Date post: | 30-Apr-2018 |
Category: |
Documents |
Upload: | trinhkhuong |
View: | 214 times |
Download: | 1 times |
WHAT DRIVES THE TOP LINE? THE DETERMINANTS OF SALES REVENUE
IN VENTURE-BACKED PRE-IPO FIRMS
John R. M. Hand Tel: (919) 962-3173 Kenan-Flagler Business School Fax: (919) 962-4727 UNC Chapel Hill [email protected] Chapel Hill, NC 27599-3490
ABSTRACT Generating and then growing sales revenue is vital for the survival and success of all businesses, particularly start-ups. This study examines the drivers of revenue in young companies using proprietary data from a large and recent cross-section of U.S. venture-backed pre-IPO firms. Based on a Cobb-Douglas-type model of revenue production, I predict and find that a combination of financial and non-financial information explains firms’ own forecasts of one-year-ahead sales and sales growth. Revenues are projected to be larger the higher are current revenues, the more rapid is the expected growth in firms’ patents and total labor force, the more business development, finance, marketing, sales, technical, and ‘other’ staff are employed, and when a formal sales commission plan is in place. Projected revenues are lower when firms are in the development or clinical trials stages of life. Because they seem to reflect economically sensible financial and non-financial fundamentals, I conclude that the top-line forecasts made by young entrepreneurial companies are not necessarily the ‘creative fiction’ as is often alleged. Key words: Employees; incentive plans; revenue forecasts; revenue growth; venture-
backed pre-IPO companies.
JEL classifications: D24, J24, J41, L25, M13, M41.
First draft: May 28, 2005
Data availability: The data in this study were released to the author under a proprietary and confidential agreement with VentureOne.
I am very grateful to B. Hughes and VentureOne for generously granting me access to their data. Financial support from the Edward O’Herron, Jr. Fund for Distinguished UNC Faculty is also appreciated. The usual disclaimer applies.
1. Introduction The goal of this paper is to investigate the economic determinants of sales revenue and
top-line growth. First establishing and then increasing revenue is vital for corporate longevity
and the creation of wealth because it is the main channel through which businesses transform
assets and growth opportunities into cash.
I focus on venture-backed companies because sales revenue is especially crucial to the
survival and success of such firms. Unlike mature companies that typically have multiple
product lines, a diversified customer base and low external financing needs, entrepreneurial
companies—particularly the young high-tech firms that receive the vast majority of venture
funding—have few (if any) products, sell to a concentrated set of customers, and consume large
amounts of outside equity capital. This is likely to make the marginal impact of new or lost
revenue more significant for them than for their publicly traded cousins.
The central hypothesis that I develop is that the nature of entrepreneurial firms’
production functions means that both financial and non-financial factors affect the level and rate
of growth in their sales revenue. The dataset I use to test this hypothesis is a large cross-section
of confidential one-year-ahead revenue forecasts made by firms that responded to VentureOne’s
Spring 2004 and 2005 surveys of U.S. venture-backed pre-IPO companies, together with firm-
specific characteristics that were collected as part of the surveys.
Like seasoned firms, entrepreneurial businesses combine capital and labor to create
economic value. However, the capital stock in a venture-backed company is far more heavily
tilted toward intangible assets such as novel ideas, intellectual property, patents and R&D
breakthroughs than it is toward tangible assets. In addition, the labor required in a high-tech
startup is frequently knowledge-based, complementary to the firm’s intangible assets (rather than
being a substitute), and highly incentivized so as to overcome agency problems.
I model these features using a multi-input Cobb-Douglas-type revenue production
function in which a firm’s sales revenue depends on its tangible and intangible assets, life-stage,
the number of personnel employed in specialized capacities, the revenue-oriented incentives it
provides, and its life-stage. Because the VentureOne database does not contain information
about firms’ tangible assets, I substitute tangible assets out via firms’ lagged revenues and the
assumption that tangible assets evolve in an exponential manner. This yields a reduced form in
which one-year-ahead revenue depends on both financial and non-financial drivers. Specifically,
2
I predict that one-year-ahead revenues are an increasing function of current period revenues, the
current level and the one-year-ahead rate of growth in intangible assets, the one-year-ahead rate
of growth in total employees, the number of personnel employed in business development,
finance, marketing, sales and technical functions, and the presence of a formal sales commission
plan. One-year-ahead revenues are also predicted to be unrelated to the number of
administrative employees, and lower in the development and clinical trials stages of life.
Empirical test results appear strongly consistent with these predictions. For the set of
1,138 venture-backed companies in VentureOne’s Spring 2004 and 2005 surveys that have
sufficient data, I find that projected revenues are indeed larger the higher are firms’ current
revenues, the more rapidly firms are growing their patents and expect to grow their total labor
force, the more business development, finance, marketing, sales, and technical staff they employ;
and when a formal sales commission plan is in place. Projected revenues are also uncorrelated
with the number of administrative employees, but are reliably lower during the development and
clinical trials stages of firms’ lives. In all, the model explains 81% of the cross-sectional
variation in firms’ forecasted revenues. Of the regression total adjusted R2, 13% is unique to
financial information, 6% is unique to non-financial information, and 62% is common to both.
I further estimate the revenue model separately for each of VentureOne’s three key
industry-defined sectors: Healthcare, Information Technology, and Retail/Consumer/Business
Products/Services. Consistent with the conjecture that a more rapidly growing workforce
generates higher sales, in all three industries I find that one-year-ahead revenue forecasts are
significantly positively related to the one-year forecasted growth rate in the total number of
employees. However, supporting the view that not all employees are equal with regard to their
role in generating revenue, and that the value of specialized labor is higher the more intangible-
intensive is a firm, the explanatory power of specialized job functions is higher in the Healthcare
and Information Technology industries than in Retail/Consumer/Business Products/Services
sector. The reverse is true for non-specialized or ‘Other’-categorized personnel, where
forecasted revenues are reliably increasing in the number of ‘Other’-categorized employees for
Retail/Consumer/Business Products/Services but are unrelated to forecasted revenues for
Healthcare and Information Technology firms. Finally, in agreement with the proposition that
intangible assets play a larger role in generating revenue the more intangible-intensive is the
3
firm, revenues and revenue growth are increasing in the rate of patent growth for Healthcare and
Information Technology firms but not for Retail/Consumer/Business Products/Services firms.
Overall, this paper contributes to the accounting and entrepreneurial economics
literatures in two main ways. First, I add to what is known about the relevance of non-financial
information in predicting a key corporate financial performance measure, namely sales revenue.
Prior research using publicly traded firms has shown that the level and growth in revenues are
associated with order backlog (Liu, Livnat and Ryan, 1996), customer satisfaction (Ittner and
Larcker, 1998), operational measures of product quality (Nagar and Rajan, 2001), internet traffic
(Trueman, Wong and Zhang, 2001), and network effects (Rajgopal, Venkatachalam and Kotha,
2003). To this set of variables I add both the growth in total employees and the types of human
capital skills that employees currently possess. Although the growth in total employees is often
computable from publicly available disclosures for publicly traded firms, the breakdown of total
employees according to job function is not. Subject to the normal problems of generalizability,
venture-backed pre-IPO companies therefore offer a window into firms’ revenue productions
that publicly traded firms do not. I argue that confirming the importance of employee-related
variables is particularly noteworthy because firms often claim that their employees are their most
important assets, especially during the start-up phase of their life cycle. I also demonstrate that
employees and their human capital skills play differentiated roles in growing young firms. For
example, business development, sales, and marketing employees contribute to future revenues
but administrative personnel do not.
