WHO SEEKS ASSISTANCE AND WHAT DO THEY GET OUT OF IT?
AN IMPACT EVALUATION OF THE MSU PRODUCT CENTER
By
Adam Lovgren

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Agricultural, Food and Resource Economics – Doctor of Philosophy
2014

i

ABSTRACT
WHO SEEKS ASSISTANCE AND WHAT DO THEY GET OUT OF IT?
AN IMPACT EVALUATION OF THE MSU PRODUCT CENTER
By
Adam Lovgren
This paper seeks to evaluate the impact that the Michigan State University Product
Center Food-Ag-Bio (or Product Center for short) has had on the success and survival of the
entrepreneurs who have received assistance from it. To determine this impact, a survey was sent
out to over 2200 Michigan food and agricultural entrepreneurs in the fall of 2012 . Over 600
entrepreneurs responded to the survey, both clients and non-clients of the Product Center, and the
resulting data was used to conduct this analysis. Considerable effort was taken to overcome the
inherent selection bias in entrepreneurial assistance program (EAP) evaluations. This bias is due
to the fact that those who seek assistance are more likely to have a higher propensity to seek
information and lower overall entrepreneurial ability, and both factors are also likely to influence
venture performance. These biases are best thought of as an unobserved variable bias when
included a regression analysis. To overcome these biases, therefore, we identified observable
factors that predict the decision to seek assistance and included these in the impact evaluation
regressions in order to have unbiased estimators. In short, we found that the Product Center had
the most incremental impact on small to medium sized ventures, older entrepreneurs with less
industry-related business experience, and for those who participated in the development of the
business idea prior with the Product Center prior to the launch of their new venture. This impact
was seen most notably in increased survival rates, levels of employment and increased perceived
legitimacy with external resources holders.

ii

This work is dedicated to the brave entrepreneurs out there willing to forgo steady employment
in pursuit of achieving their dream in a very difficult and uncertain environment. Though not all
will succeed, all are worthy of admiration and respect for trying to make it on their own.

iii

ACKNOWLEDGEMENTS

First and foremost I would like to thank my wife Kimiko Lovgren, and daughters Kiana and
Zoey for all their support and understanding of my journey through the PhD process and in the
long-hours spent writing this dissertation. Next I would like to thank my major professor Dr. H.
Christopher Peterson for believing in me from my first semester in graduate school at Michigan
State University and providing me with a large amount of support and sage advice throughout
the years. I would also like to thank my dissertation supervisor Dr. R. Brent Ross, for all his
work and patience on the development of this document and the many subsequent revisions. I
also appreciate the comments and suggestions provided by my other committee members Dr.
Roger Calantone and Dr. Dave Weatherspoon. I would like to thank my father, Fred Lovgren for
his assistance with grammar checking the errors that I can no longer see, and my mother Kathy
Lovgren for her support on the home-front. Finally, I would also like to thank all the AFRE
graduate students who came along with me on this journey and spent many hours in Cook Hall
and elsewhere learning the tools of the trade.

iv

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................................... viii
LIST OF FIGURES ....................................................................................................................... xi
I.

II.

INTRODUCTION .................................................................................................................. 1
A.
Unit of Analysis ............................................................................................................... 4
B.
Research Questions .......................................................................................................... 6
C.
Dissertation Roadmap ...................................................................................................... 8
LITERATURE REVIEW ..................................................................................................... 10
Mechanisms by which EAPs create value...................................................................... 10
i.
Distinguishing by clusters of entrepreneurs ............................................................... 20
ii.
Distinguishing by the timing of assistance ................................................................. 20
iii.
Distinguishing by size of the venture ......................................................................... 21
iv.
Identifying the counterfactual to deal with selection bias .......................................... 22
B.
Different differences are affected differently ................................................................ 29
i.
Distinguishing by the timing of assistance ................................................................. 30
C.
Summary of Research Questions and Hypotheses ........................................................ 34
A.

III. CONCEPTUAL FRAMEWORK ......................................................................................... 36
A.
Dealing with Selection Bias .......................................................................................... 37
B.
Determining the Average Treatment Effect on the Treated (ATET) ............................ 38
IV. DATA ................................................................................................................................... 40
A.
Determination of the Treatment Variable – pcassist ...................................................... 41
B.
Performance Variables {Y} of Interest .......................................................................... 43
i.
Measure of Survival: Still in operation – “stillalive” ................................................ 43
ii.
Measures of Performance: Current gross annual sales and total employment .......... 43
iii.
Perceived Legitimacy Measures ................................................................................ 44
iv.
Ability to Obtain External Financing – “externfinance” ........................................... 44
v.
Total Sales Sold through Formal Distribution Channels .......................................... 45
vi.
Subjective Assessments of the Assistance Provided ................................................. 46
C.
Control Variables to be used .......................................................................................... 46
i.
Food Processor Industry Indicator – foodproc ........................................................... 47
ii.
Pushed/pulled into entrepreneurship – pushed ........................................................... 47
iii. Contact hours – contact, contact2 ............................................................................... 48
iv.
Venture Age – ventureage .......................................................................................... 48
v.
Other Controls – inherit, familyentrep, male, white ................................................... 48
D.
Instrumental Variables (IV) to be tested ...................................................................... 49
i.
Distance away from the Product Center – disttopc ................................................... 50
ii.
Social tie to Michigan State University – msuconnect .............................................. 51
v

