Microsoft Word - BernsteinKorteweg

Is it Just the Idea that Matters? A Randomized Field

Experiment on Early Stage Investments

Shai Bernstein, Arthur Korteweg, and Kevin Laws*

Abstract

Which start-up characteristics are most important to investors in early-stage firms? This paper

uses a randomized field experiment involving 4,500 active, high profile, early stage investors,

implemented through AngelList, an online platform that matches investors with start-ups that are

seeking capital. The experiment randomizes investors’ information sets on “featured” start-ups

through the use of nearly 17,000 emails. Investors respond strongly to information about the

founding team, whereas they do not respond to information about either firm traction or existing

lead investors. This result is driven by the most experienced and successful investors. The least

experienced investors respond to all categories of information. We present evidence that the

information materially impacts investment rates in start-up companies. The results suggest that,

conditional on the quality of the idea, information about human assets is highly important for the

success of early stage firms.

JEL classification: G32, L26, D23

Keywords: Angels, Early stage firms, Entrepreneurship, Crowdfunding, Theory of the firm

Current Draft: March, 2014

* Shai Bernstein ([email protected]) and Arthur Korteweg ([email protected]) are from

Stanford Graduate School of Business, and Kevin Laws is from AngelList, LLC. We thank

Wayne Ferson, Steve Kaplan, seminar participants at Harvard Business School, UCLA,

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

California, University of Texas at Austin, and brown bag participants at the UC Berkeley Fung

Institute and Stanford for helpful comments.

2

Early stage investors provide an important source of capital enabling the birth and growth of

start-up companies, which play a key role in promoting innovation and growth in the economy

(Solow (1957)). A large and growing literature analyzes the implications of early stage

investments (e.g., Kortum and Lerner (2000)) and the factors that affect the terms of financing

(e.g., Kaplan and Stromberg (2003)). There is, however, much less understanding of the

investment decision-making of early stage investors, that is, how do they choose which start-up

to fund? What leads investors to pick one company over the other? This stands in sharp contrast

to the wealth of evidence on the behavior of retail and institutional investors who target

investments in publicly traded firms.

Studying the investment decision-making of early stage investors is challenging for

several reasons. First, existing databases consist of completed deals, rather than the pool of start-

ups considered by the investor, thus preventing assessment of the investor selection process.

Second, researchers possess far less data than investors do. Two early stage ventures may seem

similar to a researcher, but may look completely different to an investor. Third, even if such data

were available, how can one separate the causal effect of different start-up characteristics? For

example, if we observe that a serial entrepreneur is more likely to attract financing, is it because

of the importance of her past experience, or because serial entrepreneurs are just more likely to

generate high-quality ideas in the first place. This endogenous relationship between ideas and

start-up characteristics is even more problematic given that the quality of ideas is subjective,

highly uncertain, and unobserved to the researcher, leading to omitted variable bias.

We study which start-up characteristics early-stage investors causally respond to using a

randomized field experiment that builds on the correspondence testing methodology that has

3

been successfully used in labor economics.

1

The experiment takes place on AngelList, an online

platform that matches start-ups and angel investors.

2

We observe the start-up companies at the

stage in which they approach investors and seek capital. The platform includes many well-known

investors that are experienced in investing in, and building, early-stage firms. These individuals

are highly involved in startup creation, taking various roles in these firms such as investors,

board members, advisors and founders, and are therefore well-suited to inform research about

early-stage investor decision-making, and to reveal which start-up characteristics are most

important for early-stage firm success.

AngelList regularly sends out emails to investors to feature start-ups who are attempting

to raise capital. We randomize the information shown in these emails, and measure investors’

responses and level of interest to causally infer what factors drive investors’ decisions. Investors’

interest is gauged by measuring whether they choose to learn more about the firm on the

platform. While all investors see similar information on the start-up idea and potential market,

we randomly expose investors to information in three categories that “package” the company:

the founding team’s background (e.g., college, prior work experience, or entrepreneurial

background), the start-up’s traction (e.g., revenues and user growth), and the identity of current

investors. Broadly speaking, the team and traction categories correspond to information about

human and non-human assets, respectively. Kaplan, Sensoy, and Stromberg (2009) show the

1

The correspondence testing methodology has been frequently used in labor economics by exploring responses to

fictitious job applications. For example, Bertrand and Mullainathan (2004) explore racial discrimination in the labor

market, Weichselbaumer (2003) study the impact of sex stereotypes and sexual orientation, and Nisbett and Cohen

(1996) study employers’ response to past criminal activity.

2

Examples of well-known companies that have raised money through AngelList are Uber, Pinterest, BranchOut,

and Leap Motion. Around 1,300 confirmed financings had been made through AngelList, raising over $200 million

of seed funding. Most of these investments were concentrated in 2012 and 2013. The companies funded through

AngelList have gone on to raise over $2.9 billion in later rounds of venture capital and exit money.

4

differential importance of these two types of capital as the firm grows. The current investors

category can be interpreted as “social proof”, or a different type of human assets, as prominent

investors may apply their expertise or network to the firm.

3

These categories are randomly

revealed to each investor, and we exploit the variation across angels’ reactions within each start-

up. In other words, conditioning on the start-up’s “idea”, we exogenously vary the

aforementioned categories of information.

We sent about 17,000 emails to nearly 4,500 angels, spanning 21 different capital-

seeking start-ups, over the summer of 2013. Our target investors include many of the most

prominent angel investors, who are very active in the startup scene. Among these investors, 82%

have past investments, with a median of eight companies in their portfolio. Almost half of the

investors serve as an advisor to one or more start-up companies. Interestingly, 60% of the

investors have an entrepreneurial background themselves, as they formed at least a single venture

in the past.

The start-ups in the experiment are at a very early stage. The median company seeks to

raise $1.3 million, and approximately half the companies have already raised some capital, with

a median amount raised of $290,000. The median startup has two founders and employs three

additional workers. Only 23% of these firms already formed a board, and 57% graduated from an

incubator.

The randomized experiment reveals that angels are highly responsive to information

about the founding team, whereas information about the traction and current investors does not

lead to a significantly higher response rate. This shows that the information about the human

capital of the firm is very important to potential investors. Interestingly, we find significant

3

The investor category does not, however, reveal information about the level of funding, as this is separately

disclosed in the email.

5

heterogeneity among angels. On one hand, the most highly experienced investors, as measured

by a battery of metrics (including the number of prior investments, a success metric, and their

network centrality), react only to the team information. On the other hand, the least experienced

investors react to all categories of information. This suggests that the significance of team

information is not simply due to a better signal-to-noise ratio.

A unique feature of our setting is that it allows us to empirically confirm that absent the

randomization of information revelation, we would overestimate the importance of the various

categories, as we can run the same models on the subset of emails with the information set that

would have been sent outside of the experiment. This is consistent with the notion that start-up

characteristics may be mistakenly considered as highly influencing investor decision-making

because of their positive correlation with high quality ideas, and thus underscoring the

importance of randomization for establishing causality.

We mitigate external validity concerns by showing that the start-ups in the experiment

are fairly representative for the more than 5,500 start-ups raising capital on AngelList that

attracted a minimal level of attention by investors, across a long list of observable characteristics.

A common drawback of the correspondence testing methodology is that the use of

fictitious correspondence does not allow the researcher to observe real outcomes. For example,

Bertrand and Mullainathan (2004) sent fictitious resumes with randomized names to recruiters in

order to study discrimination, but they cannot observe actual hiring outcomes. Since our

experiment involves real start-ups and real investors, we can in fact observe real outcomes such

as investor introduction requests and investments. We show that when investors are interested in

learning more about the firm (henceforth ‘click’), this interest converts to introductions between

founders and investors, as well as actual investments, at a rate of 15.1% and 3.0%, respectively.

6

The click-to-investment rate may be as high as 6.0% due to underreporting of investments. These

conversions rates indicate that clicks have a material impact on real outcomes.

As discussed in Kaplan, Sensoy, and Stromberg (2009), existing theories of the firm

yield different predictions towards the importance of key assets that the organization is built

around. The property rights theory (Grossman and Hart (1986), Hart and Moore (1990),

Holmstrom (1999), amongst others) places the ownership of non-human assets at the core of the

firm, whereas the contrasting view puts the human assets of the firm at its core (e.g., Wernerfelt

(1984), Rajan and Zingales (1998, 2001) and Rajan (2012)).

4

Our results present suggestive evidence of the importance of human capital assets at the

earliest stages of the firm, around the firm’s birth. Our results, however, do not suggest that non-

human assets are not essential. Kaplan, Sensoy, and Stromberg (2009) explore the evolution of

50 venture capital backed companies from the business plan stage to initial public offering (IPO).