The study adds to the literature in entrepreneurial economics by enlarging our knowledge
of venture-backed firms. As noted by Gompers and Lerner (2000) and the National Venture
Capital Association (2004), venture-backed firms are becoming an increasingly important force
in the modern U.S. economy. A number of studies have identified variables that explain venture-
backed firms’ equity values and returns (Gompers and Lerner, 1997, 1998, 2000; Lerner, 1994;
Seppä, 2003; Hand, 2004, 2005a; Armstrong, Davila and Foster, 2005), while other research has
sought to appreciate the varied roles that accounting systems and financial information play in
the business development and maturation of venture-backed companies (Davila, 2004; Davila
and Foster, 2005) and growth in employees, average salaries and turnover around funding dates
(Davila, Foster and Gupta, 2003). My study complements this stream of literature by focusing
on the heretofore unexamined but undeniably relevant driver of profitability and equity value,
4
namely sales revenue. In doing so, I show that the revenue forecasts made by venture-backed
pre-IPO companies are not necessarily the ‘creative fiction’ that is sometimes alleged. Rather,
they sensible reflect a variety of economically fundamental financial and non-financial drivers.
The remainder of the paper proceeds as follows. In Section 2, I summarize the literature
that bears on the research questions examined in this paper. In Section 3, I describe the
VentureOne survey dataset. In Section 4, I put forward a Cobb-Douglas-type revenue
production model tailored to venture-backed companies and use it to develop predictions about
which variables in the VentureOne dataset will explain variation in firms’ projected one-year-
ahead revenue growth. Section 5 reports the results of several empirical analyses that test these
predictions. Section 6 notes the major limitations of the study, while Section 7 concludes.
2. Literature review Academics and practitioners alike agree that sales revenue it the most important driver of
corporate value (Holliday, 2000; Wall Street Journal, C1, Sept. 25, 2000; Lundholm and Sloan,
2004, p.144; Penman, 2004, p.402; Ghosh, Gu and Jain, 2005). However, the goal of this
paper—understanding the economic determinants of sales revenue and top-line growth in young
firms—has been the subject of limited academic study. There are two streams of research that
provide relevant contextual background for my paper. The first are studies in accounting that
investigate the relevance of non-financial information for predicting financial outcomes, and the
second is research in entrepreneurial economics into venture-backed pre-IPO companies.
During the past ten years, several accounting scholars have studied the relevance or lack
thereof that non-financial information has for predicting financial performance in publicly traded
firms. Specifically with respect to sales revenue, this work suggests that the level and rate of
growth in revenues are associated with order backlog (Liu, Livnat and Ryan, 1996), customer
satisfaction (Ittner and Larcker, 1998), operational measures of product quality (Nagar and
Rajan, 2001), internet traffic (Trueman, Wong and Zhang, 2001), and network effects (Rajgopal,
Venkatachalam and Kotha, 2003). A key objective of my paper is to explore the relevance of
new non-financial variables, particularly the quantity of employees and the types of human
capital skills that they possess. I argue that such employee-related variables are important to
understand because firms frequently claim that their employees are their most important assets,
especially during the start-up phase of their life cycle (Davila, Foster and Gupta, 2003).
5
The second stream of research that bears on this paper is the work on venture-backed
companies has developed over the past decade as an indirect result of the heightened public
prominence of and interest in venture capital funds.1 Firms that are funded by venture capital are
economically quite different from the majority of publicly traded companies. First and foremost,
venture-backed firms are chiefly technology- or innovation-driven young private companies in
the most formative years of their life. They derive the lion’s share of their value comes from the
technology intensive growth options that dominate their business models and production
functions. As a result, venture-backed firms tend to be more innovative and produce more, and
more valuable, patents (Kortum and Lerner, 2000). Second, the dominant role played by their
highly heterogeneous types of intangibles creates large agency costs and information
asymmetries between firms and potential suppliers of capital (Gompers and Lerner, 2000; Lev,
2001). Related to this, Hand (2005b) argues that this leads to venture-backed pre-IPO companies
granting employee stock options very deeply into their organizations. Third, venture-backed
firms are more dynamic in the sense that they typically grow in a rapid, flexible and professional
manner (Hellman, 2000; Davila, Foster and Gupta, 2003; Davila and Foster, 2005). They tend to
uncover new investment opportunities faster, to develop them quicker and more professionally,
to engage in cooperative commercialization strategies such as strategic alliances and technology
licensing, and to be more ready to discard or abandon bad projects than more seasoned
companies (Hellman and Puri, 2000; 2002; Hsu, 2004a). Fourth, firms obtain quicker venture
backing when the entrepreneur has previous start-up experience or ties to venture capitalists
(Hsu, 2004b). By virtue of the novel dataset used in this study, my work adds to this literature
by measuring the importance of employees and the types of human capital skills they possess.
3. Data
The data in this study come predominantly from proprietary surveys undertaken by
VentureOne in Spring 2004 and Spring 2005.2 At those times, VentureOne emailed a web-based
compensation questionnaire to each of the roughly 5,000 venture-backed firms in its financing
1 An excellent summary of venture capital fund research is provided by Gompers and Lerner (2000a), spanning topics such as the compensation of venture capitalists (Gompers and Lerner 1999); the optimal investment, monitoring and staging of venture capital (Gompers 1995); the decision to go public (Lerner 1994b); and the long-run performance of venture-backed IPOs (Brav and Gompers 1997). 2 The author was granted access to VentureOne’s data only after signing a non-disclosure agreement.
6
database that it classified as still being private and independent. The questionnaire asked firms
to provide information on a broad set of compensation- and business-related items. For example,
companies were asked to report the dollar values of the base salary, bonus, and other cash
compensation of every employee (up to a maximum of 50 people from the most senior person
down); the number of employees that receive stock options; the total number of shares of
founder’s stock and exercised and unexercised options that each held; and the total number of
both fully diluted and common shares that the firm had outstanding. Of primary interest to this
study, VentureOne also asked each firm to report its actual revenues for its most recently
completed fiscal year; its forecast of one-year-ahead revenues; the number of employees at the
end of the most recently completed fiscal year, both in total and by job function (administrative,
business development, finance, marketing, sales, technical, and ‘other’); and the total number of
employees it expected to have at the end of its next fiscal year.3
The data from VentureOne’s surveys were then merged with VentureOne’s financing and
general support databases and patent data. VentureOne’s financing database contains a record of
each firm’s equity financing history, where available. For each round of funding, the financing
database reports the amount of money raised, when the round closed, the pre- and post-money
valuations, the type of round (e.g., First, Second, Individual Investor), the firm’s business status
(e.g., Startup, Product Development, Shipping), and the ID code and type of each investor that
participated in the round. The general support database contains general information about each
firm, such as its industry, state, and telephone area code, as well as details on current and former
senior management and Board members, such as their title, type (e.g., outside Board member,
venture investor Board member) and whether they are or were one of the firm’s founders. The
number of patents granted to each firm at June 30 of the most recently completed fiscal year as
well as June 30 of the forecasted year ahead were obtained from www.uspto.gov.4
Table 1 lists the restrictions that I placed on VentureOne’s databases in order to be able to
estimate a Cobb-Douglas-type production function model of firms’ sales revenue. Of the
roughly 5,000 venture-backed firms to whom VentureOne emailed its compensation survey in
3 A full listing of the items requested from survey participants is available from the author upon request. 4 In this draft, I use the number of patents granted at May 14, 2005 rather than June 30, 2005. The next (post June 30, 2005) draft of the paper will remove this approximation.
7
Spring 2004 and 2005, a total of 1,396 responded.5 Of these, 42 were eliminated because they
had been acquired or merged, were already public or in IPO registration, or had gone out of
business. This yielded 1,254 firms that were truly private and independent venture-backed pre-
IPO companies. Untabulated analysis indicated that respondents were not significantly different
from non-respondents.6 From this set, 116 firms were excluded because they were missing one
or more data items needed to estimate the revenue model. Missing data items included the lack
of actual revenues, forecasted revenues, the number of sales personnel, the firm’s life-stage,
whether a formal sales commission plan was in place, the number of employees at the time of the
survey, or the number of patents granted. These sample selection criteria resulted in a set of
1,138 U.S. venture-backed pre-IPO firm-years contributed by 970 different firms.
Descriptive statistics for this sample are shown in Table 2. Panel A reports the minimum,
mean, maximum and standard deviation of financial and non-financial characteristics that I
model in Section 4 as being direct or indirect candidates for explaining variation in firms’
revenues. Panel B reports general firm characteristics, some of which are employed in
robustness tests the results of which are summarized in Section 5.
