senate-bill/973).

44
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
These findings shed new light on how angel investors make decisions. They are likely related

to the importance of nonmonetary factors such as certification and advice that angel investors

provide, as opposed to capital constraints being the primary scarce factor. This is a promising

topic for future research.

Our findings raise questions about the ability of investor tax credits to stimulate

entrepreneurial activity. Angel tax credits, relative to direct programs such as grants, have

the attractive feature of being more market-based tools that do not require the government

to identify which companies deserve subsidies. However, this flexibility presents problems of

its own as the targeted investors may not be sensitive to the policy. Our results highlight the

importance of program design and investor type. Targeting investors who can identify and

monitor high-growth startups is an important element of government programs focused on

subsidizing capital for high-growth entrepreneurship.

Finally, angel tax credits likely represent a regressive tax policy. The credits accrue to

rich people given the income and wealth requirements to become an accredited angel investor.

If the credits had large job creation effects, there might be an argument for “trickle down”

benefits to poorer people. However, since we find no effects on job creation and instead find

evidence of crowding out, it seems likely that the programs lead to transfers from less wealthy

to more wealthy taxpayers, creating potentially large opportunity costs from alternative uses

of these public funds.

Initial Submission: August 21, 2020; Accepted: September 14, 2022

Editors: Stefan Nagel, Philip Bond, Amit Seru, and Wei Xiong

45
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
REFERENCES

Abadie, Alberto, 2020, Statistical nonsignificance in empirical economics, American Economic

Review: Insights 2, 193–208.

Acs, Zoltan, Thomas Åstebro, David Audretsch, and David T. Robinson, 2016, Public policy
to promote entrepreneurship: A call to arms, Small Business Economics 47, 35–51.

Andreoni, James, and A. Abigail Payne, 2003, Do government grants to private charities crowd
out giving or fund-raising?, American Economic Review 93, 792–812.

Andrews, Raymond J., Catherine Fazio, Jorge Guzman, Yupeng Liu, and Scott Stern, 2020,
The startup cartography project, Working paper, Columbia University.

Arefeva, Alina, Morris A. Davis, Andra C. Ghent, and Minseon Park, 2020, Who benefits
from place-based policies? Job growth from opportunity zones, Working paper, Univeristy
of Wisconsin-Madison.

Bergemann, Dirk, and Ulrich Hege, 1998, Venture capital financing, moral hazard, and
learning, Journal of Banking & Finance 22, 703–735.

Bernstein, Shai, Arthur Korteweg, and Kevin Laws, 2017, Attracting early-stage investors:
Evidence from a randomized field experiment, Journal of Finance 72, 509–538.

Bernstein, Shai, Josh Lerner, and Filippo Mezzanotti, 2019, Private equity and financial
fragility during the crisis, Review of Financial Studies 32, 1309–1373.

Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan, 2004, How much should we trust
differences-in-differences estimates?, Quarterly Journal of Economics 119, 249–275.

Bhargava, Saurabh, and Dayanand Manoli, 2015, Psychological frictions and the incomplete
take-up of social benefits: Evidence from an IRS field experiment, American Economic
Review 105, 3489–3529.

Black, Bernard, Alex Hollingsworth, Leticia Nunes, and Kosali Simon, 2019, Simulated power
analyses for observational studies: An application to the Affordable Care Act Medicaid
expansion, Working paper, Northwestern University.

Bloom, Nicholas, and John Van Reenen, 2010, New approaches to surveying organizations,
American Economic Review 100, 105–109.

46
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Brander, James A., Qianqian Du, and Thomas Hellmann, 2015, The effects of government-
sponsored venture capital: International evidence, Review of Finance 19, 571–618.

Bronzini, Raffaello, and Eleonora Iachini, 2014, Are incentives for R&D effective? Evidence
from a regression discontinuity approach, American Economic Journal: Economic Policy 6,
100–134.

Callaway, Brantly, and Pedro H.C. Sant’Anna, 2021, Difference-in-differences with multiple
time periods, Journal of Econometrics 225, 200–230.

Chetty, Raj, and Amy Finkelstein, 2020, Public economics, NBER Reporter 1, 1–6.

Chow, Shein-Chung, Hansheng Wang, and Jun Shao, 2007, Sample size calculations in clinical
research (Chapman and Hall/CRC, Boca Raton, FL).

Cohen, Susan, Daniel C. Fehder, Yael V. Hochberg, and Fiona Murray, 2019, The design of
startup accelerators, Research Policy 48, 1781–1797.

Crane, Leland, and Ryan Decker, 2020, Research with private sector business microdata: The
case of NETS/D&B, Working paper, Federal Reserve Board.

Cumming, Douglas J., and Jeffrey G. MacIntosh, 2006, Crowding out private equity: Canadian
evidence, Journal of Business Venturing 21, 569–609.

Cummins, Jason G., Kevin A. Hassett, R. Glenn Hubbard, Robert E. Hall, and Ricardo J.
Caballero, 1994, A reconsideration of investment behavior using tax reforms as natural
experiments, Brookings Papers on Economic Activity 1994, 1–74.