E.
i.
ii.
iii.
F.
i.
ii.
iii.
iv.
G.
i.
ii.
iii.
H.
i.
ii.
V.

Proxies for Omitted Selection Bias Variables ............................................................... 51
Propensity to Seek Information – propinfoseek ........................................................ 51
Propensity to Grow the Business – propgrowbiz ...................................................... 52
Entrepreneurial ability - Education and Experience –baormore and indusexp, ........ 52
Entrepreneurial Orientation Proxies .............................................................................. 53
Innovativeness - propinnovate ................................................................................... 54
Proactiveness – numbernewlines; changesinlines ..................................................... 54
Risk-Taking – proprisktake; propaggressiveposture ................................................. 55
Competitive Aggressiveness ...................................................................................... 55
Summary Statistics ......................................................................................................... 56
Performance Variables .............................................................................................. 56
Instrumental Variables, Omitted Variable Proxies and EO measures ....................... 58
Other Control Variables ............................................................................................. 60
Cluster Analysis on Entrepreneur Types ........................................................................ 62
Cluster characteristics ................................................................................................. 64
Comparison of performance statistics between clusters ............................................. 65

WHO SEEKS ASSISTANCE? ............................................................................................. 69
A.
Fit within the decision-tree framework .......................................................................... 69
B.
Methods .......................................................................................................................... 69
i.
Probit regression analysis ........................................................................................... 70
C.
Factors that influence the decision to seek assistance .................................................... 72
D.
Instrumental Variables ................................................................................................... 75
E.
Who seeks assistance prior to the launch of their venture?............................................ 78
F.
Implications of these Results.......................................................................................... 80

VI. WHO FINDS ASSISTANCE USEFUL? ............................................................................. 83
A.
Methods .......................................................................................................................... 83
B.
Influence of timing of assistance on perceived usefulness ............................................ 84
C.
Perceived usefulness amongst those who launch new ventures ..................................... 87
D.
Influence of firm size on perceived usefulness .............................................................. 90
E.
Why certain entrepreneurs find assistance useful when others do not. .......................... 92
i.
Reported Satisfaction with Assistance ....................................................................... 93
ii.
Reported Usefulness of Assistance............................................................................. 95
F. Implications of these Results ............................................................................................. 99
VII. IMPACT OF PC ASSISTANCE ON PERFORMANCE ................................................... 102
A.
Methods ........................................................................................................................ 102
i.
Ordinary Least Squares (OLS) ................................................................................. 102
ii.
Probit Regressions ................................................................................................... 103
iii.
Tobit Regressions ..................................................................................................... 104
B.
Impact of EAP assistance on new ventures’ success and survival ............................... 105
C.
Impact of assistance on the decision to launch ............................................................ 106
i.
Testing the “weeding out” component of assistance ................................................ 108
ii.
Testing the “planting in” component of assistance .................................................. 109
vi

D.
E.
F.
G.
H.
I.
J.

EAP Impact on Legitimacy with Resource Holders ..................................................... 113
EAP Impact on Legitimacy with Trading Partners ....................................................... 116
EAP Impact on Gross Annual Sales.............................................................................. 119
EAP Impact on Employment ........................................................................................ 121
Impact of the timing of assistance on performance ...................................................... 123
EAP impact on small ventures (grossing under $200,000 per year) ............................. 126
Implications of these findings ....................................................................................... 132

VIII. SUMMARY OF FINDINGS........................................................................................... 136
IX.

EAP MANAGERIAL IMPLICATIONS ......................................................................... 145

X.