They find that business lines remain stable from birth to IPO, while management turnover is

substantial. Combined with this evidence, the evidence in our paper tells a story that is consistent

with the model in Rajan (2012). Rajan argues that the entrepreneur’s human capital is important

early on to differentiate her enterprise. However, to raise substantial funds (for instance, when

going public), the entrepreneur needs to go through a standardization phase that will make

human capital in the firm replaceable, so outside financiers can obtain control rights. Our results

indeed suggest the importance of human assets at the earliest stages of a firms’ life. Kaplan,

Sensoy and Stromberg (2009) show that at the later stages, human capital is frequently replaced,

4

For example, Rajan and Zingales (1998, 2001) view the firm as a hierarchy of people who gain different degrees of

access to critical resources in the firm. The critical resources can be a person, a business idea, or key customers.

These resources are providing an incentive to specialize human capital towards the firm’s goals. In newer firms,

therefore, the competitive advantage comes from specific human capital rather than from non-human assets, which

can be bought or sold easily (Zingales (2000)).

7

as different human capital skills are needed to run a larger, more mature firm, and to ready the

firm for the injection of significant outside capital.

This paper adds to the literature on early stage investments. While some papers attempt to

illustrate causality by linking early stage investors’ actions to firm success (e.g., Kerr, Lerner and

Schoar (2013), Kortum and Lerner (1998), Sorensen (2007), Samila and Sorenson (2010), and

Bernstein, Giroud, and Townsend (2013)), there is little disagreement that early stage investors

are skilled pickers of successful companies (e.g., Kaplan and Schoar (2005), Sorensen (2007),

Puri and Zarutskie (2012), Korteweg and Sorensen (2013)). Yet, little is known about how

exactly this class of investors selects the companies to which they provide funding. Several

papers explore investors’ behavior using surveys and interviews (e.g., Pence (1982), MacMillan,

Siegel, and Narasimha (1986), MacMillan, Zemann, and Subbanarasimha (1987), and Fried and

Hisrich (1994)), but our paper provides the first large sample systematic evidence on this issue,

spanning thousands of investors.

The paper is structured as follows. In section 1, we give a brief overview of the AngelList

platform. Section 2 describes the randomized experiment, and section 3 presents descriptive

statistics. In section 4 we analyze investors’ reactions to the emails. Section 5 discusses issues of

internal and external validity. In section 6 we dive deeper into the real effects of disclosed

information on investment and introductions, and section 7 concludes.

1. The AngelList platform

AngelList is a platform that connects start-ups with potential angel investors. The

platform was founded in 2010 by Naval Ravikant (the co-founder of Epinions) and Babak Nivi

(a former Entrepreneur-in-Residence at Bessemer Ventures and Atlas Capital), and has

8

experienced rapid growth since. Start-up companies looking for funding may list themselves on

the platform and post information about the company, its product, traction (e.g., revenues or

users), current investors, the amount of money they aim to raise and at which terms, and any

other information they would like to present to potential investors. Examples of well-known

companies that have raised money through AngelList are Uber, Pinterest, BranchOut, and Leap

Motion.

On the angel side, investors that are accredited following the rules set by the U.S.

Securities and Exchange Commission

5

can join the platform to look for potential investments.

Investors typically list information on their background as well as their portfolio of past and

current investments. The platform is host to many prominent and active angels with extensive

experience investing in, building, and operating early stage companies. Examples are Marc

Andreessen and Ben Horowitz (of Andreessen-Horowitz), Reid Hoffman (co-founder of

LinkedIn), Yuri Milner (founder of Digital Sky Technologies), Marissa Mayer (president and

CEO of Yahoo), Max Levchin (co-founder of Paypal), and Dave McClure (of 500 Startups).

Using AngelList, interested investors request an introduction to the start-up’s founders.

From there the parties can negotiate their way to a final investment. Usually, investors decide to

invest following a phone call with the founders or, depending on geographical closeness, a face-

to-face meeting.

There is a strong social networking component to the platform: Investors can “follow”

each other as well as start-ups, they can post comments and updates, and they can “like”

comments made by others.

5

For individuals, an accredited investor is a natural person with either at least $1 million in net worth (either

individually or jointly with their spouse, but excluding the value of their primary residence) or with income of at

least $200,000 (or $300,000 jointly with a spouse) in each of the two most recent years and a reasonable expectation

of such income in the current year.

9

By the fall of 2013, about 1,300 confirmed financings had been made through AngelList,

raising over $200 million. Most of these investments were concentrated in 2012 and 2013. The

companies funded through AngelList have gone on to raise over $2.9 billion in later rounds of

venture capital and exit money. There is no exact benchmark to compare these numbers to, but to

give a rough comparison, the University of New Hampshire’s Center of Venture Research

estimates total 2012 angel investments at $22.9 billion, while seed rounds of start-ups totaled

$731 million in 2012 and $893 million in 2013.

6

2. Randomized field experiment

The field experiment builds on correspondence testing methodology

7

and uses so-called

“featured” emails about start-ups that AngelList regularly sends out to investors listed on its

platform. These start-ups are chosen by AngelList for being promising companies that could be

appealing to a broad set of investors that have previously indicated an interest in the industry or

the location of the start-up.

An example of a featured email is shown in Figure 1. The email starts with a description

of the start-up and its product. Next, the email shows up to three categories of information about:

i) the start-up team’s background; ii) current investors; iii) traction. Outside of the experiment, a

category is shown if it passes a certain threshold as defined by AngelList. The thresholds are

AngelList’s determination of what information investors might be most interested in. For

example, the team category is shown if the founders were educated at a top university such as

6

Data on total seed funding is from CB Insights: http://www.cbinsights.com/blog/trends/2013-seed-venture-capital,

accessed February 17, 2014. The University of New Hampshire’s angel market analysis reports are accessible at

https://paulcollege.unh.edu/research/center-venture-research/cvr-analysis-reports.

7

This approach has been used in the context of job application recruiting in various studies. Written applications are

sent to job openings and applications are constructed such that they differ only in the aspect of interest. These

studies explore the employer reactions to the fictitious job applications. A few recent examples include Bertrand and

Mullainathan (2004), Weichselbaumer (2003), and Nisbett and Cohen (1996). Our study does not rely on fictitious

subjects, but rather uses real investors and real early stage ventures seeking for capital.

10

Stanford, Harvard, or MIT, or if they worked at a top company such as Google or Paypal prior to

starting the company. As we discuss below, this algorithm is important for the interpretation of

the experiment’s results, and to provide a sense of the information content in featured emails, the

appendix shows the information that passed the disclosure threshold for each start-up in the

experiment. Finally, the email shows information about the amount of money that the company

aims to raise, and how much has been raised to date.

In the experiment, we randomly choose which of the team, current investors, or traction

categories are shown in each email, from the set of categories that exceed their threshold. For

example, suppose 3,000 angels receive a featured email about a given start-up. Outside of the

experiment, all investors would receive the same email, and let’s assume that this email would

show information about the team and traction, while the current investors category for this

company does not meet the threshold to be included in the email. In the experiment, 1,000

investors receive the original email with both team and traction shown, 1,000 receive the

identical email except that it does not show the team category, and another 1,000 receive the

email that shows the team information but with the traction category hidden. We do not send any

emails with all categories hidden, as this would not happen outside of the experiment, and could

raise suspicion among investors.

Investors respond to the emails using the “View” and “Get an Intro” buttons that are

included in each email (see Figure 1). If an investor is interested in the start-up, she can click on

the “View” button to be taken to the AngelList website and view the detailed company profile.

We record if this happens. If the investor is particularly interested, she can click the “Get an

Intro” button to request an introduction to the company straight away. However, this is a very

rare event as nearly all investors take a look at the full company profile on the AngelList website

11

before asking for an introduction. Hence, instead of clicks on the “Get an Intro” button, we

record whether the angel asks for an introduction within three days of viewing the email through

either the email or the website. Naturally, we need to exercise caution in interpreting the results

on introductions, as investors will likely have gleaned more information from the website.

3. Summary statistics

A. Emails

We ran the experiment over an eight week period in the summer of 2013. Table 1 reports

descriptive statistics. Panel A shows that a total of 16,981 emails were sent to 4,494 active

investors, spanning 21 unique start-ups. Active investors are angels that have requested at least

one introduction to a start-up while they have been enrolled in AngelList. Investors come to the

platform for a variety of purposes: to research, to confirm their affiliation with a startup that is

fundraising, or to invest. Restricting the sample to active investors excludes those are not on the

platform to seek new investments

For each start-up, we sent an average (median) of 2.76 (3) versions of the email, each

with an exogenously different information set. This means that in total we sent 58 unique emails

(2.76 emails per start-up times 21 start-ups). Each unique email was sent to 293 recipients on

average (median 264). Within a start-up, the number of recipients per unique email is roughly

equal, but there is some variation across start-ups in how many angels receive the featured

emails, as some start-ups are in more popular industries or locations than others. On average, 809

investors receive a featured email about a given start-up, with a minimum of 202 and a

maximum of 1,782 recipients per start-up. An investor in the sample receives on average 3.78

12

emails (median: 3 emails) of different featured start-ups, and importantly, no investor receives

more than one email for a given start-up.