Curtis, E. Mark, and Ryan Decker, 2018, Entrepreneurship and state taxation, Working paper,
Wake Forest University.

Da Rin, Marco, and Ludovic Phalippou, 2017, The importance of size in private equity:
Evidence from a survey of limited partners, Journal of Financial Intermediation 31, 64–
76.

Dai, Zhonglan, Edward Maydew, Douglas A. Shackelford, and Harold H. Zhang, 2008, Capital
gains taxes and asset prices: Capitalization or lock-in?, Journal of Finance 63, 709–742.

Davis, Jesse, Adair Morse, and Xinxin Wang, 2020, The leveraging of Silicon Valley, Working
paper, Univeristy of North Carolina.

47
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Dechezleprêtre, Antoine, Elias Einiö, Ralf Martin, Kieu-Trang Nguyen, and John Van Reenen,
2020, Do tax incentives for research increase firm innovation? An RD design for R&D,
Working paper, London School of Economics.

Decker, Ryan A., John Haltiwanger, Ron S. Jarmin, and Javier Miranda, 2020, Changing
business dynamism and productivity: Shocks versus responsiveness, American Economic
Review 110, 3952–3990.

DellaVigna, Stefano, and Elizabeth Linos, 2020, RCTs to scale: Comprehensive evidence from
two nudge units, Working paper, University of California-Berkeley.

Denes, Matthew, 2019, When do firms risk shift? Evidence from venture capital, Working
paper, Carnegie Mellon University.

Denes, Matthew, Xinxin Wang, and Ting Xu, 2019, Financing entrepreneurship: Tax
incentives for early-stage investors, Working paper, Carnegie Mellon University.

Deshpande, Manasi, and Yue Li, 2019, Who is screened out? Application costs and the
targeting of disability programs, American Economic Journal: Economic Policy 11, 213–
248.

Edwards, Alexander, and Maximilian Todtenhaupt, 2020, Capital gains taxation and funding
for start-ups, Journal of Financial Economics 138, 549–571.

Ersahin, Nuri, Ruidi Huang, and Naveen Khanna, 2021, Competition, reputation, and venture
capital investment, Working paper, Michigan State University.

Evans, David S., and Boyan Jovanovic, 1989, An estimated model of entrepreneurial choice
under liquidity constraints, Journal of Political Economy 97, 808–827.

Ewens, Michael, and Joan Farre-Mensa, 2020, The deregulation of the private equity markets
and the decline in IPOs, Review of Financial Studies 33, 5463–5509.

Ewens, Michael, and Nadya Malenko, 2020, Board dynamics over the startup life cycle,
Working paper, California Institute of Technology.

Ewens, Michael, Ramana Nanda, and Matthew Rhodes-Kropf, 2018, Cost of experimentation
and the evolution of venture capital, Journal of Financial Economics 128, 422–442.

Ewens, Michael, and Richard R. Townsend, 2020, Are early stage investors biased against
women?, Journal of Financial Economics 135, 653–677.

48
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fazio, Catherine, Jorge Guzman, and Scott Stern, 2019, The impact of state-level R&D tax
credits on the quantity and quality of entrepreneurship, Working paper, Columbia University.

Fehder, Daniel C., and Yael V. Hochberg, 2019, Spillover effects of startup accelerator
programs: Evidence from venture-backed startup activity, Working paper, University of
Southern California.

Freedman, Matthew, David Neumark, and Shantanu Khanna, 2021, Combining rules and
discretion in economic development policy: Evidence on the impacts of the California
Competes Tax Credit, Working paper, University of California-Irvine.

Gompers, Paul, Steven N. Kaplan, and Vladimir Mukharlyamov, 2016, What do private equity
firms say they do?, Journal of Financial Economics 121, 449–476.

Gompers, Paul A., Will Gornall, Steven N. Kaplan, and Ilya A. Strebulaev, 2020, How do
venture capitalists make decisions?, Journal of Financial Economics 135, 169–190.

Gonzalez-Uribe, Juanita, and Daniel Paravisini, 2019, How sensitive is young firm investment
to the cost of outside equity?, Working paper, London School of Economics.

Graham, John R., and Campbell R. Harvey, 2001, The theory and practice of corporate finance:
Evidence from the field, Journal of Financial Economics 60, 187–243.

Hall, Bronwyn, and John Van Reenen, 2000, How effective are fiscal incentives for R&D? A
review of the evidence, Research Policy 29, 449–469.

Hall, Robert E., and Susan E. Woodward, 2010, The burden of the nondiversifiable risk of
entrepreneurship, American Economic Review 100, 1163–1194.

Hellmann, Thomas, Paul Schure, and Dan H. Vo, 2021, Angels and venture capitalists:
Substitutes or complements?, Journal of Financial Economics 141, 454–478.

Hellmann, Thomas, and Veikko Thiele, 2015, Friends or foes? The interrelationship between
angel and venture capital markets, Journal of Financial Economics 115, 639–653.

Hochberg, Yael V., Carlos J. Serrano, and Rosemarie H. Ziedonis, 2018, Patent collateral,
investor commitment, and the market for venture lending, Journal of Financial Economics
130, 74–94.