FUTURE RESEARCH IMPLICATIONS ........................................................................ 151

REFERENCES ........................................................................................................................... 158

vii

LIST OF TABLES

Table 1: Summary of the services provided by the MSU Product Center ......................................6
Table 2: Comparison of EAP impact studies (adapted f. Kosters and Obschonka, 2010)............19
Table 3a: Summary of Research Question (R1) and Associated Hypothesis
and Sub-Hypotheses......................................................................................................34
Table 3b: Summary of Research Question (R2) and Associated Hypothesis...............................35
Table 3c: Summary of Research Question (R3) and Associated Hypothesis
and Sub-Hypotheses.......................................................................................................35
Table 4: Summary of the Treatment Variable .............................................................................42
Table 5: Performance Variable summary statistics tabulated by treatment status ........................58
Table 6: Summary of Instrumental, Proxy and EO variables ......................................................61
Table 7: Summary Statistics for Control and Instrumental Variables
Tabulated by Treatment .................................................................................................62
Table 8: Comparison of other relevant control variables between PC and
Non-PC clients ...............................................................................................................63
Table 9: Results of Cluster analysis on control variables ............................................................64
Table 10: Comparison of performance statistics between clusters ..............................................65
Table 11: Comparison of performance statistics between PC clients
and non-clients by cluster. ............................................................................................67
Table 12: Comparison of performance statistics between Product Center clients
and non-clients for firms grossing under $200,000 .......................................................69
Table 13: Unrestricted probit regression of treatment on controls and instruments ....................73
Table 14: Probit regression of treatment on controls and instruments restricted to
food processors ...........................................................................................................76
viii

Table 15: Probit regression of treatment on controls and instruments restricted to those
providing sales data ......................................................................................................78
Table 16: Probit regression of the decision to seek assistance prior to launch .............................79
Table 17: Probit regression of probability to seek assistance and find it useful on timing of
assistance and controls restricted to respondents who launched a product and had
some form of contact with the Product Center .............................................................87
Table 18: Probit regression of probability to seek assistance and find it useful restricted to
respondents who launched and had some form of contact with the Product Center ..89
Table 19: Probit regression of finding assistance useful on timing of assistance and controls
restricted to respondents who launched and had contact with the Product Center ......90
Table 20: Probit regression of usefulness of assistance on controls and firm size .....................92
Table 21: Regression of launch status on controls restricted to treated .......................................99
Table 22: Probit regression of launch status on treatment (at decision node a1) and controls ..110
Table 23: Probit regression of the probability of the new venture being still in operation on
treatment and control variables ..................................................................................112
Table 24: Left- and Right-hand censored Tobit of external financing level on controls and
treatment. ....................................................................................................................115
Table 25: Probit regression of receiving External Financing on Treatment and Controls ........116
Table 26: Tobit on volume of sales sold through a retail outlet on treatment and controls for
respondents grossing under $200,000 per annum ......................................................119
Table 27: Right-hand censored Tobit regression of gross annual sales on treatment and
controls .......................................................................................................................121
Table 28: OLS regression total employment on treatment and controls ....................................123
Table 29: Right-hand censored Tobit on Gross Annual Sales restricted to only those respondents
who have received assistance. ....................................................................................125
Table 30: Probit regression of the ability to obtain external financing on controls and prelaunch
assistance restricted to Product Center clients who have launched a product. .........127
ix

Table 31: Probit regression of firm size on control variables ...................................................129
Table 32: OLS of gross annual sales on treatment and controls, restricted to respondents
grossing less than $200,000 per annum .....................................................................130
Table 33: OLS regression of Total Employment on treatment and controls restricted to those
grossing under $200,000 per annum in sales. ................................................................131
Table 34: Probit of receiving external financing on treatment and controls restricted to those
grossing under $200,000 per annum ..........................................................................132
Table 35: Summary of research question 2, hypotheses and findings .......................................139
Table 36: Summary of research question 3, relevant hypotheses and findings .........................140
Table 37: Summary of research question 1, relevant hypotheses and findings .........................143
Table 38: Implications for EAP managers .................................................................................151
Table 39: Implications for future research. ................................................................................158

x

LIST OF FIGURES

Figure 1: Decision-tree framework for an entrepreneur with a business idea .............................36
Figure 2: Histogram of Product Center Clients’ Reported Satisfaction Levels. ...........................95
Figure 3: Histogram of Product Center Clients’ Reported Perceived Usefulness
of Assistance Levels. .....................................................................................................96
Figure 4: Comparison of launch status between Product Center clients and non-clients ..........108
Figure 5: Comparison between treated and non-treated entrepreneurs on the reasons why
those that did not launch a product gave for not launching. ........................................109
Figure 6: Survival Rates between the Treated and Non-Treated Groups ..................................111

xi

I.