In terms of response, recipients opened nearly half (48.3%) of their emails. Some

investors open none of their emails, but 2,925 investors open at least one. Of the opened emails,

16.45% of investors clicked on the “View” button to see more information about the start-up.

This click rate provides the first hint that investors pay attention to the emails: they do not click

on every company, but they also do not ignore the emails and the information therein altogether.

Of the investors who clicked on the email, we see that 15.1% requested an introduction within

three days of viewing the email. However, this includes not only direct introductions from the

email but also introductions that were made later, when investors have seen more information

about the start-up. Finally, 2.98% of investors who clicked on one of the email buttons ended up

investing in the company. Since investors are not required to report their investment to

AngelList, this statistic may underestimate the real click-to-investment conversion. AngelList

estimates that only half of the investments are being reported, potentially leading to a click-to-

investment rate of approximately 6%.

Panel B of Table 1 shows that there is no statistically significant difference in the

frequency with which each information category passes the threshold set by AngelList. This

means that the salience of the presence of an information category in an email is roughly equal

across information categories. Outside of the experiment, categories that pass the threshold

would always be shown in the email. Within the experiment, these categories are randomly

excluded. Conditional on passing the threshold, the information regarding, team, current

investors, and traction is shown about 73% of the time, with no material difference in

frequencies across categories. Note that these frequencies are different from 50% because we

13

randomize across different versions of the emails. For example, if team and traction pass the

threshold, there are three versions of the email: one that shows team only, one that shows traction

only, and one that shows both (we don’t use the empty set to avoid raising suspicion amongst

investors). In that case, if each email is shown at random then team and traction would each be

shown 67% of the time.

B. Start-ups

Table 2 presents detailed descriptive statistics of the 21 start-ups in the randomized email

experiment. Panel A shows the geographical distribution of firms. The most popular location is

Silicon Valley with six firms, but the dispersion is quite wide, with firms spread across the

United States, Canada, the United Kingdom, and Australia. Panel B shows that most firms

operate in the Information Technology and the Consumers sectors. Other represented sectors are

Business-to-business, Cleantech, Education, Healthcare, and Media. Note that the sector

designations are not mutually exclusive. For example, a Consumer Internet firm such as Google

would be classified as belonging to both the Information Technology and Consumers sectors. In

terms of company structure, panel C shows that the median start-up has two founders, and 17

start-ups (81%) have (non-founder) employees. The median firm with employees has three

workers, though there is some variation, with the largest company having as many as nine

employees. Counting both founders and employees, the largest start-up consists of 11 people.

Only a quarter of firms have a board of directors at this stage of fund-raising. Of those that do

have a board, the median board size is two, and no board is larger than three members.

8

Almost

8

It is not clear how much of an outside governance role the board fulfills at this stage of the firm, rather than simply

fulfilling a legal requirement of incorporation.

14

all companies (19 out of 21) have advisors,

9

and the median number of advisors for the

companies that have any, is three.

Panel D reports details on the financing of the sample firms. Twelve companies (57%)

had previously gone through an incubator or accelerator program. Eleven companies (52%)

received funding prior to coming to AngelList for further financing, and had raised an average

(median) of $581 thousand ($290 thousand). For the sixteen companies for which a pre-money

valuation is available, the average (median) valuation is $5.5 million ($5 million), and ranges

between a minimum of $1.2 million and a maximum of $10 million. Eighteen companies

explicitly state their fundraising goal, which ranges from $500 thousand to $2 million (not

tabulated), with an average (median) of $1.2 million ($1.3 million). Most companies (76%) are

selling shares, with the remaining 24% selling convertible notes.

C. Investors

Table 3 reports descriptive statistics of the 2,925 angel investors who received the

featured emails in the field experiment, and who opened at least one email. This is the set of

investors that is the focus of our empirical analysis in the next section. Panel A shows that

virtually all investors are interested in investing in the Information Technology and Consumers

sectors, while other key sectors of interest are Business-to-business, Healthcare and Media. Panel

B reveals that investors are very active on the platform, with the average (median) investor

requesting ten (three) introductions to start-ups from the time that they joined the platform until

we harvested the data in the late summer of 2013. However, there is considerable heterogeneity

in the number of introductions requested, with the lowest decile of investors requesting only one

introduction, while the top decile requested over twenty.

9

Advisors are typically high profile individuals, and are compensated with stocks and options.

15

In order to provide an indication of the past success of investors, AngelList computes a

“signal” for each investor and start-up that ranges from zero to ten. The algorithm that assigns

signals works recursively, and is seeded with high exit value companies (from Crunchbase) such

as Google or Facebook getting assigned a value of ten, as well as a set of hand-picked (by

AngelList) highly credible investors. The signal then spreads to start-ups and investors through

past investments: any start-up that has a high signal investor gets a boost in its own signal.

Likewise, an investor who invests in a high signal company gets a boost in his or her signal. This

signal construction, rather than crediting investors only for realized past successes, also gives

credit for investing in very young but highly promising firms that may have great exits in the

future, but are still too young to have made it to the exit stage. The average (median) investor

signal is 6.4 (6.3), with a standard deviation of 2.3. The wide distribution of the signal in Figure

2 shows that there is significant heterogeneity in signal across investors.

10,11

The social network on the platform is extensive, and the investors in the sample are well-

connected: The average (median) investor had 591 (202) followers at the time of data collection.

Again, we see large heterogeneity in investors, with the 10

th

percentile having only 26 followers

while the 90

th

percentile investor has 1,346 followers.

Over 90% of investors are actively involved with start-ups (as with the signal calculation,

these numbers are not limited to start-ups that tried to raise money through AngelList). Panel B

shows that most (82%) have a track record as investors. Conditional on making an investment,

10

There are few signal scores below three, because we limit the set of investors to those that have requested at least

one introduction through the platform.

11

The signal calculations use all declared investments on the AngelList platform. This data represents self-declared

investments by both angels and start-ups on the platform that were subsequently verified by AngelList with the party

on the other end of the transaction (i.e., investments declared by start-ups are verified with the investors and vice

versa). Importantly, the data are not limited to companies that (tried to) raise money through the platform. There are

many thousands of companies, such as Facebook, that are on the platform but never have, or ever intend to, raise

money through AngelList. Instead, they are there only because an investor declared to have invested in them (or

declared to have served another role in the firm, such as founder or advisor) that was subsequently verified by

AngelList. In addition, the signal calculation includes investment data available from Crunchbase.

16

the average (median) number of investments is 13 (8), though some investors invest in as many

as 30 companies. Roughly 44% of angels are active as advisors to start-ups, with the median

advisor advising two firms. Also, 17% of investors served as a board member on a start-up. Last,

but certainly not least, 60% of investors were at one point founders themselves.

12

The median of

these founder-investors founded two companies.

The investors in Table 3, those that opened the emails, tend to be more active and

involved than the investors that received features emails but did not open any of them: they

request more introductions (9.72 on average versus 4.97 for the investors who did not open any

emails), have a higher signal (average 6.44 versus 5.89), more followers (average 591 versus

480), more of them are involved with start-ups (91.93% versus 85.15%), and conditional on

being involved, they are involved with more start-ups (average 12.55 versus 10.56) . These

differences are all statistically significant at the 1% level (results not tabulated).

Taken together, the evidence presented here shows that the group of investors in our

sample are active, successful, connected, and highly experienced not only in investing in very

early-stage firms, but also in building companies from the ground up. As such, these individuals

form a sample that is ideally suited to inform about the assets that are most important to very

early stage firms. Moreover, there is significant heterogeneity within this group that may help to

distinguish between theories.

4. Analysis of investors’ responses in the randomized experiment

Table 4 shows results of regressions that explore how the three randomized categories of

information (team, traction, and current investors) affect angels’ interest level and trigger

12

Declarations of advisor, board member, or founder roles are verified using the same procedure as was followed

for investments.

17

response. The dependent variable equals one when an investor clicked on the “View” button in

the email, and zero otherwise. All models in Table 4 have standard errors clustered at the

investor level, to account for investors making correlated decisions across the emails they receive

for various start-ups.

13

In column (1) we run a simple ordinary least squares (OLS) regression

that explores how the three information categories affect click rates.

14

Revealing information

about the team significantly changes the click rate, raising the unconditional click rate by 2.6%.

Given a base click rate of 16.5% (Table 1), this represents a 16% increase. Recall that investors

are calibrated to think that if the information is not shown, it has not crossed the threshold and is

therefore of insufficient significance for AngelList to report. This helps for interpretation, as the

increase in the click rate is thus the effect of the team’s background being above the importance

threshold. Showing information about the current investors or traction does not significantly alter

the click rate. This means that knowing whether a notable investor (by AngelList’s definition) is

investing in the company, or if the start-up has material traction, does not make investors more

likely to click.