Howell, Sabrina T., 2017, Financing innovation: Evidence from R&D grants, American
Economic Review 107, 1136–1164.

49
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Howell, Sabrina T., 2020, Reducing information frictions in venture capital: The role of new
venture competitions, Journal of Financial Economics 136, 676–694.

Howell, Sabrina T., and J. David Brown, 2019, Do cash windfalls affect wages? Evidence from
R&D grants to small firms, Working paper, New York University.

Howell, Sabrina T., and Filippo Mezzanotti, 2019, Financing entrepreneurship through the tax
code: Angel investor tax credits, Working paper, New York University.

Hsu, David H., 2007, Experienced entrepreneurial founders, organizational capital, and venture
capital funding, Research Policy 36, 722–741.

Huang, Laura, Andy Wu, Min Ju Lee, Jiayi Bao, Marianne Hudson, and Elaine Bolle, 2017,
The American angel, Working paper, Angel Capital Association.

Inderst, Roman, and Holger M. Mueller, 2009, Early-stage financing and firm growth in new
industries, Journal of Financial Economics 93, 276–291.

Isakov, Leah, Andrew W. Lo, and Vahid Montazerhodjat, 2019, Is the FDA too conservative or
too aggressive?: A Bayesian decision analysis of clinical trial design, Journal of Econometrics
211, 117–136.

Ivković, Zoran, James Poterba, and Scott Weisbenner, 2005, Tax-motivated trading by
individual investors, American Economic Review 95, 1605–1630.

Kerr, William R., Josh Lerner, and Antoinette Schoar, 2014a, The consequences of
entrepreneurial finance: Evidence from angel financings, Review of Financial Studies 27,
20–55.

Kerr, William R., Ramana Nanda, and Matthew Rhodes-Kropf, 2014b, Entrepreneurship as
experimentation, Journal of Economic Perspectives 28, 25–48.

Khanna, Naveen, and Richmond D. Mathews, 2022, Skill versus reliability in venture capital,
Journal of Financial Economics 145, 41–63.

Knight, Brian, 2002, Endogenous federal grants and crowd-out of state government spending:
Theory and evidence from the Federal Highway Aid Program, American Economic Review
92, 71–92.

Korinek, Anton, and Joseph E. Stiglitz, 2009, Dividend taxation and intertemporal tax
arbitrage, Journal of Public Economics 93, 142–159.

50
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Lach, Saul, 2002, Do R&D subsidies stimulate or displace private R&D? Evidence from Israel,
Journal of Industrial Economics 50, 369–390.

Lafontaine, Francine, and Kathryn Shaw, 2016, Serial entrepreneurship: Learning by doing?,
Journal of Labor Economics 34, S217–S254.

Lee, Samuel, and Petra Persson, 2016, Financing from family and friends, Review of Financial
Studies 29, 2341–2386.

Lerner, Josh, 2009, Boulevard of Broken Dreams: Why Public Efforts to Boost
Entrepreneurship and Venture Capital Have Failed–and What to Do About It (Princeton
University Press, Princeton, NJ).

Lerner, Josh, 2020, Government incentives for entrepreneurship, Working paper, National
Bureau of Economic Research.

Lerner, Josh, Antoinette Schoar, Stanislav Sokolinski, and Karen Wilson, 2018, The
globalization of angel investments: Evidence across countries, Journal of Financial
Economics 127, 1–20.

Lindsey, Laura A., and Luke C.D. Stein, 2020, Angels, entrepreneurship, and employment
dynamics: Evidence from investor accreditation rules, Working paper, Arizona State
University.

Malenko, Andrey, Ramana Nanda, Matthew Rhodes-Kropf, and Savitar Sundaresan, 2020,
Investment committee voting and the financing of innovation, Working paper, University of
Michigan.

Manso, Gustavo, 2011, Motivating innovation, Journal of Finance 66, 1823–1860.

McCahery, Joseph A., Zacharias Sautner, and Laura T. Starks, 2016, Behind the scenes: The
corporate governance preferences of institutional investors, Journal of Finance 71, 2905–
2932.

McKenzie, David, 2017, Identifying and spurring high-growth entrepreneurship: Experimental

evidence from a business plan competition, American Economic Review 107, 2278–2307.

Moretti, Enrico, Claudia Steinwender, and John Van Reenen, 2019, The intellectual spoils of
war? Defense R&D, productivity and international spillovers, Working paper, University of
California-Berkeley.

51
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Mumford, Jeanette A., 2012, A power calculation guide for fMRI studies, Social Cognitive and
Affective Neuroscience 7, 738–742.

Othman, Abraham, 2019, Startup growth and venture returns, Working paper, AngelList.

Poterba, James M., 1989, Capital gains tax policy toward entrepreneurship, National Tax
Journal 42, 375–389.

Robb, Alicia, and David Robinson, 2012, The capital structure decisions of startup firms,
Review of Financial Studies 1, 1–27.

Sakpal, Tushar, 2010, Sample size estimation in clinical trial, Perspectives in Clinical Research
1, 67–67.