INTRODUCTION
Entrepreneurship is often discussed as a primary driver of economic growth in a society

(Knight, 1921; Schumpeter, 1934). One study suggests that nearly 70 percent of US economic
growth can be attributed to entrepreneurial activity (Reynolds et al, 2000). Research has also
consistently shown that over 50% of new ventures fail within the first five years (Cooper, Woo
and Dunkelburg, 1988; Shane, 2008). Therefore, given that new venture creation is both valuable
to society but difficult to achieve, it is not surprising that a significant amount of public dollars
have been allocated to create programs to assist entrepreneurs. These organizations are known
broadly as Entrepreneurial Assistance Programs (EAPs).
The primary goal of most EAPs is to aid prospective small business owners in new
venture creation by providing pre-venture assistance (Chrisman, Hoy, Robinson, 1987). While
prior research seems to indicate that EAPs such as the Small Business Development Center
(SBDC) can positively impact the formation of new ventures (e.g Clark et al., 1984; Stevenson
and Sahlman, 1988; Solomon and Weaver, 1983; Robinson, 1982), these programs are costly.
The initial Small Business Development Act of 1980 authorized an annual funding level of $20
million dollars. This number has since grown to fund the roughly 1,000 full time service centers
that operate with an overall budget of roughly $200 million dollars (ASBDC, n.d.). Evaluation
of the effectiveness of EAPs and EAP practices is therefore of non-trivial importance (Yusuf,
2010).
The evaluation of EAPs first began n the early 1980s to the mid-1990s there with studies
that measured the impact of EAPs in terms of comparing EAP clients to non-EAP clients on
performance measures such as increase in sales, employments and profits (Robinson, 1982;
Chrisman, Nelson, Hoy and Robinson 1983; and Chrisman, Hoy and Robinson 1987). In general,

1

these studies found that EAP-clients outperformed their non-EAP counterparts. However, these
studies often compared the performance of a small number of EAP clients (usually under 100)
from high performing EAPs to state averages of entrepreneurship (e.g. Chrisman, Hoy and
Robinson 1987).
In addition to performance measurement comparisons there have been other studies
aimed at the subjective assessment of client satisfaction as a measure of the benefit of an EAP
(Ibrahim and Goodwin, 1986; Solomon, 1983; Nahavandi 1988). These studies have found that
EAP clients have, on average, rated EAPs positively. However, one study found no correlation
between client satisfaction levels and the previously indicated performance indicators
(McMullan, Chrisman and Vesper, 2001). Instead, they concluded satisfaction came more from
the personal experience the client had with the counselor.
Furthermore, other researchers (.g. Wren & Storey, 2002) have pointed out that these
early studies do not account for the selection bias that is inherent in the EAP process. This
selection bias occurs for two separate reasons. On the one hand, some have suggested that the
fact that an entrepreneur has sought assistance indicates the entrepreneur is likely having some
problems with the launch of their new venture (e.g. Kosters and Obschonka, 2010). These
problems could be due to a lack of business experience or entrepreneurial ability that in turn will
affect the overall performance of the new venture. Therefore, any comparison of assisted
entrepreneurs to non-assisted ones would have an inherent negative bias in its estimation if this
sample-selection is not accounted for. On the other hand, the propensity to seek as much
information as possible before making important strategic business decisions, as opposed to
relying on cognitive biases and heuristics, has also been shown to increase the overall probability
of success for a new venture (e.g. Baron, 2004). Given that entrepreneurs often come to EAPs to

2

seek information, thus potentially signaling a higher propensity to seek information than those
who do not come to an EAP, some researchers have suggested that this will cause an upward bias
on estimators that compare EAP clients to a control population that has not sought assistance
(e.g. Rotgers et al, 2012).
In essence if one were to conduct a simple regression such as:
ܻ௜ = 	 ܺ௜ ߚ + ܼ௜ ߛ + ߳௜
Where Y denoted a particular performance variable, X was a vector of explanatory control
variables and Z was a dummy variable indicating participation in an EAP with a corresponding β
and γ vectors of coefficients to be estimated, then the biases mentioned above would result if
some unobserved variables, such as entrepreneurial ability or the propensity to seek information,
influences the participation variable, as well as the elements of performance not directly
controlled for (i.e. the error term). Then the estimator of interest, γ, would be correlated with the
error term (which would include the omitted variable) and hence lead to inconsistent estimation
(Wooldridge, 2002).
This concern is particularly problematic because, given that the two causes of selection
bias act in opposite directions, not only is the magnitude of the bias uncertain but so too is the
sign.
Another complicating factor involved in the evaluation of EAP effectiveness is the
unclear mechanism by which an EAP can create value for its clients. Recently, Yusuf (2010)
assessed EAPs based on client participation, satisfaction and entrepreneurs’ subjective
assessments of overall effectiveness. In regard to the last measure, Yusuf found that EAP
programs were effective at meeting the nascent entrepreneur’s support need only 25.8% of the
time. However, despite this lack of effectiveness, Yusuf still found that 96% of the surveyed