In column (2) we introduce controls for investors’ pre-existing knowledge of the start-up

company, and the number of emails an investor has already received in the experiment. We will

discuss these results in more detail in the next section. What is important at this stage of the

analysis is that adding these controls does not change the coefficients on the randomized

information categories.

13

The results are nearly identical if we include investor fixed effects. We do not cluster standard errors at the start-

up level because there are too few clusters to produce reliable estimates (see Angrist and Pischke (2009), chapters

8.2.1 and 8.2.3). The start-up fixed effects in the regressions remove unobserved heterogeneity in click rates for each

start-up.

14

The regressor indicator variables equal one when the information is shown in the email, and zero otherwise. It is

not necessary to interact these indicators with dummy variables whether the disclosure threshold was passed, as the

results are mechanically the same.

18

Columns (3) and (4) replicate the regressions of the first two columns with the addition of

start-up fixed effects. These fixed effects control for the effect on click rates of any information

conveyed in the descriptive paragraph, the amount that the company aims to raise, has already

raised, or any other common knowledge about the specific start-up. The coefficients on the

information categories are slightly lower, but remain significant at the 5% level. The final four

columns show that the results are robust to using a logit model instead of OLS regressions.

A unique feature of our setting is that we can show the importance of the randomized

experiment for identification, by re-running the regressions of Table 4 on the subset of 2,992

opened emails that show every piece of information that crossed the threshold. These are the

only emails that would have been sent outside of the experiment. Note that with this subsample

of emails we cannot include start-up fixed effects as there is by construction no variation across

emails for a given start-up. Moreover, given the random allocation of these “full-information”

emails, the regression results should reflect the non-randomized population estimates.

Focusing on the OLS regression with the information categories as the only explanatory

variables, Table 5 shows that the coefficients on revealed information about the team, investors,

and traction are 0.046, 0.013, and 0.037, respectively, where team is significant at the 5% level,

investors is insignificant and traction is significant at the 10% level. These coefficients are

uniformly higher than the coefficients of 0.026, 0.011 and 0.010 using the full set of randomized

emails (replicated for ease of comparison in the four right-most columns of Table 5), and where

the coefficients on both investors and traction are insignificant. Clearly, the randomization of

information is important: without the experiment, we would overestimate the importance of

traction, and to some extent, team. In fact, one would expect to overestimate the importance of

19

good teams, investors, and traction if they are positively correlated with good ideas, which is

likely to be the case. The results for the other models are similar, as seen in Table 5.

A key question at this point is whether team matters because it is a measure of quality of

the business idea, or whether there is something special about the team as a human asset that is

critical to the firm? For example, if a team was trained at MIT, does this signal high human

capital necessary for execution, or does it serve as a signal of the quality of the technology or the

business plan separate from management and future implementation? If it is the latter case, and

investors care only about observing signals of idea quality, then one would expect traction and

current investor information to also correlate with click rates. This is not what we observe,

suggesting that there is something special about team information. Still, it is possible in theory

that the team information carries a higher signal-to-noise ratio then the other information

categories. We exploit the rich heterogeneity in angel investors in the sample, and in particular

heterogeneity in investment experience, in order to disentangle these stories.

The regression results in Table 6 show the difference in response between experienced

and inexperienced investors, where we use investors’ total number of investments as a measure

of experience. The first column shows that investors who have made at least one investment

behave similarly to the overall sample, and react only to the team information. The relatively

inexperienced investors, with no prior investments, who make up about 18% of the sample, not

only react to the team information, but also to the traction and current investor information.

Columns (2) and (3) redefine the cutoff between inexperienced and experienced investors at the

25

th

and 50

th

percentile of investors, ranked by their number of investments, respectively. The

results for the experienced investors remains the same, as they respond only to information about

20

the team. The significance of the response to traction and current investors categories among

inexperienced investors weakens somewhat as we broaden the definition of inexperience.

The results of Table 6 are consistent with the most inexperienced investors interpreting

all information categories as signals of the quality of the start-up. This has important

implications for the interpretation of the reaction of the experienced investors. In particular, it

means that the absence of a reaction to the traction and current investors categories is not due the

fact that the quality signal contained in this information is too low relative to the noise. Rather, it

suggests that the experienced investors believe these categories are simply less relevant to the

success of the company, and that there is something special about the information regarding the

team.

We should be careful to point out that the fact that human capital appears to matter more

to experienced investors than information about traction, does not mean that the business idea of

the start-up is irrelevant. We explore variation about information shown on human capital

conditional on the information about the company that is shown in the descriptive paragraph of

the email. This description contains information on the market, technology and other aspects of

the idea that may be important to investors. We do claim that, conditional on this information,

that is, conditional on the idea, the information about human capital matters to investors. In other

words, our results point to the jockey being important at this stage of the firm, irrespective of

whether the horse matters or not.

In Tables 7 to 9 we explore other measures of experience as well as measures of an

investor’s importance in the network. In Table 7 we use investors’ signal as an alternative

measure of investor experience and importance. In Table 8 we use the number of followers as a

measure of an investor’s importance, and in Table 9 we use the weighted number of followers

21

(weighted by investors’ signal quality). All these measures are as defined in Table 3. Overall, the

results are very robust: investors in the lowest quartile of experience or importance respond to all

categories of information, whereas investors in the top quartile only respond to the information

in the team category.

5. Internal and external validity

The experiment is run in a highly controlled information environment, where angel

investors are making decisions about the same start-up company at the same time, with

exogenously varying information sets. Still, we should be careful to consider any concerns about

the internal and external validity of the experiment.

One potential internal validity concern is that the coefficients on the disclosed

information in the regressions in Tables 4 to 9 may be affected by investors who already know

the information in the emails, especially if these are “hot” and promising start-ups. We control

for this using an indicator variable that captures whether investors already follow the start-up

before receiving the email, and a variable that counts prior connections between the investor and

the start-up, measured as the number of people on the profile of the startup (in any role) that the

investor already follows prior to receiving the email. Not surprisingly, investors are more likely

to click if they already follow the start-up, or have pre-existing connections. To the extent that

these proxies are not perfect, our results are biased towards not finding an effect of the disclosed

information, and our estimates should be interpreted as lower bounds on the importance of the

information categories. Still, the fact that even the most experienced and well-connected

investors react to the information in the emails suggests that this is not a first-order concern.

22

Another common concern with experiments that involve repeated measurements on

subjects (here: investors) is that subjects may learn about the existence of the experiment,

contaminating the results. This concern is mitigated by three features of the experiment: First, the

experiment window of eight weeks is short. Second, the randomized information categories are

not always shown outside of the experiment, so a missing category is not out of the ordinary.

Third, no investor received more than one email for any given featured start-up, so there is no

risk of the same investor receiving and comparing emails across the same start-up and noticing

different information sets. Still, to check whether investors realize that the experiment is going

on, we also include the number of prior experiment emails that the investor received as a control

in the regressions of Tables 4 to 9. The insignificant coefficients imply that click rates do not

change as the investor receives more emails in the experiment. Unreported regression results

show that including interactions of this control with the information category dummies also

yields insignificant results, showing that investor responsiveness to the information categories

also does not change as the experiment progresses.

AngelList chooses which companies to feature through email, and this could raise

validity concerns. Since our inference exploits the variation within each start-up, internal validity

is not an issue. Similarly, the choice of recipients does not violate internal validity, as

information is varied randomly across the recipients. However, the endogenous choice of start-

ups and investors does raise questions regarding external validity (i.e., generalization) of the

results.

The experiment covers a large proportion of the active angels on the AngelList platform:

of the 5,869 angels who are active on the platform, 4,494 (77%) received at least one featured

email over the course of the experiment, and 2,925 (50%) opened at least one of these emails. To

23

get a sense of representativeness of the sample of 21 start-ups in the randomized field

experiment, Table 10 compares them to a larger sample of 5,538 firms raising money on the

AngelList platform. This larger sample consists of “serious” firms in the sense that these

companies received at least one introduction request while listed on AngelList. Table 10 shows

that the field experiment firms are slightly larger in terms of the number of founders (2.6 versus

2.1 on average), pre-money valuation ($5.6 million versus $4.9 million), funding targets for the

AngelList round ($1.2 million versus $0.9 million)), are more likely to have employees (81%

versus 53%), and are more likely to have attended an incubator or accelerator program (57%

versus 30%). Still, for the most part the differences are small on economic grounds, and the

samples are comparable on other dimensions such as board size, the fraction of companies that

get funding prior to AngelList, and the prior amount raised. Also, in both samples about three out

of four firms sell equity, while the remainder sells convertible notes. Altogether, the two samples

do not look vastly different, which mitigates the concern about generalization of the results of the

field experiment.

6. Real outcomes

The analysis up to this point has focused on the click rates. It is reasonable to ask how

meaningful these clicks really are. We address this concern in three ways. First, consider the base

click rate of 16.45%. This shows that investors do not ignore the information in the emails, nor

do they click on every featured company that lands in their inbox. Second, if investors did not

care about the information in the emails, clicks would be random and we would not see strong

reactions to any information. The fact that we find economically and statistically significant

results suggests that investors do care, and pay attention to the information provided at this stage.