The main financial variable of interest is revenue. In responding to VentureOne’s survey,
firms were asked to select a category for actual revenues in the most recently completed fiscal
year and estimated revenues for the next fiscal year. The categories were as follows, with the
reported revenue ranges shown in parentheses: 0 ($0-$0.5 mil.), 1 ($0.5-$1.0 mil.), 2 ($1.0-$2.0
mil.), 3 ($2.0-$3.0 mil.), 4 ($3.0-$5.0 mil.), 5 ($5.0-$10 mil.), 6 ($10-$20 mil.), and 7 (> $20
mil.). For revenue categories 0-6, I converted each firm’s responses into dollars using the
midpoint of the range (e.g., a category 5 response was recoded as $7.5 mil.). Revenue category 7
5 The annual response rate of approximately 20% compares favorably with other compensation surveys. For example, in a study of companies with broad-based stock option plans, Weeden et al. (2001) report a response rate of approximately 10%. 6 I developed and estimated a logistic regression model to discriminate between respondents and non-respondents. Using data available in VentureOne’s aggregate financing and valuation database (which did not include compensation data), less than 1% of the variance could be explained. The independent variables used were the age of the firm, the date of its most recent round of equity financing, the number of its most recent round of equity financing (e.g., 1st, 2nd, etc.), the amount raised in that round, the state in which it is headquartered, its life-stage (e.g., start-up, product development, beta testing, shipping, profitable, clinical trials, and restart), and the industry sector the firm was in. This analysis notwithstanding, in private correspondence VentureOne indicated that it is their belief that it is firms that anticipate seeking further funding from venture capital funds that are most likely to respond to the survey.
8
was coded as $60 mil.7 Panel A reports that the mean revenue categories for actual and
forecasted revenues are 2.5 and 3.7, respectively. Mean dollar revenues are $7.5 mil. and $13.0
mil., respectively. As suggested by these figures, venture-backed firms believe that their
revenues will grow rapidly. This reinforces a key presumption of this study, namely that
revenues and revenue growth are key performance measures for these companies. The mean log
growth rate in revenues is 0.83 (i.e., 129%) and the unreported median percentage growth rate is
100%. Moreover, since firms know that VentureOne will not make their revenue projections
public, they have no incentive- or strategic-based reason to upwardly bias their projections.8
Panel A also reports summary statistics spanning four types of potentially relevant non-
financial information available in the augmented VentureOne dataset: intangible assets, human
capital and employee labor, incentives, and investment intensity. The most detailed of these is to
do with human capital and employee labor. In VentureOne’s survey, firms were asked for their
actual total headcount at the end of their most recently completed fiscal year and their estimate of
what their headcount would be at the end of the next fiscal year. The former was coded as both a
number and a category, while the latter was only coded as a category. The categories were as
follows, with the reported ranges shown in parentheses: 0 (0-10 employees), 1 (10-20), 2 (20-
30), 3 (30-40), 4 (40-50), 5 (50-60), 6 (60-100), and 7 (> 100). For headcount categories 0-6, I
converted each firm’s responses into a number using the midpoint of the range (e.g., a category 4
response was recoded as 45 employees). Headcount category 7 for projected headcount was
coded as 360, the mean headcount for category 7 firms in their most recently completed fiscal
year. Panel A shows that the mean actual and forecasted headcounts are 60 and 86 employees,
respectively. The mean log change in forecasted headcount is 0.32 (i.e., 38%) and the
unreported median percentage growth rate is 25%. Firms therefore expect to grow their
workforce quite rapidly, although at a markedly slower rate than revenues.
7 $60 mil. is the mean revenue of the subset of firms with category 7 revenues in the most recently completed fiscal year that had revenues for that year reported in VentureOne’s Support Data supplementary file. Similar empirical results obtain if category 7 revenues are coded as anywhere between $30 mil. and $100 mil., most likely because of the relatively few observations with category 7 actual or forecasted revenues and the log-linear nature of the regression model that I estimate. 8 This does not preclude forecasts being upward biased for imperfectly rational behavior reasons. I assess the degree of ex-post forecast bias for the small subset of firms that responded to both the 2004 and 2005 VentureOne surveys in section 6.
9
Panel A additionally indicates that there is substantial variation in the types of human
capital skills that employees hold in startups, as judged by job functions. The mean number of
employees ranges from 1.1 in business development to 9.5 in sales to 21 in ‘Other’-categorized
positions. There is also substantial variation in the number of patents granted to firms, my proxy
for intangible assets. Midway through the forecasted (prior) fiscal year, the mean number of
patents that had been granted to sample firms was 2.1 (1.6), but the maximum was 95 (94). Of
firms, 64% have a formal sales commission plan in place, 24% are in the Development life-stage,
and 6% are conducting Clinical Trials.
Panel B reports summary statistics pertaining to general firm characteristics. Sample
firms are on average 6.3 years old and half the time are headquartered in either California or
Massachusetts. Sample firms have completed an average of 3.2 rounds of financing through
which they have raised $21 million from venture funds. Their post-money valuation at their
most recent financing round averaged $41 million and an average of 62% of firms’ equity is
owned by venture investors. Perhaps because sample firms extend employee stock options to an
average of 88% of their employees, workforce turnover averages 13%. Roughly half of sample
firms still have a founder in the CEO position.
4. Revenue production model
Young entrepreneurial businesses combine capital and labor to create economic value.
However, compared to a mature company, the capital stock in a new business—particularly a
high-tech firm of the type that receives the vast majority of venture funding—consists mostly of
intangible assets such as an innovative business model, intellectual property, patents and R&D
investments rather than tangible assets such as PP&E. It is also the case that the human capital
needed in a venture-backed startup is typically knowledge-based, potentially complementary to
the firm’s intangible assets (rather than a substitute), and, in order to overcome agency issues,
highly incentivized.
Following the example of Lev and Radhakrishnan (2004), I model these economic
features beginning with a multi-input Cobb-Douglas-type sales revenue production function.9
9 For simplicity I ignore interest revenue. Although interest revenue can be large (and larger than sales revenues) for venture-backed firms, the firm is a price-taker in the money markets and therefore cannot expect to create economic value for its shareholders with regard to interest revenue. In contrast, the whole reason to start a business is to develop a new product or service that in expectation allows the firm
10
11
First, I assume that the revenue of firm i in year t, REVit, is a stochastic log-linear function of its
average tangible capital, TKit, intangible capital, IKit, and total labor force, Lit during year t:
iteLIKTKREV itititiit
εδβαφ= , where { } 0,,,,,, >LIKTKδβαφ . (1) I do not restrict the production function to have constant returns to scale. The log of revenue in
t+1 is then given by:
1,1,1,1,1, lnlnlnlnln +++++ ++++= titititiiti LIKTKREV εδβαφ (2) As noted in section 3, the VentureOne database does not contain information about firms’ assets
or shareholder equity. I finesse this problem by modeling the dynamics of tangible capital as:
θ
itti TKTK =+1, (3) Equation (3) permits TKi,t+1 to be expressed in terms of REVit, IKit and Lit and therefore
substituted out in equation (2).
Because venture-backed firms typically convert the external equity capital they raise
from venture capitalists into intangible capital (say, by spending the cash on R&D or patenting)
before they subsequently convert it back into tangible capital (say, by making large profits on
their patented drugs), I propose that tangible capital depreciates, viz., 10 << θ .10 I also assume
that the number of patents granted to a firm, NPATit, is a good proxy for its intangible capital:
γ)1( itit NPATIK += (4) Taken together, equations (1) – (4) yield the following expression for one-year-ahead revenues:
1,1,
1,1,
ln)1(ln
)1ln()1(1
1lnlnln)1(ln
++
++
+−+⎥⎦
⎤⎢⎣
⎡+
+−+⎥⎦
⎤⎢⎣
⎡+
+++−=
tiitit
ti
itit
tiititi
LL
L
NPATNPAT
NPATREVREV
εθδδ
θβγβγθφθ
(5)
to be a price-maker and thereby secure positive abnormal profits. Sales revenue is therefore more indicative of the firm’s potential to create positive abnormal profits than is interest revenue. 10 In a mature firm that is profitable, it might be reasonable to expect that θ > 1.