Scherer, Frederic M., and Dietmar Harhoff, 2000, Technology policy for a world of skew-
distributed outcomes, Research Policy 29, 559–566.

Shane, Scott, 2009, Fool’s Gold?: The Truth behind Angel Investing in America (Oxford
University Press, New York, NY).

Shapiro, Bradley, Günter J. Hitsch, and Anna Tuchman, 2021, TV advertising effectiveness
and profitability: Generalizable results from 288 brands, Econometrica 89, 1855–1879.

Slemrod, Joel, and Shlomo Yitzhaki, 2002, Tax avoidance, evasion, and administration, in
Handbook of Public Economics, volume 3, 1423–1470 (Elsevier).

Sun, Liyang, and Sarah Abraham, 2021, Estimating dynamic treatment effects in event studies
with heterogeneous treatment effects, Journal of Econometrics 225, 175–199.

Thiel, Peter A., and Blake Masters, 2014, Zero to One: Notes on Startups, or How to Build
the Future (Currency, New York, NY).

Xu, Ting, 2019, Learning from the crowd: The feedback value of crowdfunding, Working paper,
University of Toronto.

Yagan, Danny, 2015, Capital tax reform and the real economy: The effects of the 2003 dividend
tax cut, American Economic Review 105, 3531–3563.

Zwick, Eric, and James Mahon, 2017, Tax policy and heterogeneous investment behavior,
American Economic Review 107, 217–248.

52
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
.
Panel A. States with Angel Tax Credit Programs

Panel B. Timing of State Angel Tax Credit Programs

Figure 1. State angel tax credit programs. Panel A provides a map of states that adopted angel tax credit
programs from 1988 to 2018. Blue shading indicates the tax credit percentage, with darker shades representing larger
tax credits. Slanted lines denote states with no state income tax. Panel B shows the introduction and termination
of each program in our sample, starting with the earliest program and ending with the most recent one.

53
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Panel B. Entry by Young
Panel A. Number of Angel Investments High-Tech Firms

Panel C. Entry Rate of Young Panel D. Jobs Created by Young

High-Tech Firms High-Tech Firms

Panel E. Job Creation Rate by Young

High-Tech Firms

Figure 2. Dynamics effects of angel tax credit introduction. This figure shows the dynamic effects of
introducing angel tax credits using equation (2). Dots denote the point estimates of dynamic differences-in-differences
coefficients and bars indicate 95% confidence intervals. The year before policy introduction is normalized to zero.
Panel A shows the number of angel investments, Panel B examines the entry by young (age 0-5) high-tech firms in a
state, Panel C shows the entry rate among young high-tech firms, Panel D examines the number of new jobs created
by young high-tech firms, and Panel E looks at the job creation rate among young high-tech firms. All outcome
variables are log transformed and are defined at the state-year level. The sample period is 1988 to 2018. Detailed
variable definitions are in Section II of the Internet Appendix. Standard errors are clustered at the state level.

54
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Panel A. Entry by Young High-Tech Firms Panel B. Entry Rate of Young High-Tech Firms

Panel D. Job Creation Rate by Young High-Tech

Panel C. Job Creation by Young High-Tech Firms Firms

Figure 3. Power and prior for the effect of angel tax credits on real outcomes. This figure shows the
relationship between the estimated power of our differences-in-differences model and the minimum detectable effect
(MDE) for the four main real outcomes considered in Table IV at the statewide level. Power is computed using the
simulation method detailed in Section IV of the Internet Appendix and represents the likelihood that our test detects
a significant effect of angel tax credits (at the 10% significance level) when we induce an effect equal to MDE in the
data. Each dot represents the MDE for a given power. Solid horizontal lines denote our prior effect (see Section V
of the Internet Appendix for the calculation) and dotted lines denote 80% power.

55
15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Panel A. Distribution of Responses to Factor Importance Question

56
15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Panel B. Distribution of Responses to Importance of Angel Tax Credits by Respondent Type

57
15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Panel C. Why Angel Tax Credits Are Unimportant (If Rated as Unimportant)

58
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Panel D. Distribution of Responses to Why Have Not Used Angel Tax Credits

Figure 4. Survey results. The graphs in Panel A show the distribution of responses to question 1 in the survey
for each of the nine investment factors. Respondents could only choose one importance level for each factor. The
order in which the factors were presented was randomized across survey participants. N=1,364. The graphs in Panel
B show the distribution of responses to the question of whether angel tax credits are important to the decision to
invest in a startup. Each graph presents a different sample. The top graph shows the subset of respondents who
were angel tax credit recipients from our state-provided data or who reported having used an angel tax credit in
the survey (N=268). The second graph shows the subset of respondents from AngelList data who reported having
never used an angel tax credit in the survey (N=1,028). The third graph shows the subset of respondents from
AngelList data who identify as professional investors (N=241). The bottom graph shows the subset of respondents
from AngelList data whose number of deals are in the top 10% among all AngelList responders (N=84). For this
graph, no respondents answered “Very important.” Respondents could only choose one importance level. The order
in which the factors were presented was randomized across survey participants. Panel C shows the distribution of
responses to the question of why angel tax credits are unimportant (N=948) to the decision to invest in a startup.
Respondents were prompted to answer the question of why the credits are unimportant if they rated them as not
at all or slightly important. Panel D shows the distribution of responses to the question of why an investor has not
used angel tax credits, conditional on not using them (N=1,028). For this question, respondents could only choose
one option.