3

clients found the assistance at least somewhat valuable, with 50% finding it extremely valuable
(Yusuf, 2010). This high degree of satisfaction, yet unclear correlation to how the assistance
actually improves the business also suggests deeper analytical research is required in order to
uncover the mechanisms of EAP impact on the success and survival of its clients’ new ventures.
To do this, an appropriate unit of analysis must be selected in order to test the hypotheses relative
to the research questions that will be further developed below.
A. Unit of Analysis
The EAP chosen as a unit of analysis to uncover the impact that an EAP can have on the
success and survival of its clients’ new ventures is the Michigan State University Product Center
for Agriculture and Natural Resources (or The Product Center for short). The Product Center was
created by a memorandum of understanding among the MSU College of Agriculture and Natural
Resources (CANR), Michigan State University Extension (MSUE), and the Michigan
Agricultural Experiment Station (MAES) on March 1, 2003 and is still in operation as of today
(2012) more than eight years later. The original mission was, “To be a catalyst for the creation of
profitable futures for businesses and industries engaged in Michigan’s agriculture, food and
natural resources systems.” This was then expanded into a three part framework that emphasized
the Product Center’s role as a business and technical assistance program, a market research
institution and an entrepreneurial education provider (from MSU Product Center Strategic Plan,
2007). However, over time, it became clear that the entrepreneurial education component was
not highly valued by the entrepreneurs themselves, and this component was dropped.
The Product Center’s team consists of a core group of self-directed staff members
involved in all or most of the organization’s processes, a small executive group comprised of the
Product Center director and the two associate directors, who take actions and make commitments
4

on behalf of the organization, and two operating subgroups: a research subgroup–composed of
university faculty and students who engage in interdisciplinary research aimed at identifying and
supporting actual and potential clients’ needs; and a venture development subgroup–who work
with the actual and potential business clients, as well as the internal and external partners, to
provide the analysis and services the clients require. In addition, the Product Center has a vast
network of affiliates, including previous clients, partners and stakeholders who support the
organization in its operations.
The Product Center’s central offices are housed on the campus of Michigan State
University, but its innovation counselor network is dispersed throughout the entire state of
Michigan, operating through MSU’s extension network. This structure allows clients to have
their first contact with an innovation counselor in their local extension offices, with more
advanced services offered on campus.
At the extension level, selected extension agents are trained to be “Innovation
Counselors” who serve as a first contact for individuals interested in receiving services from the
Product Center. The on-campus specialized service unit assists entrepreneurs by either directly
providing services to clients or connecting them to on-campus departments. The services
provided are: packaging, nutritional labeling, assistance in obtaining financing, feasibility
studies, food-safety testing, assistance in supply-chain entry, product testing, strategic advice and
legal assistance.
Since it began operations in 2004 the Product Center’s staff has had over 15,805 one-on-one
client sessions, helped 1,434 clients with developing their venture concept, has gotten 881 clients

5

to the start-up stage and helped to produce 164 launches. In addition the Product Center has also
provided over 917 clients with different specialized service assistance (see table 1 below).
Table 1: Summary of the services provided by the MSU Product Center
SERVICES PROVIDED

SINCE 2004
21,205 sessions

One-on-one client counseling sessions
Assistance with business concept development

1,797 clients

Venture start-ups

1,039 clients

Specialized services

1,031 clients

Venture launches

229 ventures

B.

Research Questions

While prior research seems to indicate that EAPs such as the Product Center or the Small
Business Development Center (SBDC) can positively impact the formation of new ventures (e.g
Clark et al., 1984; Stevenson and Sahlman, 1988; Solomon and Weaver, 1983; Robinson, 1982),
these programs are costly. The initial Small Business Development Act of 1980 authorized an
annual funding level of $20 million dollars. This number has since grown to fund the roughly
1,000 full time service centers that operate with an overall budget of roughly $200 million
dollars (ASBDC, n.d.). Evaluation of the effectiveness of EAPs and EAP practices is therefore
of non-trivial importance (Yusuf, 2010). Therefore, the primary research question this paper
seeks to answer in relation to the chosen unit of analysis is:
How much value has the Product Center, as an example EAP, created for its clients?

6

(R1)

Before R1 can be answered, however, we must first further develop how EAPs create
value for their clients in order to determine the proper measures and methods to quantify the
value created. We must also determine what a suitable “control” group of entrepreneurs should
be to base our evaluation upon. Unfortunately, we cannot observe what would have happened to
the entrepreneur’s new venture if they had not received assistance, nor can we randomly assign
who receives treatment and who does not across a group of similar entrepreneurs. Therefore,
comparing entrepreneurs who have received assistance from an EAP to those who have not is not
a sufficient method of evaluating EAP impact. Instead, we will need to further develop a method
to ensure the estimated differences are valid and consistent. To do this, we must identify the
factors behind the supposed selection bias associated with comparing assisted entrepreneurs to
non-assisted ones. Therefore, we must also answer the following research question:
What types of entrepreneurs seek assistance and why?