12
Equation (5) says that firms’ one-year-ahead revenues are not just positively auto-correlated—
they are also increasing in the rate of growth in patents and total employees, and the current
stock or level of patents and employees.
I further expand equation (5) by making five refining assumptions. First, I propose that
the contribution to revenues of firms’ M different types of employees {EMP1, EMP2, … EMPM}
can be modeled as:
∏=
+=M
mmitit
mEMPL1
)1( νδ ψ , where ∑=
=M
mmitit EMPL
1. (6)
Second, venture-backed firms tend to be plagued by high agency costs due to incentive
and information problems between entrepreneurs, managers and outside investors (Gompers and
Lerner, 2000). A common response to these agency concerns is that venture investors provide
management strong incentives to build equity value rather than consume perquisites. Since early
and rapid revenue growth is frequently a necessary condition for creating equity value in young
companies, I add a dummy variable DUMSPLANi to equation (5), where DUMSPLANi = 1 if the
firm has a sales commission plan in place and zero otherwise.
Third, I accommodate the fact that the level and growth in revenues will depend on the
degree to which a firm is converting tangible capital into longer-payback intangible capital rather
than shorter-payback sales. That is, all else held constant, revenues are likely to be low when the
firm is in the Development or Clinical Trials stage of its life (assuming that these stages apply to
the company) because in those stages the firm is deliberately taking the cash it has raised from
venture investors and investing it into long-term intangible assets such as R&D and patents. I
therefore add two dummy variables to equation (5)—DUMDEVTi and DUMCLINi. The former
is set to one if VentureOne classifies the firm as being in the Development life-stage, and zero
otherwise. The latter is set to one if according to VentureOne the firm is conducting Clinical
Trials, and zero otherwise. Fourth, because nominal revenues will increase with inflation, I
include a dummy variable DUM2005i set to one if the VentureOne survey year is 2005 rather
than 2004. The inclusion of DUMSPLANi, DUMDEVTi and DUMCLINi and DUM2005i is
accomplished by assuming that the intercept term φi in equation (5) can be rewritten as:
iiii DUMDUMCLINDUMDEVTDUMSPLAN
i eeee 20054321 ηηηηλφ = (7)
13
Finally, I assume firms’ forecasts of future revenues, FREV, and total future employees,
FL, are rational in the sense that:
1,1,1, lnln +++ += tititi REVFREV ς (8) 1,1,1, lnln +++ += tititi LFL ω (9)
Combining equations (5) – (9) yields the final model that I test using VentureOne’s
survey data:
1,321
1
1,
1,41,
)1()1()1(
)1ln()1(ln)1ln()1(
1
1lnln2005)1(ln)1(ln
+
=
+
++
+−+−+−+
+−+⎥⎦
⎤⎢⎣
⎡++−+
⎥⎦
⎤⎢⎣
⎡+
+++−+−=
∑
tiiii
mit
M
m
m
it
tiit
it
tiititi
DUMCLINDUMDEVTDUMSPLAN
EMPL
FLNPAT
NPATNPAT
REVDUMFREV
µηθηθηθ
νθδθβγ
βγθηθλψθ
(10)
where 1,1,1,1, ++++ −−−= titiittiti ωδςθεεµ . The empirical analog of equation (10) that I estimate is:
1,321
1
1,
1,41,
)1ln(ln )1ln(
11
lnln2005ln
+
=
+
++
++++
++⎥⎦
⎤⎢⎣
⎡+++
⎥⎦
⎤⎢⎣
⎡+
++++=
∑
tiiii
mit
M
mm
it
tiit
it
tiititi
DUMCLINhDUMDEVThDUMSPLANh
EMPgL
FLfNPATd
NPATNPAT
cREVbDUMhaFREV
µ
(11)
The assumptions used to derive equation (10) lead to the predictions that a, b, c, d, f, h1 and h4 in
equation (11) will be positive but h2 and h3 will be negative. Equation (10) also implies that
d = c (1 – b).
I further propose that the signs expected for g1, g2, … gM will vary depending on the job
function involved. As highlighted in Table 2, the VentureOne survey asks firms to report the
number of employees in seven job functions. In terms of equation (11), M = 7. I denote
VentureOne’s job functions as follows: Administration (m = 1), business development (m = 2),
finance (m = 3), marketing (m = 4), sales (m = 5), technical (m = 6), and non-categorized ‘other’
(m = 7). I propose that g1 will be zero because all else held equal an incremental person
employed in administration will increase expenses but will have no effect on revenues.11 The
nature of business development, marketing and sales job functions dictate that g2, g4 and g5
should be positive. The importance of raising external equity capital, the importance of
managing internal cash flows and the more immediate sales payback of technical activities in
venture-backed firms (which are often high-technology centered) suggest that g3 and g6 will be
positive. Lastly, without knowing the activities of those employed in ‘other’ job functions, I do
not make a sign prediction for g7. What can reasonably be supposed, however, is that by virtue
of being a catchall category rather than a specific job function, g7 < {g2, g3, g4, g5, g6}.
5. Results 5.1 Simple correlations
Table 3 reports the Pearson and Spearman correlations between firms’ log transformed
one-year-ahead revenue forecasts, denoted lnFREV, and the explanatory variables detailed in
equation (11). Pearson (Spearman) correlations are shown above (below) the diagonal. Because
the log-transformation results in very similar Pearson and Spearman correlations, I discuss only
the former.
As seen in the first row of table 3, every explanatory variable is reliably correlated with
lnFREV, although the signs of three correlations are opposite to those predicted by equation (11).
Specifically, the correlations between lnFREV and lnPATGROWTH and between lnFREV and
ln(1 + NPAT) are reliably negative, not positive, while the correlation between lnFREV and ln(1
+ EMPADMIN) is reliably positive, not zero. However, the signs predicted by equation (11) are
partial correlations, not simple correlations, so such results are only suggestive, not definitive.
5.2 Regressions
The results of estimating equation (11) using OLS are reported in Tables 4 and 5. Table
4 presents the parameter estimates of three models. Model #1 includes all the financial and non-
financial variables; model #2 includes only financial variable, sales revenue; and model #3
includes only the non-financial variables. In all three models in Table 4 observations are pooled
11 Although obvious, it bears noting that there is nothing inconsistent between administrative personnel making no contribution to revenues and firm value-maximization. This is because administrative personnel almost certainly add value by controlling costs and facilitate the smooth internal running of the firm.
14
across firms from different industries. Table 5 estimates equation (11) separately for each of
VentureOne’s three key industry-defined sectors: Healthcare, Information Technology, and
Retail/Consumer/Business Products/Services.
The findings shown in Table 4 are strongly consistent with predictions. First, model #1
indicates that all but one of the estimated coefficients have the predicted signs. Firms’ projected
revenues are higher the greater are their current revenues, the more rapid is the expected growth
in patents granted to them and their total labor force, the more business development, finance,
marketing, sales, technical, and ‘other’ staff they employ, when they have a formal sales
commission plan in place, and when the forecast is made for 2005 versus for 2004.12 Projected
revenues are lower when firms are in the development or clinical trials stages of life. Only the
estimated coefficient on the number of patents mid-way through the most recently completed
fiscal year is inconsistent with predictions, being insignificantly different from zero rather than
positive. The log-linear nature of the regression means that coefficient estimates on log-
transformed variables are elasticities. Thus a one percent increase in firms’ current period
revenue is associated with an 0.59% increase in their forecasts of one-year-ahead revenue.13 Per
equation (3), the coefficient estimate of 0.59 on lnREV also implies that tangible capital
depreciates very rapidly in young venture-backed companies, with $100 of tangible capital in
year t depreciating to only $15 in year t+1.