24

Third, we can measure the conversion rate from clicks into actual investments. This addresses a

common weakness of the correspondence testing literature, where real outcomes can usually not

be observed. For example, in their study on racial discrimination in hiring, Bertrand and

Mullainathan (2004) cannot measure the conversion rate of call backs into actual hiring, since

the randomized resumes that they send out are fictitious. Since our experiment involves real

companies and real investors, we can observe the real outcomes.

Table 1 reports that the conversion rate from clicks to ultimate investment is 2.98%. This

is fairly high compared to venture capital, as venture capitalists invest in about one in every 50 to

100 deals that they look into. Moreover, no investments were made without an initial click in

these emails. Thus, clicks are a prerequisite for investment.

The investment results should be interpreted with some degree of caution, as investors are

not required to report their investment to AngelList. The company’s best estimate is that only

half of investments are reported.

15

This means that the click-to-investment rate could be as high

as 6%.

We look at introductions as an alternate measure of real outcome that is less subjective to

the underreporting issue. AngelList records whenever an investor requests an introduction with a

start-up through the website, and introductions are therefore more precisely measured than

investments. Table 1 shows that the conversion rate of clicks to introductions is 15.14%. This

rate is higher than the investment rate not only due to the underreporting of investments, but also

because introductions are a lower screen than investments, as only a certain proportion of

introductions lead to investments. Still, the click-to-introduction conversion rate shows that

clicks are meaningful for real outcomes as they lead to a significant number of introductions.

15

We have tried to supplement the investments data using Crunchbase but did not identify any further investors in

the angel round.

25

7. Conclusion

This paper uses a field experiment to study early stage investors’ responses to

information about start-up firms. We randomly vary investors’ information sets in a tightly

controlled information environment that uses emails regarding featured start-ups, sent through

AngelList’s platform. We find that investors react most strongly to the information about the

start-up’s founding team. However, there is considerable heterogeneity among investors, and

while experienced and successful investors react only to the team information, inexperienced

investors also react to information about the firm’s traction and current investors.

We also present evidence that clicks on emails, our response measure that indicates

investor interest in the start-up, matters for real outcomes such as investors requesting

introductions to start-ups, and ultimately making investments.

Our results suggest that, conditional on the quality of the idea, human assets are

important to the success of the early stage firm. This is important in light of the debate in the

literature about which assets are central to the firm at an early stage. However, our results do not

suggest that non-human assets are not essential.

Finally, this paper opens up a set of new questions for future work. In particular, the

study of long-run outcomes for the companies that obtain funding is an important avenue for

future work.

26

References

Angrist, Joshua, and Jörn-Steffen Pischke, 2009, Mostly harmless econometrics, Princeton

University Press, Princton, NJ.

Bernstein, Shai, Xavier Giroud, Richard Townsend, 2013, The impact of venture capital

monitoring: Evidence from a natural experiment, Working paper.

Bertrand, Marianne, and Sendhill Mullainathan, 2004, Are Emily and Greg more employable

than Lakisha and Jamal? A field experiment on labor market discrimination, American

Economic Review 94, 991-1013.

Coase, Ronald, 1937, The nature of the firm, Economica 4, 386–405.

Ewens, Michael, Ramana Nanda, and Matthew Rhodes-Kropf, 2014, Entrepreneurship and the

cost of experimentation, Working paper, Carnegie Mellon University and Harvard

University.

Fried, Vance H, and Robert D. Hisrich, 1994, Toward a model of venture capital investment

decision making, Financial Management 23, 28-37.

Gompers, Paul, and Josh Lerner, 2001, The money of invention, Harvard Business School Press,

Boston, MA.

Grossman, Sanford J., and Oliver D. Hart, 1986, The costs and benefits of ownership: A theory

of vertical and lateral integration, Journal of Political Economy 94, 691-719.

Hart, Oliver D., and John Moore. 1990, Property rights and the nature of the firm, Journal of

Political Economy 98, 1119-1158.

Holmstrom, Bengt R., 1999, The firm as a subeconomy, Journal of Law, Economics, and

Organization 15, 74-102.

27

Kaplan, Steven N., and Antoinette Schoar, 2005, Private equity performance: Returns,

persistence, and capital flows, Journal of Finance 60, 1791-1823.

Kaplan, Steven N., Berk A. Sensoy, and Per Stromberg, 2009, Should investors bet on the jockey

or the horse? Evidence from the evolution of firms from early business plans to public

companies, Journal of Finance 64, 75-115.

Kerr, William R., Josh Lerner, and Antoinette Schoar. "The consequences of entrepreneurial

finance: Evidence from angel financings." Review of Financial Studies (2011).

Korteweg, Arthur G., and Morten Sorensen, 2013, Skill and luck in private equity performance,

Working paper, Stanford University and Columbia University.

Kortum, Samuel and Josh Lerner, 1998, Stronger protection or technological revolution: What is

behind the recent surge in patenting?, Carnegie-Rochester Conference Series on Public

Policy 48, 247-304.

MacMillan, Ian C, Robin Siegel, and P.N. Narasimha, 1986, Criteria used by venture capitalists

to evaluate new venture proposals, Journal of Business Venturing 1, 119-128.

MacMillan, Ian C, Lauriann Zemann, and P.N. Subbanarasimha, 1987, Criteria distinguishing

successful from unsuccessful ventures in the venture screening process, Journal of Business

Venturing 2, 123-137.

Metrick, Andrew, 2007, Venture capital and the finance of innovation, John Wiley & Sons,

Hoboken, NJ.

Nisbett, Richard E. and Cohen, 1996, Culture of honor: The psychology of violence in the South,

Boulder, CO: Westview Press.

Pence, Christine Cope, 1982, How venture capitalists make investment decisions, UMI Research

Press.

28

Puri, Manju, and Rebecca Zarutskie, 2012, On the lifecycle dynamics of venture-capital- and

non-venture-capital-financed firms, Journal of Finance 67, 2247-2293.

Quindlen, Ruthann, 2000, Confessions of a venture capitalist, Warner Books, New York, NY.

Rajan, Raghuram G., 2012, Presidential address: The corporation in finance, Journal of Finance

57, 1173-1217.

Rajan, Raghuram G., and Luigi Zingales. 1998, Financial dependence and growth, American

Economic Review 88, 559-586.

Rajan, Raghuram G., and Luigi Zingales. 2001, The firm as a dedicated hierarchy: A theory of

the origins and growth of firms, Quarterly Journal of Economics 116, 805-851.

Rin, Marco Da, Thomas F. Hellmann, and Manju Puri. 2013, A survey of venture capital

research, in: Handbook of the Economics of Finance Volume 2 Part A, eds. George M.

Constantinides, Milton Harris and Rene M. Stulz, Elsevier, 573-648.

Samila, Sampsa, and Olav Sorenson. 2010, Venture capital, entrepreneurship, and economic

growth, Review of Economics and Statistics 93, 338–349.

Solow, Robert M., 1957, Technical change and the aggregate production function, Review of

Economic and Statistics 39, 312-320.

Sorensen, Morten, 2007, How smart is smart money? a two-sided matching model of venture

capital, Journal of Finance 62, 2725–2762.

Weichselbaumer, Doris, 2003, Sexual orientation discrimination in hiring, Labour Economics

10, 629-642.

Wernerfelt, Birger, 1984, A resource-based view of the firm, Strategic Management Journal 5,

171-180.

Zingales, Luigi, 2000, In search of new foundations, Journal of Finance 55, 1623-1653.

29

Appendix: Information disclosed in featured emails

For each start-up in the randomized experiment, we show what information passed AngelList’s

disclosure threshold. This information would be shown in the featured emails outside of the experiment.

To protect the companies’ identities each column is shown in a different order, so that rows do not

correspond to companies. The team, investors, and traction information passed AngelList’s disclosure

threshold for 19, 17, and 18 start-ups, respectively.

Team information

Investors information

Traction information

Team worked at Microsoft,

Google and Ask.com.

Great Oaks and Josh

Abramowitz are investing in this

round.

$125K revenue in first 5 months,

15 companies, 1.7K testers.

Team worked at Starbucks and

Nabisco.

500 Startups is investing.

Incubated by Startmate.

$30M in transaction volume.

Team worked at Royal Bank of

Canada and went to University

of Toronto.

Summit Partners are investing in

this round.

3.2K health providers, 15%

monthly growth.

Team worked at IBM and went

to the University of Waterloo.

Hadi Partovi, Keith Rabois and

Tony Hsieh are investing in this

round.

$1M/year revenue, 70K users,

20% monthly growth.

Team worked at Accel. Went to

Cambridge and Oxford.

Quest Venture Partners are

investing in this round.

Incubated by Y Combinator.

$800K revenue/year, 60%

annual growth, 1K customers.

Team went to Stanford and

Berkeley.