(R2)

Finally, as entrepreneurs are an extremely heterogeneous group, relying on the average
impact of the entire sample set alone will ignore this heterogeneity and therefore will not tell the
whole story of EAP impact. We also wish to know: Do entrepreneurs with more experience find
assistance less useful?; Are there other observable factors that impact perceived usefulness?;
Does time of contact with the EAP within the venture creation process affect the perception of
usefulness of the EAP’s services? Therefore, the third research question this paper will address
is:
What types of entrepreneurs find assistance useful and why?

(R3)

It is hoped that by answering the stated questions above, future researchers will have a
better understanding of how to evaluate EAP impact, what to expect the impact will be and how
7

that impact varies across different types of entrepreneurs. For the sake of clarity, it should be
noted that treatment and assistance are used interchangeably in this analysis in an effort to stay
true to both the EAP stream of literature (which uses assistance) and the impact evaluation
methodological stream of literature (which uses treatment).
C.

Dissertation Roadmap
In order to answer these questions many steps will be taken. First, in chapter 2, a

theoretical rationale on the mechanisms by which an EAP can create value for its clients will
built through a literature review of both the entrepreneurial process literature as well as on the
work done by past EAP evaluation researchers. Included within this literature review will be a
deeper discussion of the problem of selection bias in EAP evaluation studies as well as
techniques that can be employed to overcome this problem. The results of this literature review
will be a set of testable hypotheses that can answer the three research questions stated above.
Next, the Michigan State University (MSU) Product Center, as an example EAP, will be
introduced as the unit of analysis in chapter 3. A discussion of the history of the Product Center,
its organization, guiding philosophy and clientele will be provided to justify its use an
appropriate unit of analysis.
Chapter 4 will provide a conceptual model of the entrepreneurial assistance process that
ties the hypotheses to each research questions within the context of the model. This model
provides a framework of the decision-making process an entrepreneur makes with deciding
(consciously or not) whether to seek assistance in the launch of a new venture. This framework
will clarify what stage of the process each hypothesis is testing. This will prove to be important
in interpreting the quantitative results both in terms of selection bias and the value that assistance
plays at different points in the entrepreneurial process.
8

Chapter 5 will lay out the performance and control variables that were collected in order
to test the proposed hypotheses and the rationale for the inclusion of each variable. It will then
discuss the sample population used to collect data on the variables from, the method used to
collect this data, the response rate and summary statistics for each variable. In addition, a cluster
analysis will be performed to identify different types of entrepreneurs within the sample.
Chapter 5 will then seek to answer the research question on who seeks assistance and
why. This chapter will discuss the data and methods used to test the proposed hypotheses for this
question, the results associated with each test, and implications that come from these results.
Similarly Chapters 6 & 7 will do this for the other two research questions (“What types of
entrepreneurs find assistance useful and why?” and “How much value has an example EAP
created?” respectively).
Finally, Chapter 8 will summarize the results of the entire dissertation and draw
conclusions from the results of the preceding chapters. From here implications will be drawn for
both future EAP researchers and EAP managers.

9

II.

LITERATURE REVIEW
A. Mechanisms by which EAPs create value
To begin with, we will first look into finding a theoretical basis to answer the primary

research question:
How much value has an example EAP, the Product Center, created for its clients? (R1)
Given the level of public funding and the widespread nature of EAPs nationwide, we
propose the following hypothesis to be tested as an answer to this question:
Entrepreneurs who receive assistance from an EAP will be more likely to have higher
performance, survival rates and legitimacy than had they not sought assistance. (H1)
One theory proposed by Chrisman and McMullan (2000; 2005) on the mechanism by
which EAPs can help improve clients’ new venture survival and success is that assistance from
outside advisors facilitates the development of knowledge, as a special type of resource available
to the firm. They argue that the knowledge possessed by the entrepreneur or entrepreneurial team
creates the foundation for many if not all of the new venture’s competitive advantages (Alvarez
and Busenitz, 2001; Chrisman et al., 1998, c.f. Chrisman and McMullan, 2005). Given that
individual entrepreneurs have imperfect knowledge of market conditions (Hayek, 1945) and may
not know how to write business plan, obtain financing, optimally locate their business or deal
with trading partners, EAPs can help improve new venture success by providing the tacit and
explicit knowledge needed to fill those gaps (Chrisman and McMullan, 2005).
In addition, individuals facing uncertainty have certain heuristics and biases that also
influence their decision-making processes (Busenitz & Barney, 1997). More importantly, these
heuristics and biases vary across individuals and will have a significant impact on who decides to
become an entrepreneur and the probability of success of the venture (Venkatraman, 1997). This