59
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Figure 5. Model prediction—investment probability and investor tax credit rate. This figure plots
investment probability against tax credit rate τ and shows how the relationship varies with the shape of the return
distribution α. We consider cases in which α is equal to 1, 1.5, 2, 5, and 10. A lower α represents a Pareto distribution
with a fatter tail. We assume cost of capital k = 10% and C = 1. Section VIII of the Internet Appendix details the
investment probability function and the associated parameters.

60
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table I
Summary Statistics on Angel Tax Credit Programs
This table presents the program parameters for the 36 angel tax credit programs in our sample. Column (1) reports
the percentage of programs that have a particular restriction in place. Columns (2) and (3) report the mean and
median values of these restrictions.

% with restriction Mean Median

(1) (2) (3)
Tax credit % 34% 33%

Company restrictions
Age cap 31% 7.1 6.0
Employment cap 39% 64.6 50.0
Revenue cap ($ million) 47% 5.4 5.0
Asset cap ($ million) 22% 11.5 7.5
Prior total external financing cap ($ million) 19% 5.7 4.0

Investment and investor restrictions

Minimum investment per investor ($) 36% 19,231 25,000
Minimum holding period (years) 50% 3.2 3.0
Ownership cap before investment 64% 35% 30%
Exclude owners and their families 61%
Exclude full-time employees 22%
Exclude executives and officers 33%

Tax credit restrictions

State tax credit allocation per year ($ million) 86% 9.0 5.0
Maximum tax credit per company per year ($ million) 42% 0.81 0.6
Maximum tax credit per investor per year ($ million) 78% 0.21 0.11
Non-refundable 72%
No carry forward 11%
Non-transferrable 72%

61
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table II
Summary Statistics
This table reports summary statistics for the state-year level variables used in our analyses and investor-level
characteristics. All variables are defined in Section II of the Internet Appendix.

N Mean Std. dev. p5 p50 p95

Treatment variables

1(ATC) 1,200 0.25 0.43 0.00 0.00 1.00

Tax credit % 1,200 0.09 0.18 0.00 0.00 0.50
Ln(agg. TC cap) 1,168 3.36 6.31 0.00 0.00 15.65
Ln(agg. supported investment) 1,168 3.60 6.77 0.00 0.00 16.59

Real outcomes

Entry by young HT firms (statewide) 1,550 1,475 1,994 135 833 5,706
Entry by young HT firms (top MSAs) 1,550 1,122 1,740 50 531 5,447
Jobs created by young HT firms (statewide) 1,550 11,330 17,196 810 5,831 42,277
Jobs created by young HT firms (top MSAs) 1,550 8,940 15,430 329 3,877 37,291
Entry rate of young HT firms (statewide) 1,550 0.271 0.027 0.225 0.272 0.314
Entry rate of young HT firms (top MSAs) 1,550 0.272 0.032 0.218 0.273 0.319
Jobs creation rate by young HT firms (statewide) 1,550 0.356 0.059 0.270 0.352 0.453
Jobs creation rate by young HT firms (top MSAs) 1,550 0.360 0.070 0.259 0.354 0.465

Financing outcomes

Number of angel investments (unrestricted 1,550 133.5 330.1 1.0 40.0 465.0
sample)
Number of angel investments (NETS-matched 1,200 24.2 65.9 0.0 8.0 78.0
sample)
Aggregate early-stage financing amount 1,200 1,394 5,847 0 185 5,362
Aggregate non-angel financing amount 1,200 1,058 5,399 0 87 3,861
Aggregate angel financing amount 1,200 336 1,049 0 38 1,319
Angel share among early-stage financing 1,200 0.42 0.32 0.02 0.33 1.00

Investor characteristics on AngelList

In-state 89,146 0.51 0.50 0.00 1.00 1.00

Inexperienced 89,146 0.73 0.45 0.00 1.00 1.00
Had no exit 89,146 0.41 0.49 0.00 0.00 1.00
No founder experience 89,146 0.81 0.39 0.00 1.00 1.00

62
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table III
Angel Tax Credits and Angel Investments
Panel A reports differences-in-differences estimates for the effect of angel tax credit programs on the log number
of angel investments in the high-tech sector (IT, biotech, and renewable energy). Columns (1) and (2) use the
unrestricted sample of angel deals from 1988 to 2018. Columns (3) and (4) use the sample of deals that can be
matched to NETS from 1993 to 2016. 1(AT C) is an indicator variable equal to one if a state has an angel tax credit
program in that year. Tax Credit % is a continuous variable equal to the maximum tax credit percentage available in
a state-year with an angel tax credit program and is zero in state-years without a program. Panel B splits the angel
volume in the NETS-matched sample by different pre-investment startup characteristics at the median (employment,
employment growth, fraction of serial entrepreneurs on founding team, and age). Control variables are defined in
equation (1). Each observation is a state-year. All specifications include state and year fixed effects. Standard errors
are reported in parentheses and clustered by state. ***, **, and * denote significance at the 1%, 5%, and 10% level,
respectively.