Laurent Drion is investing

$500K in this round.

$250K in pre-sales, 960 pre-

orders.

Team worked at Microsoft,

Groupon and went to Stanford

GSB.

Grishin Robotics is investing

$500K in this round.

40 vending machines, 3 pilot

contracts.

Team members worked at

Accenture.

Lightbank is investing in this

round.

350 subscribing businesses, 700

active users, 125% monthly

growth.

Team worked at E*TRADE and

studied at Stanford.

Adventure Capital is investing in

this round.

$10K/month revenue, 120K

users, 7.5K courses.

Team includes the founder of

SIMMS - radiology software

used by 2 million patients.

Incubated by AngelPad.

$20K/month revenue, 25%

monthly growth, 12K monthly

active users.

Team worked at Intel and went

to The University of Chicago.

SoftTech VC and Matt

Mullenweg are investing in this

round.

$1K revenue, 12 customers.

Team worked at JPMorgan and

went to MIT.

Jeff Fluhr and Great Oaks are

investing in this round.

130K users, 10% monthly

growth, $6K/month revenue.

30

Team worked at Yahoo!, Oracle

and went to Stanford.

Dave McClure is investing in

this round.

80 users. Waiting list includes

BHP Billiton, USGS and the

WWF.

Team members went to Harvard.

Sandbox Industries are investing.

90K items for sale, 10K monthly

active users, 30% monthly

growth.

Team worked at Google and

went to The University of

Cambridge. Includes 2 Artificial

Intelligence PhDs.

Golden Gate Ventures is

investing in this round.

$20K revenue/month, 10K

engineers.

Team worked at Microsoft, GE

and went to Cornell.

Patrick Condon (co-founder of

Rackspace) investing. Incubated

by TechStars.

$1.4M revenue/year, 10K units

sold.

Team founded well.ca

($40M/year revenue), worked at

IBM and RIM.

Boris Wertz is investing in this

round. Incubated by Y

Combinator.

$10M/year revenue, 60% annual

growth, 13 stores, 25% profit

margin.

Team went to University of

British Columbia.

$70K revenue/month, 500K

monthly active users, 100K daily

active users.

The founder’s last company

designed and built 12

composting facilities in the US.

31

Table 1: Descriptive Statistics of Emails in Randomized Field Experiment

This table reports summary statistics for the sample of emails about featured start-ups in the randomized

field experiment. Each featured start-up has up to three information categories (team, traction, and current

investors) that would normally be shown in the email if the information for that category reaches a

threshold as defined by AngelList (see Figure 1 for an example). For each start-up, various unique

versions of each email are generated that randomly hide these pieces of information. These emails are

sent to investors registered on the AngelList platform. The sample is limited to active investors who have

in the past requested at least one introduction to a start-up on AngelList. Panel A shows basic descriptive

statistics regarding the emails, the investors who received the emails, and the start-ups covered by the

experiment. Each email contains a button that, when Clicked, takes the investor to the AngelList platform

where more information about the company is shown, and introductions to the company’s founders can

be requested. Intro means an investor requested an introduction to the start-up’s founders within three

days of viewing the email. Investment means the investor invested in the company at some time after

receipt of the email. Panel B shows the frequency with which each information category passed the

threshold where it would normally be shown, and how often this information was actually shown in the

emails conditional on the threshold being passed. The rightmost column shows the p-value for Pearson’s

chi-squared test with null hypothesis that the proportions in the first three columns are all equal.

Panel A: Experiment descriptive statistics

mean

st. dev.

percentile

10

50

90

Emails

Total

16,981

Unique

58

Investors / unique email

293

149

86

264

468

Active investors emailed

4,494

Active investors who opened at least one email

2,925

Start-ups

21

Investors / start-up

809

468

338

676

1,451

Unique emails / start-up

2.76

0.62

2

3

4

Start-ups / investor

3.78

2.45

1

3

7

Emails opened (%)

48.28

Of opened emails:

Clicked (%)

16.45

Of clicked emails:

Intro (%)

15.14

Investment (%)

2.98

32

Panel B: Information in emails

Team

Investors

Traction

p-value

Information passed threshold (% of start-ups)

90.48

80.95

85.71

0.678

Information shown if passed threshold (% of

unique emails)

73.24

73.02

72.06

0.987

33

Table 2: Descriptive Statistics of Start-ups

This table shows descriptive statistics of the 21 start-ups in the randomized field experiment at the time of

fundraising. Panel A shows the distribution across cities and countries. Panel B reports the distribution

across sectors, where sectors are not mutually exclusive. Panel C shows the structure of the start-up in

terms of number of founders, employees, board size, advisors, and whether or not the company has an

attorney. Employees (%) is the fraction of start-ups that has non-founder employees. The If > 0, #

employees variable shows how many employees are working for those start-up that have employees. The

variables for board members, advisors and attorney follow a similar pattern. Panel D reports the

percentage of start-ups that had funding prior to the current round (Pre-round funding (%)), and if any

prior money was raised, the amount raised (If > 0, pre-round funding raised). Incubator (%) is the

fraction of start-ups that have been part of an incubator or accelerator program in the past, and Equity

financing (%) is the percentage of firms selling stock, with the remainder selling convertible notes.

Panel A: Start-up Distribution across Cities

N

fraction (%)

Austin, TX

1

4.76

Chicago, IL

1

4.76

Kitchener, Canada

1

4.76

London, United Kingdom

1

4.76

Melbourne, Australia

1

4.76

New York City, NY

3

14.28

San Antonio, TX

1

4.76

Silicon Valley, CA

6

28.57

Singapore

1

4.76

Sydney, Australia

1

4.76

Toronto, Canada

3

14.28

Vancouver, Canada

1

4.76

Panel B: Start-up Distribution across Sectors

N

fraction (%)

Information Technology

18

85.71

Consumers

13

61.90

Clean Technology

1

4.76

Healthcare

3

14.28

Business-to-business

8

38.10

Media

2

9.52

Education

2

9.52

34

Panel C: Start-up Structure

N

mean

st. dev.

percentile

10

50

90

# Founders

21!

2.62

0.92

2

4

Employees (%)

21

80.95

If >0, # employees

17

3.35

2.21

1

3

7

Board members (%)

21

23.81

!

If >0, # board members

5

1.80

0.84

1

2

3

Advisor (%)

21

90.48

!

If >0, # advisors

19

4.74

6.00

1

3

7

Attorney (%)

21

71.43

Panel D: Start-up Funding

N

mean

st. dev.

percentile

10

50

90

Incubator (%)

21

57.14

Pre-round funding (%)

21

52.38

If > 0, pre-round funding raised ($000s)

11

580.95

855.33

50.00

290.00

950.00

Pre-money valuation ($000s)

16

5,465.63

2,133.60

3,000.00

5,000.00

8,000.00

Fundraising goal ($000s)

18

1,183.06

462.88

570.00

1,250.00

2,000.00

Equity financing (%)

21

76.19

35

Table 3: Descriptive Statistics of Investors

This table reports descriptive statistics of the active investors (defined as having requested at least one

introduction through the AngelList platform) who received featured emails about the start-ups in the

randomized field experiment, and opened at least one such email. Panel A shows in which sectors

investors have stated they are interested in investing. A single investor can indicate multiple sectors of

interest. Panel B shows the number of introductions requested by investors, the signal of an investors’

success as computed by AngelList (see the main text for a description of the algorithm), the number of

followers that investors have on the platform, both the raw number and weighted by the followers’ signals,

the percentage of investors that were involved with start-ups in the past, and for those involved with start-

ups, the number of start-ups the investor was involved with. Panel C breaks down these involvements into

various roles. Investor (%) shows the percentage of angels who have invested in start-ups. For the subset

of angels who invested in start-ups, If > 0, # start-ups funded reports the number of start-ups that they

invested in. The variable definitions for advisor, board member, and founder follow a similar pattern.

Panel A: Investor Stated Interest across Sectors

Sector

N

fraction (%)

Information Technology

2,884

98.59

Consumers

2,769

94.66

Clean Technology

861

29.43

Healthcare

1,239

42.35

Business-to-business

2,328

79.58

Finance

949

32.44

Media

1,420

48.54

Energy

165

5.64

Education

685

23.41

Life Sciences

414

14.15

Transportation

307

10.49

Other

26

0.8

Panel B: Investor Characteristics

N

mean

st. dev.

percentile

10

50

90

# Introductions requested

2,925

9.72

31.09

1

3

21

Signal

2,925

6.44

2.26

3.28

6.30

9.87

# Followers

2,925

591.12

1,493.10

26

202

1346

Weighted number of followers

2,925

2,527.30

5,763.70

108.97

915.70

5,896.90

Involved in start-ups (%)

2,925

91.93

If > 0, # start-ups involved with

2,689

12.55

17.18

2

8

27

36

Panel C: Investor Roles in Start-up Companies

N

mean

st. dev.

percentile

10

50

90

Investor (%)

2,925

82.36

If > 0, # start-ups funded

2,409

13.10

16.81

2

8

28

Advisor (%)

2,925

43.49

If > 0, # start-ups as advisor

1,272

3.47

4.54

1

2

7

Board member (%)

2,925

16.92

If > 0, # start-ups as board member

495

1.93

1.82

1

4

Start-up founder (%)

2,925

60.00

If > 0, # start-ups founded

1755

2.05

1.44

1

2

4

Table 4: Investor Response to Randomized Emails

This table reports regression results of investor responses to the featured emails in the randomized field experiment. The dependent variable is one

when an angel investor clicked on the “View” button in the featured email, and zero otherwise. Only opened emails are included in the sample.