10

variation is useful in determining not only why some individuals become entrepreneurs when
others don’t, but also in evaluating why some ventures fail and others do not. Specifically,
though certain biases and heuristics will increase the likelihood of deciding to exploit a perceived
entrepreneurial opportunity, it can also negatively impact the probability of success of that
entrepreneurial opportunity if the entrepreneur does not full appreciating the costs or risks
involved and take appropriate measures to deal with them (Baron, 2004). Therefore, the more
that the entrepreneur can build and engage “safety nets” that provide checks on whether a
cognitive bias is leading them down a potentially negative path, the more successful they will be
at exploiting an entrepreneurial opportunity (Simon et al, 2000). By leveraging an EAP
counselor’s tacit strategic knowledge into the entrepreneur’s information set when developing
the entrepreneur’s new venture strategy, entrepreneurs can thereby make better decisions as to
whether or not to exploit a particular opportunity and how to do so. This will lead to a “weeding
out” of bad ideas, which is beneficial in light of the opportunity cost of wasting resources on
untenable ideas, as well as the “planting in” of good ideas that otherwise may not have been
exploited. In this way, EAPS can provide a check on whatever cognitive biases an entrepreneur
may have in order to help them make better decisions. To determine if this is true, the following
two sub-hypotheses will be tested in relationship to research question (R1):
Entrepreneurs who receive assistance from the example EAP prior to the launch of a
new venture will be less likely to launch that new venture than those who do not (H1-a –
weeding out hypothesis).

11

Entrepreneurs who receive assistance from the example EAP prior to the launch of a
new venture and decide to launch that new venture will be more likely to stay in business
than those who launched a new venture but did not receive assistance (H1-b – planting in
hypothesis).
Another mechanism by which EAPs can create value for their clients is the signal or
“badge”(Bell et al., 2002) that completing a business plan with EAP assistance can provide. This
was found in the author’s master’s thesis (Lovgren, 2012) as well as a recent article by Rotger,
Gortz and Storey (2012).
In order to exploit a new venture, there are many types of resources an entrepreneur must
acquire. These include physical equipment, land and human capital, but perhaps the most
important capital required before launch of the new venture is financial capital. Since the
recombination of resources inherent in the exploitation of an entrepreneurial activity must occur
before the sale of the output, this exploitation must be financed (Knight, 1921; Shane, 2003). The
majority of the time this takes the form of self-financing (Alrdich, 1999). External financing is
also available in the form of equity investment which would include funding from angel
investors and venture capitalists, debt-financing, asset-based financing and governmental or nonprofit grants.
For entrepreneurs without observable track records or products, resource providers must
make financing decisions on very little information and their own intuition (Bhide, 2000). Due to
this limited information and evidence, investors face high risk in financing entrepreneurial
opportunities (Low and Srivatsan, 1994), which can result in high interest rates on loans to cover
this risk or the inability to get a loan at all.

12

There are two basic characteristics of new ventures that make acquiring resources and
developing ties to trading partners difficult for entrepreneurs – uncertainty and information
asymmetry (Shane, 2003). Entrepreneurial opportunities are, by definition, uncertain because
the profitability associated with the new resource combination is unknown before the launch of
the venture (Arrow, 1974). In addition, in an environment that does not fully understand or
acknowledge their existence it can be quite difficult for new organizations to establish ties with
the trading partners they need to organize their venture properly (Hannan & Caroll, 1992;
Stinchcombe, 1965). These characteristics create what Stinchcombe (1965) has termed “the
liability of newness” that is often a key contributor to the low survival rate of new organizations.
The uncertainty of the new venture’s profit stream also creates difficulties in determining
the net present value of the venture, which can lead to differences of opinion about the venture’s
value between the entrepreneur and the lender (Wu, 1989). This can create bargaining problems,
and result in the resource provider offering less than what the entrepreneur believes is the value
of the opportunity (Shane, 2003). Furthermore, resource providers might wish to have collateral
(Blanchflower and Oswald, 1998) against default if the entrepreneur fails and cannot pay back
the loan (Casson, 1982). This will pose significant problems for those entrepreneurs without
suitable collateral to offer.
Information asymmetry, defined in this case where the entrepreneur holds more
information about the value of the opportunity and her own dedication and work ethic in regards
to exploiting it, creates four major obstacles to capital acquisition (Shane, 2003).
The first comes from the natural tendency of entrepreneurs to want to keep key
information about the value of and the method for exploiting an opportunity. Making this
information known could cause others to attempt to exploit it as well, thus bidding up the