Panel A: Angel Investments

Ln(no. of angel investments)
(1) (2) (3) (4)
1(ATC) 0.164∗∗∗ 0.174∗∗
(0.058) (0.073)
Tax Credit % 0.348∗∗∗ 0.535∗∗∗
(0.128) (0.169)
Sample Unrestricted NETS-matched
State FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Observations 1,550 1,550 1,200 1,200
Adjusted R2 0.954 0.954 0.912 0.912

Panel B: Angel Investments by Ex-Ante Growth Characteristics

Employment Growth Serial Entrep. Age
High Low High Low High Low Young Old
(1) (2) (3) (4) (5) (6) (7) (8)
1(ATC) -0.001 0.249∗∗∗ 0.081 0.186∗∗ -0.003 0.186∗ 0.091 0.197∗∗∗
(0.073) (0.082) (0.065) (0.082) (0.105) (0.098) (0.066) (0.067)
State FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Controls Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1,200 1,200 1,200 1,200 1,200 1,200 1,200 1,200
Adjusted R2 0.880 0.872 0.892 0.860 0.733 0.874 0.899 0.833

63
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table IV
Angel Tax Credits and Real Effects
This table provides the differences-in-differences estimates of the effect of angel tax credit programs on firm entry and
job creation from BDS. In Panel A, the dependent variable is the log number of young, high-tech firms in columns (1)
and (2) and firm entry rate in columns (3) and (4). Young firms are defined as age 0 to 5. In Panel B, the dependent
variable is the log number of new jobs created by young, high-tech firms in columns (1) and (2) and job creation
rate by these firms in columns (3) and (4). The odd columns construct these variables using only data from the top
MSAs, which are defined as the largest MSAs by angel volume that account for at least 90% of angel deals in the year
before the tax credit implementation. The even columns use statewide data. MDE for 80% power is the minimum
detectable effect (MDE) for 80% power. Details are in Section II of the Internet Appendix for variable definitions
and in Section IV of the Internet Appendix for power calculations. The sample period is 1988 to 2018. 1(AT C) is
an indicator equal to one if a state has an angel tax credit program in that year. Control variables are defined in
equation (1). Each observation is a state-year. All specifications include state and year fixed effects. Standard errors
are reported in parentheses and clustered by state. ***, **, and * denote significance at the 1%, 5%, and 10% level,
respectively.

Panel A: Firm Entry

Counts Rates
(1) (2) (3) (4)
1(ATC) -0.020 -0.003 -0.010 -0.002
(0.022) (0.009) (0.012) (0.008)
Year and State FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Geography Top MSAs State Top MSAs State
MDE for 80% Power 0.040 0.019 0.022 0.017
Observations 1,550 1,550 1,550 1,550
Adjusted R2 0.993 0.996 0.496 0.620
Panel B: Job Creation
Counts Rates
(1) (2) (3) (4)
1(ATC) 0.006 0.018 0.012 0.001
(0.026) (0.019) (0.016) (0.018)
Year and State FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Geography Top MSAs State Top MSAs State
MDE for 80% Power 0.059 0.044 0.031 0.028
Observations 1,550 1,550 1,550 1,550
Adjusted R2 0.981 0.983 0.221 0.283

64
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table V
Angel Tax Credits and Firm-Level Outcomes
This table reports the effect of receiving a tax credit on firm-level outcomes, using the sample of firms that applied to
be certified for angel investors to receive a tax credit. 1(T ax Creditit ) is an indicator variable for startup i having an
investor that receives a tax credit in year t. The dependent variable in column (1) is an indicator variable denoting
that a startup received VC financing within two years after applying to be certified for angel investors to receive a
tax credit. The dependent variable in column (2) is an indicator variable equal to one if a startup experienced a
successful exit (via acquisition or IPO). The dependent variables in columns (3) and (4) are indicator variables equal
to one if a startup had more than 25 employees two or three years after it applied to be certified for angel investors
to receive a tax credit. This is repeated in columns (5) and (6) except using the 75th percentile employment among
firms in the sample. Employment data are from the Steppingblocks LinkedIn panel. 1(F inance pre-T C Y r) is an
indicator variable for whether a firm received any other external financing before its investors received a tax credit.
All specifications include sector-year and state-year fixed effects. Standard errors are clustered at the state-year level.
***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

Raised Exit Empl > 25 Empl > 75th Pctile

VC
2 Yrs 3 Yrs 2 Yrs 3 Yrs
Post-TC Post-TC Post-TC Post-TC
(1) (2) (3) (4) (5) (6)
1(Tax Credit) -0.009 -0.005 0.006 0.010 0.018 0.015
(0.016) (0.009) (0.006) (0.007) (0.014) (0.013)
1(Finance pre-TC Yr) 0.174∗∗∗ 0.086∗∗∗ 0.029∗∗∗ 0.040∗∗∗ 0.003 -0.005
(0.027) (0.015) (0.010) (0.011) (0.014) (0.017)
1(Empl > 25 in TC Yr) 0.811∗∗∗ 0.780∗∗∗
(0.040) (0.039)
1(Empl > 75th Pctile in TC 0.753∗∗∗ 0.739∗∗∗
Yr)
(0.030) (0.033)
Sector-Year FE Yes Yes Yes Yes Yes Yes
State-Year FE Yes Yes Yes Yes Yes Yes
Observations 3,218 3,218 3,218 3,218 3,218 3,218
Adjusted R2 0.260 0.047 0.480 0.421 0.612 0.575