Second, the results from model #1 support the prediction that not all employees are equal
with regard to their role in generating revenue. Thus, although the estimated coefficients on the
number of employees in business development, finance, marketing, sales, technical and ‘other’
job functions are all reliably positive, the estimated coefficient on the number of employees in
administration is insignificantly different from zero. An F-test on the hypothesis that all
employee coefficients are zero (that is, g1,= g2,= g3,= g4,= g5 = g6 = g7) is soundly rejected.
Moreover, while the data indicates that employees in non-categorized ‘other’ job functions are
expected to make positive contributions to future revenues, they do so at a rate that is on average
half that of business development, finance, marketing, sales and technical employees (g7 = 0.05
versus the average of {g2, g3, g4, g5, g6} = 0.11).
12 Estimated coefficients are not the same as primitive coefficients. Per equations (10) and (11), regression coefficients are primitive coefficients multiplied by one minus the coefficient on lnREV. 13 The departure from an exact interpretation of estimated coefficients on log-transformed variables as elasticities caused by the fact that lnREV = ln(1 + REV) rather than ln(REV) is immaterial for most firms.
15
16
Third, both financial and non-financial data explain variation in firms’ forecasts of their
one-year-ahead revenues. In model #1, the t-statistic on the sole financial variable, lnREV, is
highly significant, and the (unreported) F-statistic on the null that all the coefficients on the 13
non-financial variables are zero is 18.5 with a p-value < 0.0001. However, financial information
uniquely explains more of the cross-sectional variation in lnFREV than does non-financial
information. Of model #1’s adjusted R2 of 81%, 13% is unique to financial information, 6% is
unique to non-financial information, and 62% is common to both.
Fourth, to most accurately identify the contributions made by different types of human
capital skills, it is important to control for tangible capital. This is illustrated by comparing the
estimated coefficient on the number of administrative employees across models #1 and #3. In
model #3, which does not control for the effects of physical capital through current period
revenues, the estimated coefficient on the number of administrative employees is 0.20 (t-statistic
= 3.4) whereas in model #1 which does control for the effects of physical capital the estimated
coefficient on the number of administrative employees is 0.00 (t-statistic = 0.1).
Fifth, a comparison of the estimated coefficients on the growth in patents granted to firms
between models #1 and #3 makes it clear that most accurately identifying the contribution to
forecasted revenues made by intangible capital requires a proper control for firms’ tangible
capital. In model #3, which does not control for the effects of physical capital through current
period revenues, the estimated coefficient on the growth in patents is 0.06 (t-statistic = 0.6)
whereas in model #1 which does control for the effects of physical capital the estimated
coefficient on the growth in patents is 0.30 (t-statistic = 3.6).
Sixth, the log-linear specification of model #1 means that determinants of the log growth
rate in forecasted revenues, ln[FREV/REV], obtain by simply subtracting one from the coefficient
on lnREV. All coefficients on other independent variables remain unchanged.
Finally, a battery of untabulated sensitivity tests indicates that the results in Table 4 are
robust to a variety of potential threats to inferential reliability. For example:
• Including the log age of the firm does not materially change the size or statistical significance of any of model #1’s coefficients. Moreover, the estimated coefficient on log firm age is insignificantly different from zero.
• Excluding observations where the revenue category for both the most recently completed fiscal year and the forecasted year is the highest possible category (= 7) does not materially change the size or statistical significance of model #1’s coefficients, with the
17
exception of t-statistic on the coefficient on the number of marketing employees which is now only significantly positive at the 10% level under a one-tailed test. The reason for conducting this robustness check is that the upper bounded nature of VentureOne’s revenue categories means that a firm with very large revenues that it expects to see grow even higher will necessarily be coded as having zero forecasted revenue growth.
• Estimating model #1 using revenue categories rather than dollar revenues has no material effects on the signs or statistical significance of model #1’s estimated coefficients.
• Restricting the sample to only one observation per firm, thereby reducing the number of observations from 1,138 to 970, does not materially affect any results.
• Equation (6) fits the data well. That is, a regression of the log of total employees on the logs of employees in the seven job functions yields an adjusted R2 of 0.88. Conforming to intuition, all estimated slope coefficients lay between zero and one (the smallest is 0.09 and the largest is 0.35) and all the associated t-statistics exceeded 4.3.
Table 5 reports the results of estimating model #1 in Table 4 separately for each of
VentureOne’s three key industry-defined sectors: Healthcare, Information Technology, and
Retail/Consumer/Business Products/Services.14 Different industries are likely to have different
production functions insofar as the parameters mapping tangible assets, intangible assets and
different types of labor into revenue are concerned. Table 5 provides evidence confirming this
proposition. For example, consistent with the conjecture that a more rapidly growing workforce
generates higher sales, in all three industries I find that one-year-ahead revenue forecasts are
significantly positively related to the one-year forecasted growth rate in the total number of
employees. However, the explanatory power of specialized job functions is higher in the
Healthcare and Information Technology industries than in Retail/Consumer/Business
Products/Services sector. This supports the view that not all employees are created equal with
regard to their ability to generate revenue and that the value of specialized labor is higher the
more intangible-intensive is a firm, the explanatory power of specialized job functions is higher
in the Healthcare and Information Technology industries than in Retail/Consumer/Business
Products/Services sector. The reverse is true for non-specialized or ‘Other’-categorized
personnel, where forecasted revenues are reliably increasing in the number of ‘Other’-
categorized employees for Retail/Consumer/Business Products/Services but are unrelated to
forecasted revenues for Healthcare and Information Technology firms. Finally, in agreement
14 The industry groups are those reported by VentureOne. Retail/Services consists of Retail and Consumer/Business Products/Services. Of the sample observations, only 21 did not belong to either Healthcare, Information Technology or Retail/Services.
18
with the proposition that intangible assets play a larger role in generating revenue the more
intangible-intensive is the firm, revenues and revenue growth are increasing in the rate of patent
growth for Healthcare and Information Technology firms but not for Retail/Consumer/Business
Products/Services firms.
6. Limitations of the study
There are several limitations to this study. First, per equations (8) and (9), my regression
models assume that firms’ forecasts of one-year-ahead revenues and total headcount are
unbiased. This may be a poor assumption because venture capitalists invariably apply large
haircuts to the revenue projections that entrepreneurs provide them as part of business plans that
request financing. They do so because they believe that entrepreneurs are intrinsically optimistic
people who also inflate their true forecasts for strategic reasons. Against this argument,
however, are two points. One, firms in the study did not provide their forecasts to venture funds,
only a disinterested third party VentureOne. Moreover, they did so on a confidential basis,
meaning that current or prospective suppliers of venture capital would have no access to their
forecasts. As such, it would seem likely that the strategic component of firms’ forecasts would
be absent, leaving only entrepreneurs’ intrinsic optimism. Two, the forecasts that firms provided
were only one-year-ahead in nature—indeed, strictly speaking their forecasts were only nine
months ahead, given that the surveys were conducted in Spring, some three months into a typical
firm’s calendar fiscal year. This probably works against the degree of forecast bias being large.15
The second limitation of note is that it is not unreasonable to conjecture that while firms’
forecasts of their one-year-ahead revenues are unlikely to affect their projections of their patent
log growth rates, ln[NPATt+1/NPATt], the same cannot be said for their projections of the log
growth rate in their total employees, ln[FLt+1/Lt]. This is because firms are more able to afford
to hire more employees next year the larger is their revenue next year (particularly given the
15 One measure of the forecast bias can be obtained by regressing actual revenues and total employees in year t+1 with forecasted revenues and employees for the subset of 168 firms that responded to both the 2004 and 2005 VentureOne surveys. Thus, estimating the regression lnREVt+1 = α + βFREVt+1 + et+1 yields α = 0.70 (t-stat = 2.3) and β = 0.85 (t-stats on nulls of zero and +1 are 22.3 and –3.3, respectively). Estimating the regression lnLt+1 = α + βFLt+1 + et+1 yields α = 0.58 (t-stat = 4.8) and β = 0.80 (t-stats relative to nulls of zero and +1 are 25.3 and 4.7, respectively). However, the forecast biases suggested by these results do not take into account any distortions arising from firms’ non-random decision to participate in VentureOne’s study two years in a row.
difficulty of quickly obtaining cash in the form of external equity or debt capital). As a result,
the variable ln[FLt+1/Lt] in Tables 4 and 5 may well be endogenous, not exogenous, leading to
inconsistent estimates of the OLS parameters on the independent variables. The difficulty in
addressing this concern springs from the absence of plausible instruments in the VentureOne
database for FLt+1.16 A modicum of comfort concerning the need to rely on OLS where a key
independent variable is likely endogenous can perhaps be drawn from a recent paper by Larcker
and Rusticus (2004) that critiques the use of instrumental variables in accounting research. The
authors conclude that “Our analytical results and numerical simulations indicate that the IV
approach typically used in accounting research is in many cases unlikely to produce estimates
with desirable econometric properties. It can easily be the case that IV estimates are more biased
than simple OLS estimates that make no explicit correction for endogeneity” (p.2).