Team = 1 is an indicator variable that equals one if the team information is shown in the email, and zero otherwise. Similarly, Investors = 1 and

Traction = 1 are indicator variables for the current investors, and traction information, respectively. Connections counts the number of people on

the start-up’s profile (in any role) that the investor already follows prior to receiving the email. Prior follow = 1 is an indicator variables that

equals one if the investor was already following the start-up on AngelList prior to receiving the featured email. Prior emails is the number of

emails that the investor has received in the experiment prior to the present email. R2 is the adjusted R

2

for OLS regressions, and pseudo R

2

for

logit models. Standard errors are in parentheses, and are clustered at the investor level. ***, ** and * indicate statistical significance at the 1, 5 and

10 percent level, respectively.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Model

OLS

Logit

Team = 1

0.026***

0.028***

0.022**

0.023**

0.193***

0.208***

0.162**

0.172**

(0.009)

(0.010)

(0.064)

(0.073)

(0.074)

Investors = 1

0.011

0.013

0.010

0.009

0.077

0.100

0.070

0.067

(0.010)

(0.013)

(0.072)

(0.073)

(0.097)

Traction = 1

0.010

0.013

0.016

0.017

0.073

0.097

0.122

0.123

(0.010)

(0.014)

(0.075)

(0.076)

(0.106)

Connections

0.016***

0.010

0.093***

0.064*

(0.006)

(0.034)

(0.038)

Prior follow = 1

0.129***

0.143***

0.739***

0.835***

(0.033)

(0.163)

(0.166)

Prior emails

-0.007***

0.001

-0.051***

0.006

(0.002)

(0.003)

(0.014)

(0.022)

Start-up fixed effects

N

Y

N

Y

Number of observations

8,189

8, 189

R2

0.001

0.008

0.001

0.005

0.001

0.009

0.028

0.033

38

Table 5: Investor Response to Non-randomized Emails

This table replicates the regressions in Table 4 for the subset of featured emails that show all information that has crossed the disclosure threshold,

in the columns labeled “Full-information emails only”. The model numbers in the second row correspond to the model numbers in Table 4. For

ease of comparison, the columns labeled “Randomized sample” show the results from Table 4 for the same set of models. The dependent variable

is one when an angel investor clicked on the “View” button in the featured email, and zero otherwise. The explanatory variables are as defined in

Table 4. R2 is the adjusted R

2

for OLS regressions, and pseudo R

2

for logit models. Standard errors are in parentheses, and are clustered at the

investor level. ***, ** and * indicate statistical significance at the 1, 5 and 10 percent level, respectively.

Full-information emails only

Randomized sample

(1)

(2)

(5)

(6)

(1)

(2)

(5)

(6)

OLS

Logit

OLS

Logit

Team = 1

0.046**

0.045**

0.336**

0.337*

0.026***

0.028***

0.193***

0.208***

(0.022)

(0.171)

(0.172)

(0.009)

(0.064)

Investors = 1

0.013

0.022

0.091

0.155

0.011

0.013

0.077

0.100

(0.018)

(0.019)

(0.127)

(0.133)

(0.010)

(0.072)

(0.073)

Traction = 1

0.037*

0.043**

0.265*

0.311**

0.010

0.013

0.073

0.097

(0.020)

(0.149)

(0.154)

(0.010)

(0.075)

(0.076)

Connections

0.010

0.058

0.016***

0.093***

(0.010)

(0.054)

(0.006)

(0.034)

Prior follow = 1

0.150**

0.822***

0.129***

0.739***

(0.059)

(0.277)

(0.033)

(0.163)

Prior emails

-0.006

-0.042*

-0.007***

-0.051***

(0.003)*

(0.025)

(0.002)

(0.014)

Start-up fixed effects

N

Number of observations

2,992

8,189

R2

0.001

0.006

0.002

0.008

0.001

0.008

0.001

0.009

Table 6: Investor Response by Number of Investments

This table reports regression results of investor responses to the featured emails in the randomized field

experiment. The dependent variable is one when an angel investor clicked on the “View” button in the

featured email, and zero otherwise. Only opened emails are included in the sample. Team = 1, Investors =

1 and Traction = 1 are indicator variables that equal one if the team, current investors, or traction

information, respectively, are shown in the email. # Investments <= cutoff is an indicator variable that

equals one if number of investments by a given investor is less than or equal to the percentile of the

investments count distribution shown in the row labeled Cutoff. The other variables are as defined in

Table 4. R2 is the adjusted R

2

for OLS regressions, and pseudo R

2

for logit models. Standard errors are in

parentheses, and are clustered at the investor level. ***, ** and * indicate statistical significance at the 1,

5 and 10 percent level, respectively.

(1)

(2)

(3)

(4)

(5)

(6)

OLS

Logit

Team shown = 1

0.017*

0.021*

0.026**

0.130*

0.162*

0.203**

(0.010)

(0.011)

(0.013)

(0.079)

(0.087)

(0.099)

Investors shown = 1

-0.001

0.004

0.003

-0.012

0.025

0.016

(0.013)

(0.014)

(0.016)

(0.104)

(0.115)

(0.125)

Traction shown = 1

0.009

0.003

0.010

0.066

0.024

0.076

(0.014)

(0.016)

(0.017)

(0.108)

(0.117)

(0.129)

# Investments <= cutoff

0.037

0.007

-0.007

0.235

0.029

-0.061

x Team shown = 1

(0.025)

(0.018)

(0.017)

(0.176)

(0.137)

(0.131)

# Investments <= cutoff

0.070**

0.021

0.014

0.476**

0.146

0.103

x Investors shown = 1

(0.028)

(0.021)

(0.020)

(0.189)

(0.151)

(0.149)

# Investments <= cutoff

0.063**

0.047**

0.015

0.427*

0.351**

0.116

x Traction shown = 1

(0.031)

(0.021)

(0.020)

(0.220)

(0.163)

(0.154)

# Investments <= cutoff

-0.080*

-0.028

-0.001

-0.530*

-0.193

0.001

(0.043)

(0.031)

(0.030)

(0.320)

(0.241)

(0.231)

Connections

0.010

0.066*

0.068*

0.067*

(0.006)

(0.038)

Prior follow

0.145***

0.144***

0.145***

0.847***

0.849***

0.852***

(0.033)

(0.166)

Prior emails

0.002

0.001

0.013

0.014

0.010

(0.003)

(0.022)

Startup fixed effects

Y

Cutoff

Zero

25%

50%

Zero

25%

50%

Number of observations

8,189

R2

0.007

0.006

0.005

0.035

0.034

40

Table 7: Investor Response by Signal

This table reports regression results of investor responses to the featured emails in the randomized field

experiment. The dependent variable is one when an angel investor clicked on the “View” button in the

featured email, and zero otherwise. Only opened emails are included in the sample. Team = 1, Investors =

1 and Traction = 1 are indicator variables that equal one if the team, current investors, or traction

information, respectively, are shown in the email. Signal < cutoff is an indicator variable that equals one

if the investor signal is below the percentile of the signal distribution shown in the row labeled Signal

cutoff. See the main text for the algorithm used to compute the signals. The other variables are as defined

in Table 4. R2 is the adjusted R

2

for OLS regressions, and pseudo R

2

for logit models. Standard errors are

in parentheses, and are clustered at the investor level. ***, ** and * indicate statistical significance at the

1, 5 and 10 percent level, respectively.