65
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table VI
Crowding Out
This table examines whether angel tax credit programs crowd out alternative early-stage financing. Panel A examines
the effect of angel tax credits on aggregate early-stage financing received by young high-tech firms at the state-year
level. The dependent variables are aggregate early-stage financing, nonangel financing, angel financing, and the
fraction of angel financing in a state-year. All financing amounts are log transformed. Early-stage financing are all
early rounds (see Section IV.A for a detailed definition) identified in CVV and Form D data, including angel rounds.
Panel B examines the effect of angel tax credits on total early-stage financing received by firms at the firm level.
Columns (1) and (2) are unweighted and columns (3) and (4) weight each observation by one over the number of
firms in each state. The sample period is 1993 to 2016. Standard errors are clustered at the state level. ***, **, and
* denote significance at the 1%, 5%, and 10% level, respectively.

Panel A: Aggregate Financing at State-Year Level

Ln(early-stage) Ln(nonangel) Ln(angel) Angel share
(1) (2) (3) (4)
1(ATC) -0.068 -0.326∗ 0.268∗ 0.075∗∗
(0.118) (0.178) (0.142) (0.029)
State FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Observations 1,200 1,200 1,200 1,200
Adjusted R2 0.853 0.706 0.813 0.247
Panel B: Total Early-Stage Financing at Firm Level
Ln(early-stage)
(1) (2) (3) (4)
1(ATC) 0.005 -0.001 -0.006 0.002
(0.038) (0.045) (0.048) (0.046)
Weighted No No Yes Yes
State FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Age FE Yes Yes Yes Yes
Controls No Yes No Yes
Observations 38,487 38,487 38,487 38,487
Adjusted R2 0.098 0.099 0.088 0.094

66
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table VII
Relabeling
Panel A reports summary statistics for tax credit recipients who are insider investors, defined as angel investors who
also serve as executives or managers at the firm for which they receive angel tax credits, as well as their family
members. For company-level statistics, the unit of observation is a unique tax credit beneficiary company for which
we observe an investor-company link. For investor-level statistics, the unit of observation is a unique investor for
which we observe an investor-company link. Panel B compares the Form D filing rate by beneficiary firms (treated)
and matched nonbeneficiary firms (control). The panel also compares covariates across the two samples. Each treated
firm is matched to up to five similar control firms through a nearest-neighbor matching procedure. To match with
a treated firm, the control firm(s) must also have received angel financing, be located in a different state but the
same Census division, belong to the same sector, have similar age (within two years), and have a similar amount of
previous financing relative to the year of the treatment firm’s first tax credit.

Panel A: Tax Credit Take-up by Insiders

N Fraction
Company level
≥1 investor is executive or has family member who is executive 628 0.35
among Kentucky companies 77 0.04
among Maryland companies 81 0.38
among New Jersey companies 63 0.24
among New Mexico companies 61 0.26
among Ohio companies 346 0.44
≥1 investor is an executive 628 0.33
Investor level
Investor is executive or has family who is executive 3,560 0.14
Investor is executive 3,560 0.11
Panel B: Form D Filing Rate by Beneficiary and Matched Nonbeneficiary Firms
Got Tax Credit No Tax Credit t-Test
Mean SD Obs Mean SD Obs t-value p-value
Filed Form D 0.644 0.479 517 0.320 0.467 3,129 −14.56 0.000
Year Founded 2009.5 4.137 517 2009.3 3.803 3,129 −1.282 0.200
Total Financing 10.15 27.25 517 8.106 23.28 3,129 −1.035 0.301
Average Emp 6.450 10.55 517 7.811 88.61 3,129 0.386 0.700
Average Sales 777,256 3,390,227 517 663,931 3,015,808 3,129 −0.661 0.508

67
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table VIII
Characteristics of Investors Receiving Tax Credits
This table describes the characteristics of investors who received angel tax credits. We gather information from
LinkedIn on angel investors from seven states that publicly release the names of individual investors who received
angel tax credits. Corporate Executive is an investor who lists their current occupation as President, Vice President,
Partner, Principal, Managing Director, or Chief Officer other than Chief Executive Officer. Gender and race are
identified from pictures. An individual’s approximate age is derived by adding 22 years to the difference between the
individual’s college graduation year and the median year of investment in the sample, which is 2013.