Other limitations in this study include the fact that the study is restricted to two annual
cross-sections of survey-based information that is provided voluntarily by firms. Such data may
be of lower quality than data produced by publicly traded firms whose financial and non-
financial disclosures are mandated and audited. It is also the case that data on the forecasted
number of employees is end-of-year not middle-of-year as required by the revenue production
function laid out in equation (1). The study is limited to using patents as the sole proxy for
firms’ intangible assets and investment opportunities, at a minimum leading to noisy parameter
estimates and at worse resulting in parameter estimates that suffer from material omitted
correlated variables bias. Finally, a lack of data prevented me from controlling for variables that
prior work with publicly traded companies has found relevant to explaining variation in sales
revenue, such as order backlog (Liu, Livnat and Ryan, 1996), customer satisfaction (Ittner and
Larcker, 1998) and measures of product quality (Nagar and Rajan, 2001). The same lack of data
also meant that I am unable to model or control for organization capital which Lev and
Radhakrishnan (2004) find to be an important contributor to corporate performance and growth.
For all these reasons, the inferences I have made in this study should be viewed with
appropriate caution.
16 For example, a good case can be made for FL being a function of many of the independent variables contained in model #1 in Table 4. But that necessarily eliminates them from being valid instruments.
19
7. Conclusions
Revenues and revenue growth are vital to the survival and success of all businesses,
particularly young, very risky companies that rely on staged injections of outside equity capital.
In this paper I have sought to shed empirical light on the economic determinants of firms’ sales
revenues by analyzing the one-year-ahead revenue forecasts made by a large and recent cross-
section of U.S. venture-backed pre-IPO firms.
Using a multi-input Cobb-Douglas-type model of revenue production and a large,
proprietary, survey-based dataset from VentureOne, I predicted and found that a combination of
financial and non-financial information best explains variation in firms’ revenue forecasts.
Revenues are projected to be larger the higher are current revenues, the more rapid is the
expected growth in firms’ patents and total labor force, the more business development, finance,
marketing, sales, technical, and ‘other’ staff they employ, and when a formal sales commission
plan is in place. Projected revenues are lower when firms are in the development or clinical
trials stages of life.
Based on these results, I conclude that although it cannot be ruled out that the
confidential revenue forecasts made by venture-backed pre-IPO companies are biased and/or
otherwise distorted, it seems unlikely that they are purely the ‘creative fiction’ that is often
alleged. Rather, they appear to reflect a variety of economically sensible financial and non-
financial fundamentals. As such, and because of the increasing importance of young companies
to the U.S. economy, further research into the financial and non-financial business drivers of
entrepreneurial firms would seem worthwhile.
20
REFERENCES
Armstrong, C., Davila, A., and G. Foster, 2005. Venture-backed private equity valuation and financial statement information. Working paper, Stanford University.
Barringer, B. R., Jones, F. F., and D. O. Neubaum, 2005. A quantitative content analysis of the characteristics of rapid-growth firms and their founders. Forthcoming in Journal of Business Venturing.
Davila, A., 2004. An exploratory study on the emergence of management control systems: Formalizing human resources in small growing firms. Forthcoming in Accounting, Organizations and Society.
Davila, A., Foster, G., and M. Gupta, 2003. Venture capital financing and the growth of startup companies. Journal of Business Venturing 18, 689-708.
Davila, A., and G. Foster, 2005. Management accounting systems adoption decisions: Evidence and performance implications from startup companies. Working paper, Stanford University.
Delmar, F., Davidsson, P., and W. B. Gartner, 2003. Arriving at the high-growth firm. Journal of Business Venturing 18, 189-216.
Ghosh, A., Gu, Z., and P.C. Jain, 2005. Sustained earnings and revenue growth, earnings quality, and earnings response coefficients. Review of Accounting Studies 10, 33-57.
Gompers, P.A., and J. Lerner, 1998. Risk and reward in private equity investments: The challenge of performance assessment. Journal of Private Equity, 1 (Winter), 5-12.
Gompers, P.A., and J. Lerner, 2000. The venture capital cycle, Cambridge, MA: MIT Press.
Hand, J. R. M., 2004. Determinants of the round-to-round returns to pre-IPO venture investments in U.S. biotechnology companies. Working paper, UNC Chapel Hill.
Hand, J. R. M., 2005a. The value relevance of financial statements in venture capital markets, The Accounting Review 80, 613-648.
Hand, J. R. M., 2005b. Give everyone a prize? Employee stock options in venture-backed firms. Working paper, UNC Chapel Hill.
Hellman, T.F., 2000, Venture capitalists: The coaches of Silicon Valley, in William F. Miller, Marguerite G. Hancock and Henry S. Rowen, eds.; The Silicon Valley edge: A habitat for innovation and entrepreneurship (Stanford University Press, Stanford, CA).
Hellman, T.F., and M. Puri, 2000. The interaction between product market and financing strategy: The role of venture capital. Review of Financial Studies 13, 959-984.
Hellman, T.F., and M. Puri, 2002. Venture capital and the professionalization of start-up firms: Empirical evidence. Journal of Finance 57, 169-198.
Holliday, K.K., 2000. Forget ‘cost-cutting,’ think low-cost revenue growth. ABA Banking Journal 92, 31-35.
21
Hsu, D.H., 2004a. Venture capitalists and cooperative start-up commercialization strategy. Working paper, University of Pennsylvania.
Hsu, D.H., 2004b. Experienced entrepreneurial founders and venture capital funding. Working paper, University of Pennsylvania.
Ittner, C., and D. Larcker, 1998. Are nonfinancial measures leading indicators of financial performance? An analysis of customer satisfaction. Journal of Accounting Research 36, 1-36.
Kaplan, S. N., Sensoy, B. A., and P. Stromberg, 2005. What are firms? Evolution from birth to public companies. Working paper, University of Chicago.
Kortum, S. and J. Lerner, 2000. Assessing the contribution of venture capital to innovation. Rand Journal of Economics 31, 674-692.
Larcker, D.F., and T.O. Rusticus, 2004. On the use of instrumental variables in accounting research. Working paper, University of Pennsylvania.
Lerner, J., 1994. The importance of patent scope: An empirical analysis. Rand Journal of Economics 25: 319–333.
Lev, B., 2001. Intangibles: Management, measurement, and reporting. Brookings Institution.
Lev, B., and S. Radhakrishnan, 2004. The valuation of organizational capital. Forthcoming in Measuring Capital in the New Economy, eds. C. Corrado, J. Haltiwanger and D. Sichel, The University of Chicago Press.
Liu, C., Livnat, J., and S. G. Ryan, 1996. Forward-looking financial information: The order backlog as a predictor of future sales. Journal of Financial Statement Analysis (Fall): 89-99.
Lundholm, R., and R. Sloan, 2004. Equity valuation and analysis with e-val. New York: McGraw-Hill/Irwin.
Nagar, V., and M. V. Rajan, 2001. The revenue implications of financial and operational measures of product quality. The Accounting Review 76 (4): 495-513.
Penman, S., 2004. Financial Statement Analysis and Security Valuation, 2nd ed. New York: McGraw-Hill/Irwin.