(1)

(2)

(3)

(4)

(5)

(6)

OLS

Logit

Team shown = 1

0.019*

0.024*

0.035*

0.147*

0.184*

0.265*

(0.011)

(0.013)

(0.019)

(0.085)

(0.097)

(0.139)

Investors shown = 1

-0.003

0.001

-0.005

-0.036

0.005

-0.045

(0.014)

(0.016)

(0.022)

(0.112)

(0.127)

(0.170)

Traction shown = 1

0.005

0.010

0.009

0.037

0.077

0.067

(0.015)

(0.017)

(0.022)

(0.112)

(0.123)

(0.156)

Signal < cutoff

0.013

-0.003

-0.016

0.067

-0.027

-0.127

x Team shown = 1

(0.020)

(0.017)

(0.020)

(0.145)

(0.131)

(0.154)

Signal < cutoff

0.055**

0.016

0.019

0.385**

0.121

0.141

x Investors shown = 1

(0.022)

(0.020)

(0.024)

(0.159)

(0.151)

(0.184)

Signal < cutoff

0.063**

0.015

0.012

0.440**

0.122

0.093

x Traction shown = 1

(0.024)

(0.020)

(0.023)

(0.180)

(0.155)

(0.171)

Signal < cutoff

-0.049

-0.013

0.001

-0.320

-0.102

0.010

(0.034)

(0.030)

(0.036)

(0.261)

(0.231)

(0.276)

Connections

0.011*

0.010

0.075*

0.066*

0.067*

(0.006)

(0.038)

Prior follow

0.144***

0.145***

0.843***

0.841***

0.846***

(0.033)

(0.166)

Prior emails

0.003

0.001

0.020

0.008

0.006

(0.003)

(0.022)

Startup fixed effects

Y

Signal cutoff

25%

50%

75%

25%

50%

75%

Number of observations

8,189

R2

0.007

0.005

0.036

0.033

0.034

41

Table 8: Investor Response by Number of Followers

This table reports regression results of investor responses to the featured emails in the randomized field

experiment. The dependent variable is one when an angel investor clicked on the “View” button in the

featured email, and zero otherwise. Only opened emails are included in the sample. Team = 1, Investors =

1 and Traction = 1 are indicator variables that equal one if the team, current investors, or traction

information, respectively, are shown in the email. # Followers < cutoff is an indicator variable that equals

one if number of followers of a given investor is less than the percentile of the followers count

distribution shown in the row labeled Cutoff. The other variables are as defined in Table 4. R2 is the

adjusted R

2

for OLS regressions, and pseudo R

2

for logit models. Standard errors are in parentheses, and

are clustered at the investor level. ***, ** and * indicate statistical significance at the 1, 5 and 10 percent

level, respectively.

(1)

(2)

(3)

(4)

(5)

(6)

OLS

Logit

Team shown = 1

0.017

0.023*

0.036**

0.135

0.182*

0.293**

(0.011)

(0.013)

(0.017)

(0.085)

(0.104)

(0.141)

Investors shown = 1

-0.007

-0.018

-0.064

-0.163

-0.155

(0.013)

(0.016)

(0.020)

(0.109)

(0.127)

(0.164)

Traction shown = 1

0.006

0.003

0.010

0.046

0.022

0.078

(0.015)

(0.017)

(0.021)

(0.114)

(0.127)

(0.158)

# Followers < cutoff

0.021

-0.001

-0.018

0.115

-0.031

-0.163

x Team shown = 1

(0.020)

(0.017)

(0.020)

(0.144)

(0.132)

(0.158)

# Followers < cutoff

0.064***

0.053***

0.035

0.443***

0.412***

0.277

x Investors shown = 1

(0.024)

(0.020)

(0.022)

(0.165)

(0.151)

(0.177)

# Followers < cutoff

0.049**

0.030

0.010

0.324*

0.232

0.081

x Traction shown = 1

(0.025)

(0.020)

(0.022)

(0.174)

(0.156)

(0.177)

# Followers < cutoff

-0.041

-0.023

0.014

-0.247

-0.163

0.132

(0.034)

(0.029)

(0.034)

(0.252)

(0.230)

(0.277)

Connections

0.013**

0.088**

0.089**

0.085**

(0.006)

(0.040)

(0.039)

Prior follow

0.145***

0.146***

0.145***

0.855***

0.860***

0.854***

(0.033)

(0.165)

(0.166)

Prior emails

0.003

0.001

0.020

0.008

0.004

(0.003)

(0.022)

(0.021)

Startup fixed effects

Y

Cutoff

25%

50%

75%

25%

50%

75%

Number of observations

8,189

R2

0.008

0.007

0.037

0.036

42

Table 9: Investor Response by Weighted Number of Followers

This table reports regression results of investor responses to the featured emails in the randomized field

experiment. The dependent variable is one when an angel investor clicked on the “View” button in the

featured email, and zero otherwise. Only opened emails are included in the sample. Team = 1, Investors =

1 and Traction = 1 are indicator variables that equal one if the team, current investors, or traction

information, respectively, are shown in the email. Weighted # followers < cutoff is an indicator variable

that equals one if number of followers of a given investor, weighted by their signal, is less than the

percentile of the weighted followers count distribution shown in the row labeled Cutoff. The other

variables are as defined in Table 4. R2 is the adjusted R

2

for OLS regressions, and pseudo R

2

for logit

models. Standard errors are in parentheses, and are clustered at the investor level. ***, ** and * indicate

statistical significance at the 1, 5 and 10 percent level, respectively.

(1)

(2)

(3)

(4)

(5)

(6)

OLS

Logit

Team shown = 1

0.020*

0.026**

0.033*

0.158*

0.207**

0.268*

(0.011)

(0.013)

(0.017)

(0.085)

(0.104)

(0.140)

Investors shown = 1

-0.003

-0.017

-0.013

-0.036

-0.156

-0.111

(0.013)

(0.016)

(0.020)

(0.109)

(0.127)

(0.164)

Traction shown = 1

0.006

0.003

0.014

0.048

0.023

0.112

(0.015)

(0.017)

(0.021)

(0.113)

(0.126)

(0.158)

Weighted # followers < cutoff

0.009

-0.008

-0.014

0.035

-0.082

-0.129

x Team shown = 1

(0.020)

(0.017)

(0.020)

(0.144)

(0.132)

(0.157)

Weighted # followers < cutoff

0.051**

0.051***

0.028

0.354**

0.399***

0.221

x Investors shown = 1

(0.024)

(0.020)

(0.022)

(0.165)

(0.152)

(0.176)

Weighted # followers < cutoff

0.049**

0.030

0.003

0.338*

0.229

0.023

x Traction shown = 1

(0.024)

(0.020)

(0.022)

(0.177)

(0.157)

(0.176)

Weighted # followers < cutoff

-0.035

-0.018

0.021

-0.216

-0.117

0.183

(0.034)

(0.029)

(0.034)

(0.254)

(0.231)

(0.276)

Connections

0.013**

0.083**

0.090**

0.084**

(0.006)

(0.039)

(0.040)

(0.039)

Prior follow

0.145***

0.146***

0.145***

0.855***

0.861***

0.854***

(0.033)

(0.166)

Prior emails

0.002

0.001

0.018

0.008

0.005

(0.003)

(0.022)

(0.021)

Startup fixed effects

Y

Cutoff

25%

50%

75%

25%

50%

75%

Number of observations

8,189

R2

0.007

0.008

0.006

0.036

0.037

0.035

Table 10: Start-ups in Field Experiment Sample versus Broad Sample

This table compares the sample of 21 start-ups in the randomized field experiment (the “experiment firms”) with a broad sample of 5,538 firms

raising funding on AngelList (the “non-experiment firms”). The non-experiment firms are those firms that attempted to raise money through

AngelList and received at least one introduction request. The variables are as defined in Table 2. The rightmost column shows the p-value for a

differences-in-means test between the experiment and non-experiment samples.

Experiment firms (N = 21)

Non-experiment firms (N = 5,538)

Means

N

mean

median

st. dev.

N

mean

median

st. dev.

test p

# Founders

21

2.62

2

0.92

5,538

2.11

2

1.06

0.028

Employees (%)

21

80.95

5,538

52.56

0.009

If > 0, # employees

17

3.35

3

2.21

2,911

2.91

2

2.45

0.453

Board members (%)

21

23.81

5,538

16.78

0.390

If > 0, # board members

5

1.80

2

0.84

929

1.96

2

1.14

0.749

Advisor (%)

21

90.48

5,538

60.74

0.005

If > 0, # advisors

19

4.74

3

6.00

3,364

2.94

2

2.18

0.000

Incubator (%)

21

57.14

5,538

29.70

0.006

Pre-round funding (%)

21

47.62

5,538

45.76

0.865

If > 0, amount raised ($000s)

10

605.05

234.00

897.66

2,534

674.27

250.00

1,874.28

0.904

Pre-money valuation ($000s)

12

5,579.17

5,000.00

2,383.22

2,616

4,857.83

3,500.00

15,747.91

0.873

Fundraising goal ($000s)

15

1,226.33

1,325.00

488.96

4,321

923.99

500.00

1,135.56

0.303

Equity financing (%)

21

76.19

4,912

69.04

0.603

Figure 1: Sample featured start-up email to investors

This figure shows an example of a featured start-up email that is sent to investors. Each featured start-up

has up to three information categories (team, traction, and current investors) that would normally be

shown in the email if the information for that category reaches a threshold as defined by AngelList. For

each start-up, various unique versions of each email are generated that randomly hide these pieces of

information (the Randomization categories). Each email contains a “View” button that, when clicked,

takes the investor to the AngelList platform where more information about the company is shown, and

introductions to the company’s founders can be requested. The “Get an Intro” button requests such an

introduction straight from the email.

Randomization