N Fraction N Fraction
Number of investor-tax credit pairs 8,218 Profession 3,286
Corp. Exec. 0.82
Number of unique investors 5,637 Doctor 0.073
Illinois 0.14 Entrepreneur 0.062
Kentucky 0.05 Lawyer 0.041
Maryland 0.16 Investor 0.007
Minnesota 0.39 Other 0.003
New Jersey 0.09
New Mexico 0.03 Race 4,446
Ohio 0.14 White 0.95
South Asian 0.03
Location is in state 4,694 0.79 East Asian 0.02
Black 0.007
Male 4,702 0.87 Hispanic 0.002
Middle Eastern 0.001
N Mean
Age 2,363 41.9

68
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table IX
Which Investors Respond to Angel Tax Credits?
This table examines changes in investor composition due to angel tax credit programs. Panel A reports differences-in-
differences estimates for the effects of angel tax credits on investor characteristics using AngelList. Each observation
is an investor-startup pair (i.e., investment) and is weighted by one over the number of observations in each state. The
dependent variables are indicators if an investor was in-state, inexperienced (fewer than five years of deal experience),
had no prior exit, or had no prior founder experience. All specifications include CBSA and year fixed effects. Panel
B reports the differences-in-differences estimates for the effects of angel tax credits on the entry of investors using
AngelList. Each observation is a state-year. The dependent variable is the log number of investors in each category
(in-state, out-of-state, inexperienced, experienced, had no prior exit, had exit, no prior founder experience, had
founder experience) who invested in a state-year. All specifications include state and year fixed effects. 1(AT C) is
an indicator equal to one if a state has an angel tax credit program in that year. Control variables are defined in
equation (1). The sample period is 2003 to 2017 in both panels. Standard errors are reported in parentheses and
clustered by state. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

Panel B: Investor Entry at the State-Year Level

No founder Has founder
In-state Out-of-state Inexperienced Experienced Had no exit Had exit
experience experience
(1) (2) (3) (4) (5) (6) (7) (8)
1(ATC) 0.284∗∗ 0.194∗ 0.275∗∗ 0.103 0.277∗∗∗ 0.150 0.262∗∗ 0.134
(0.119) (0.099) (0.106) (0.112) (0.095) (0.114) (0.101) (0.131)
State FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Controls Yes Yes Yes Yes Yes Yes Yes Yes
Observations 735 735 735 735 735 735 735 735
Adjusted R2 0.867 0.848 0.863 0.821 0.854 0.846 0.867 0.803

Panel A: Investor Characteristics at the Investment Level

No founder
In-state Inexperienced Had no exit
experience
(1) (2) (3) (4)
1(ATC) 0.075∗∗ 0.041∗∗ 0.073∗∗ 0.069∗∗
(0.031) (0.018) (0.027) (0.032)
Year FE Yes Yes Yes Yes
CBSA FE Yes Yes Yes Yes
Controls Yes Yes Yes Yes
Observations 89,146 89,146 89,146 89,146
Adjusted R2 0.202 0.102 0.176 0.109

69
 15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table X
Survey Analysis
This table examines investors’ perception of the importance of angel tax credits based on survey data. In Panel A,
the dependent variable is ATC importance, a score that takes a value of 1 to 5 (1 = “not at all important” and 5 =
“extremely important”). Column (1) examines whether a respondent has done an above-median number of angel deals
since January 2018. Column (2) focuses on investor experience measured by matching respondents to AngelList data.
Column (3) examines investor profession. Column (4) examines surveyed importance of other investment factors.
The first four independent variables describe any past experience using AngelList data. The remaining variables are
based on the survey. Panel B examines how deal experience correlates with why a respondent perceives angel tax
credits to be unimportant. All regressions include state fixed effects. Standard errors are clustered by state. ***, **,
and * denote significance at the 1%, 5%, and 10% level, respectively.

Panel A. ATC Importance

ATC importance
(1) (2) (3) (4)
Above median no. of deals since 2018 -0.229***
(0.041)
Has exit (AL) -0.199***
(0.039)
Has founder exper (AL) -0.118*
(0.061)
Has invested as insider (AL) 0.103**
(0.049)
Top school (AL) -0.138***
(0.033)
Corp Executive -0.144
(0.110)
Entrepreneur -0.193*
(0.105)
Investor -0.375***
(0.136)
Team importance -0.103**
(0.040)
Business importance 0.127***
(0.034)
Location importance 0.055*
(0.031)
Financial return importance 0.117***
(0.020)
Add value importance 0.041**
(0.017)
Valuation importance 0.001
(0.031)
Gut reaction importance -0.02
(0.021)
Deal terms importance 0.141***
(0.029)
State FE Yes Yes Yes Yes
Observations 1,202 1,199 1,242 1,331
Adjusted R2 0.126 0.048 0.121 0.170

70
15406261, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jofi.13267 by Cochrane Peru, Wiley Online Library on [14/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Cannot use

(0.021)
0.003

1,202
0.090
Yes
(5)
Coordination Nonfinancial Too small

(0.021)
-0.021

1,202
0.018
Yes
(4)

(0.016)
0.006

1,202
0.007
Yes
(3)
Panel B. Reasons for ATC Unimportance

0.051**
(0.024)

1,202
0.025
Yes
(2)

71
Home run

0.046**
(0.020)

1,202
0.018
Yes
(1)
Above median no. of deals since 2018

Observations
Adjusted R2
State FE