Rajgopal, S., Venkatachalam, M., and S. Kotha, 2003. The value relevance of network advantages: The case of e-commerce firms. Journal of Accounting Research 41, 135-162.
Seppä, T.J., 2003. Essays on the valuation and syndication of venture capital investments. Ph.D. dissertation, Helsinki University of Technology.
Trueman, B., Wong, F., and X.-J. Zhang, 2001. Back to basics: Forecasting the revenues of Internet firms. Review of Accounting Studies, 6 (2-3): 305-329.
Weeden, R., Rosen, C., Carberry, E., and S. Rodick, 2001. Current practices in stock option plan design. Oakland, CA: The National Center for Employee Ownership.
22
23
Table 1
Description of the sample selection criteria applied to VentureOne’s Spring 2004 and Spring 2005 compensation surveys that yielded a set of pre-IPO venture-backed firms
with the variables needed for analyzing their forecasted one-year-ahead revenues Firms in VentureOne’s financing database that were ≈ 5,000 invited by VentureOne to participate in its Spring 2004 or Spring 2005 compensation survey. Firms that responded to the 2004 survey 796 Firms that responded to the 2005 survey 500 1,396 Less those that:
Had been acquired or merged 20 Were in IPO registration 10 Were out of business 8 Were publicly traded 4 42 = Pre-IPO venture-backed firms that responded to VentureOne’s Spring 2004 or 2005 surveys 1,254 Less firms with:
Missing data on actual or forecasted revenue 57 Missing data on # employees in the sales function 41 Missing data on firm life stage or sales commission plan 9 Missing data on # patents or # employees 9 116 = Pre-IPO venture-backed firms with sufficient data for analysis of their forecasted revenues 1,138 Of which:
Firms with only 2004 data 507 Firms with only 2005 data 295 Firms with both 2004 and 2005 data 168 1,1381 1 Note that 1,138 = 507 + 295 + (2 x 168).
Table 2
Descriptive statistics on key characteristics of the sample
The sample is the set of firms in VentureOne’s Spring 2004 + 2005 compensation survey database that were venture-backed, had not yet filed for an IPO, merged or been acquired, and satisfied the data requirements described in table 1.
Panel A: Firm characteristics pertaining to explaining one-year-ahead forecasted revenues
Mean Std. dev. Min. Max. Financial MRFY actual revenue (category = 0 to 7) 2.5 2.5 0 7 One-year-ahead forecasted revenue (category = 0 to 7) 3.7 2.4 0 7 Forecasted change in revenue (# categories) 1.1 1.3 –4 6 MRFY actual revenue ($ mil) $7.5 $15.2 $0.3 $60 One-year-ahead forecasted revenue ($ mil) $13.0 $19.9 $0.3 $60 Forecasted change in revenue ($ mil) $5.5 $12.0 –$7.5 $59 Forecasted log change in revenue 0.83 0.91 –2.8 4.4
Non-financial Intangible assets # patents issued at June 30 of prior year 1.6 5.5 0 94 # patents issued at June 30 of forecasted year 2.1 6.6 0 95
Human capital and employee labor Headcount at end of prior year 60 145 1 2,838 Headcount at end of forecasted year 86 116 5 360 Forecasted change in headcount 26 135 –2,478 332 Forecasted log change in headcount 0.32 0.47 –2.1 2.6 # employees at end of prior year # administrative employees 4.5 12 0 284 # business development employees 1.1 2.1 0 35 # finance employees 2.5 4.0 0 50 # marketing employees 2.3 4.0 0 65 # sales employees 8.7 31 0 650 # technical employees 20 27 0 348 # non-categorized or ‘other’ employees 21 126 0 2,726
Incentives Indicator = 1 if firm has a sales commission plan 0.64 0.48 0 1
Investment intensity Indicator = 1 if firm is in Development life-stage 0.24 0.42 0 1 Indicator = 1 if firm is in Clinical Trials life-stage 0.06 0.23 0 1 Revenue categories are: 0 ($0-$0.5 mil.), 1 ($0.5-$1.0 mil.), 2 ($1.0-$2.0 mil.), 3 ($2.0-$3.0 mil.), 4 ($3.0-$5.0 mil.), 5 ($5.0-$10 mil.), 6 ($10-$20 mil.), and 7 (> $20 mil.). Headcount categories are: 0 (0-10 employees), 1 (10-20), 2 (20-30), 3 (30-40), 4 (40-50), 5 (50-60), 6 (60-100), and 7 (> 100).
24
25
Table 2 (continued)
Panel B: General characteristics of sample firms Variable Mean Std. dev. Min. Max. Firm age (years) 6.3 3.9 0.4 38 Indicator = 1 if firm is headquartered in CA or MA 0.50 0.50 0 1 Fraction of shares held by venture investors (%) 62% 26% 0% 100% Workforce turnover in prior year (%) 13% 14% 0% 100% # rounds of equity financing completed 3.2 1.2 1 6 Total funding raised from VCs ($ mil.) $21 $23 $0.2 $256 Funding raised from VCs in most recent round ($ mil.) $11 $11 $0.1 $134 # years since most recent round of venture funding 1.8 1.4 0.001 14 Fraction of employees who receive stock options (%) 88% 25% 0% 100% Indicator = 1 if CEO is a founder 0.49 0.50 0 1 Post-money firm value at most recent round ($ mil.) $41 $64 $1.4 $727 The number of observations in panel B varies between 577 and 1,138.
[1] ln(1 + FREV) 0.87 –0.10 –0.08 0.18 0.50 0.33 0.62 0.58 0.71 0.35 0.45 0.56 –0.46 –0.26
26
Table 3
Correlations among one-year-ahead forecasted revenues, lnFREV, and proposed economic determinants (n = 1,138 U.S. venture-backed pre-IPO firms in VentureOne’s Spring 2004 + 2005 compensation survey)
The sample is the set of firms in VentureOne’s Spring 2004 + 2005 compensation survey database that were venture-backed, had not yet filed for an IPO, merged or been acquired, and satisfied the data requirements described in table 1. Pearson (Spearman) correlations are shown above (below) the diagonal. Correlations of absolute magnitude greater than 0.06 are reliably non-zero at the 5% significance level. The signs predicted between LnFREV and proposed economic determinants are in parentheses. (+) (+) (+) (+) (0) (+) (+) (+) (+) (+) (+/–) (+) (–) (–) [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]
[2] ln(1 + REV) 0.88 –0.15 –0.03 0.10 0.53 0.31 0.65 0.53 0.67 0.31 0.49 0.48 –0.44 –0.19 [3] lnPATGROWTH –0.12 –0.16 0.18 –0.03 –0.02 –0.03 –0.05 –0.03 –0.14 0.10 –0.03 –0.14 0.07 0.13 [4] ln(1 + NPAT) –0.08 –0.05 0.50 –0.12 0.04 –0.07 0.03 –0.07 –0.11 0.00 0.04 –0.13 –0.04 0.23 [5] lnFEMPGROWTH 0.16 0.06 –0.10 –0.15 0.03 0.08 0.06 0.16 0.09 0.12 –0.03 0.03 –0.01 –0.06 [6] ln(1 + EMPADMIN) 0.50 0.50 0.01 0.07 0.05 0.30 0.58 0.40 0.46 0.41 0.42 0.22 –0.19 –0.03 [7] ln(1 + EMPBUSDEV) 0.31 0.27 –0.05 –0.08 0.07 0.25 0.37 0.24 0.23 0.26 0.13 0.10 –0.13 –0.09 [8] ln(1 + EMPFIN) 0.64 0.63 –0.01 0.03 0.06 0.53 0.32 0.49 0.54 0.37 0.49 0.29 –0.30 –0.04 [9] ln(1 + EMPMKTG) 0.58 0.51 –0.07 –0.08 0.12 0.38 0.21 0.49 0.64 0.37 0.32 0.42 –0.31 –0.15 [10] ln(1 + EMPSALES) 0.71 0.66 –0.17 –0.12 0.08 0.43 0.18 0.52 0.64 0.32 0.36 0.67 –0.43 –0.22 [11] ln(1 + EMPTECH) 0.37 0.30 0.09 0.01 0.11 0.42 0.25 0.40 0.40 0.34 –0.03 0.21 –0.12 –0.08 [12] ln(1 + EMPOTH) 0.42 0.43 0.00 0.04 –0.02 0.32 0.06 0.42 0.29 0.32 –0.01 0.21 –0.25 0.02 [13] DUMSPLAN 0.54 0.49 –0.17 –0.15 0.03 0.23 0.09 0.30 0.44 0.72 0.21 0.21 –0.43 –0.25 [14] DUMDEVT –0.46 –0.46 0.04 0.03 0.03 –0.19 –0.11 –0.31 –0.33 –0.45 –0.12 –0.25 –0.43 –0.14 [15] DUMCLIN –0.24 –0.20 0.17 0.21 –0.06 –0.03 –0.09 –0.03 –0.16 –0.24 –0.08 0.03 –0.25 –0.14 FREV is one-year-ahead forecasted revenue and REV is the most recent fiscal year’s actual revenue (in $000s). PATGROWTH is the ratio of one plus the number of patents granted at June 30 of the forecasted year to the number granted at June 30 of the prior year. NPAT is one plus the number of patents granted at June 30 of the prior year. FEMPGROWTH is the ratio of total employee headcount forecasted at the end of the forecasted year to the actual headcount in place at the end of the prior year. The EMP variables are the number of employees in administration, business development, finance, marketing, sales, technical, and other, respectively, at the end of the year prior to the forecasted year. DUMSPLAN is an indicator set to one if the firm has a formal sales commission incentive plan. DUMDEVT and DUMCLIN are indicators set to one if the firm is in the development or clinical trials stage of its life, respectively.
27
TABLE 4
Pooled regressions of one-year-ahead forecasted revenues, lnFREV, on proposed financial and non-financial determinants (n = 1,138 U.S. venture-backed pre-IPO firms, Spring 2004 + 2005)
1,3211
1,1,41,
)1ln(
ln )1ln(1
1lnln2005ln
+=
+++
++++++
⎥⎦
⎤⎢⎣
⎡+++⎥
⎦
⎤⎢⎣
⎡+
++++=
∑ tiiiimit
M
mm
it
tiit
it
tiititi
DUMCLINhDUMDEVThDUMSPLANhEMPg
LFL
fNPATdNPAT
NPATcREVbDUMhaFREV
µ
The sample is the set of firms in VentureOne’s Spring 2004 + 2005 compensation survey database that were venture-backed, had not yet filed for an IPO, merged or been acquired, and satisfied the data requirements described in table 1. The dependent variable is the log of the firm’s forecast of its one-year-ahead revenues (in $000s). T-statistics computed using robust standard errors are in parentheses. An intercept is estimated but not reported. *, ** and *** denote coefficient estimates that are reliably significant at the 5%, 2.5% and 1% levels under a one-tailed test of the predicted sign, unless there is no sign prediction in which case the significance level pertains to a two-tailed test. Independent variables Pred. sign Model #1 Model #2 Model #3
Dummy = 1 if year t+1 is 2005 + 0.13 (2.7)*** 0.09 (1.7)* 0.15 (2.5)***
Financial ln (1 + actual $ revenues in year t) + 0.59 (27.8)*** 0.86 (58.4)***
Non-financial Intangible assets ln (patent growth rate, t to t+1) + 0.30 (3.6)*** 0.06 (0.6) ln (1 + # patents at t) + –0.04 (–1.4) –0.01 (–0.4)
Human capital and employee labor ln (Forecasted employee growth, t to t+1) + 0.32 (6.4)*** 0.42 (6.5)*** ln (1 + # administrative employees at t) 0 0.00 (0.1) 0.20 (3.4)*** ln (1 + # business dev. employees at t) + 0.15 (3.5)*** 0.24 (4.3)*** ln (1 + # finance employees at t) + 0.14 (2.7)*** 0.52 (8.2)*** ln (1 + # marketing employees at t) + 0.11 (2.6)*** 0.17 (3.1)*** ln (1 + # sales employees at t) + 0.11 (3.3)*** 0.32 (7.2)*** ln (1 + # technical employees at t) + 0.06 (2.4)*** 0.09 (2.5)*** ln (1 + # other employees at t) +/– 0.05 (2.9)*** 0.16 (6.9)***
Incentives Sales commission plan indicator at t + 0.38 (5.6)*** 0.49 (5.6)***
Investment intensity Development life-stage indicator at t – –0.36 (–5.5)*** –0.80 (–9.6)*** Clinical trials life-stage indicator at t – –0.70 (–6.3)*** –1.27 (–9.0)***
Adj. R2 0.81 0.75 0.68 # obs. 1,138 1,138 1,138
28
TABLE 5
By-industry regressions of one-year-ahead forecasted revenues, lnFREV, on proposed financial and non-financial determinants (n = 1,104 U.S. venture-backed pre-IPO firms, Spring 2004 + 2005)
1,3211
1,1,41,
)1ln(
ln )1ln(1
1lnln2005ln
+=
+++
++++++
⎥⎦
⎤⎢⎣
⎡+++⎥
⎦
⎤⎢⎣
⎡+
++++=
∑ tiiiimit
M
mm
it
tiit
it
tiititi
DUMCLINhDUMDEVThDUMSPLANhEMPg
LFL
fNPATdNPAT
NPATcREVbDUMhaFREV
µ
The sample is the set of firms in VentureOne’s Spring 2004 + 2005 compensation survey database that were venture-backed, had not yet filed for an IPO, merged or been acquired, satisfied the data requirements described in table 1, and were not classified as in industry = “Other”. The dependent variable is the log of the firm’s forecast of its one-year-ahead revenues (in $000s). T-statistics computed using robust standard errors are in parentheses. An intercept is estimated but not reported. *, ** and *** denote coefficient estimates that are reliably significant at the 5%, 2.5% and 1% levels under a one-tailed test of the predicted sign, unless there is no sign prediction in which case the significance level pertains to a two-tailed test. Predicted Industry (as defined by VentureOne) Independent variables coef. sign Healthcare Info. Tech. Retail/Services
Dummy = 1 if year t+1 is 2005 + 0.11 (1.2) 0.14 (2.1)** 0.05 (0.6)
Financial ln (1 + actual $ revenues in year t) + 0.67 (13.1)*** 0.53 (18.6)*** 0.69 (17.0)***
Non-financial Intangible assets ln (patent growth rate, t to t+1) + 0.30 (1.8)* 0.28 (2.9)*** 0.21 (0.7) ln (1 + # patents at t) + –0.04 (–0.9) –0.04 (–0.9) –0.08 (–0.7)
Human capital and employee labor ln (Forecasted employee growth, t to t+1) + 0.30 (3.0)*** 0.38 (5.2)*** 0.28 (3.4)*** ln (1 + # administrative employees at t) 0 –0.04 (–0.4) 0.05 (0.7) 0.10 (1.4) ln (1 + # business dev. employees at t) + 0.36 (3.7)*** 0.10 (1.7)* 0.07 (1.0) ln (1 + # finance employees at t) + 0.10 (1.0) 0.18 (2.5)*** 0.09 (0.9) ln (1 + # marketing employees at t) + 0.31 (2.7)*** 0.02 (0.3) 0.00 (0.1) ln (1 + # sales employees at t) + 0.12 (1.4) 0.14 (2.6)*** –0.02 (–0.5) ln (1 + # technical employees at t) + 0.02 (0.3) 0.15 (3.6)*** –0.08 (–1.5) ln (1 + # other employees at t) +/– 0.01 (0.2) 0.07 (2.4)** 0.06 (2.1)**
Incentives Sales commission plan indicator at t + 0.54 (3.5)*** 0.35 (3.7)*** –0.05 (–0.4)
Investment intensity Development life-stage indicator at t – –0.14 (–1.0) –0.35 (–3.8)*** –0.07 (–0.4) Clinical trials life-stage indicator at t – –0.31 (–2.1)**
Adj. R2 0.82 0.79 0.83 # obs. 324 577 203