13

requisite resource prices associated with that opportunity (Casson, 1982) or may cause the
resource providers to exploit it themselves. This secrecy leads to resource providers to make
decisions with incomplete information, and my cause them to refuse to finance the venture or not
provide enough financing as is needed (Shane, 2003).
Second, this information asymmetry also allows for opportunism on the part of the
entrepreneur through obtaining more resources than their venture requires (Shane and Cable,
2002) or by getting better concessions on the investments or loans than a resource provider with
a full information set would allow (Shane, 2003).
Third, there is also some moral hazard risk in the entrepreneur engaging in unnecessarily
risky behavior once he has received the loan, or not putting forth an appropriate effort level as is
required to fully exploit the opportunity. This, in turn, will make resource providers take
safeguards against opportunism and increase the transaction costs involved in obtaining
financing (Williamson, 1975; 1985; 1986). These safeguards often involve resource providers
offering fixed rate financing at high interests rates so the entrepreneur bears the risk. At high
rates, however, entrepreneurs will be more likely to favor risky projects with high potential
returns (Barzel, 1987) as low to moderately risky projects with low to moderate returns will be
less likely to be profitable, given the high interest rates.
Lastly, given there is a range of ability in both discovering and exploiting entrepreneurial
opportunities, those who are less skilled at doing so will seek financing in addition to those that
are more skilled. If the resource provider cannot distinguish between the varying skill levels,
then this can create a “market for lemons” (Akerlof, 1970) and lead to adverse selection in the
lending market (Amit et al, 1990). This will have the net effect of lenders either refusing to
participate in the lending market, or participating only by offering interest rates appropriate for

14

less skilled individuals. Since they are less skilled, and therefore more likely to default, these
rates will be significantly higher than would be appropriate for those who are more likely to
succeed in their new venture, potentially to the point of being cost-prohibitive.
To overcome this “liability of newness” Zimmerman and Zeitz (2002) have suggested
that organizations can build legitimacy, as a special type of resource, to allow them to access
other resources that are needed to survive and grow. Legitimacy, is defined by Suchman (1995)
as "a generalized perception or assumption that the actions of an entity are desirable, proper, or
appropriate within some socially constructed system of norms, values, beliefs, and definitions,"
and can act as the antidote to the information asymmetry and uncertainty ailments mentioned
above. However, new ventures often have few resources available to them, so Zimmerman and
Zeitz postulate that the degree to which new ventures can engage in legitimacy building
strategies that cost little or no money, “a certain threshold can be obtained that will allow them to
access the capital and other resources it needs. … below which the new venture will probably
perish,”(Zimmerman and Zeitz, 2002).
One method by which an EAP can help entrepreneurs to build legitimacy is through the
assistance they provide in business planning and the development of a formal business plan.
Business plans, or documents that present the entrepreneur’s conjectures in written and visual
form, are important tools in overcoming many of the uncertainty and information asymmetry
issues in the resource acquisition process (Shane, 2003). Business plans allow entrepreneurs to
tell their stories in an institutional way that helps legitimize the new venture (Zimmerman and
Zeitz, 2002), and thereby provides an appropriate platform to reduce the level of information
asymmetry to investors.

15

Business plans can also signal the quality of the entrepreneur and opportunity to
investors, which can help overcome adverse selection and moral hazard issues (Shane, 2003).
Shane (ibid.) has shown that business plans can provide legitimacy when external validation of
the opportunity is difficult (Aldrich, 1999), by rationalizing the opportunity in a way that is easy
for others to accept (Lounsbury and Glynn, 2001). This can generate confidence (Fisher, 1985)
and convince resource providers that the opportunity has potential, making it more likely that the
entrepreneur can get the resources necessary to actually make it a reality (Gartner et al., 1992).
However, because business plans are created by entrepreneurs, they may tend to be overly
optimistic and cause resource holders to wish for additional forms of validation (De Meza and
Southey, 1996; Busenitz and Barney, 1997). Creating a business plan with the support and
endorsement of EAP, therefore, can further strengthen the plan’s validity. This, in turn, will
increase the probability the entrepreneur is able to obtain the appropriate level of capital she
needs to adequately exploit the entrepreneurial opportunity.
In addition, an EAP can help build client legitimacy through developing socials ties with
key resource holders and trading partners and then leverage the EAP’s relationship on behalf of
the client. In this case, the indirect social tie between the entrepreneur and resource holder who
each have a direct relationship to the EAP (Burt, 1987), can be help to mitigate problems of
uncertainty and information asymmetry. One way this is done is through reducing people’s
tendency to act in a self-interested manner by adding additional social disincentives (Marsden,
1981; Uzzi, 1996; Granovetter, 1985; Gulati, 1995). That is to say, to the extent that both
entrepreneur and trading partner wants to maintain the relationship with the EAP, they will try to
honor the agreements facilitated by the indirect tie. It can also provide another type of “badge”

16

for the resource holder, based on the trust the resource holder has for the EAP’s
recommendations and processes.
To test this notion that EAPs can create value for entrepreneurs through increasing their
perceived legitimacy with resource holders the following sub-hypothesis will also be tested:
Entrepreneurs who receive assistance from the example EAP will be more likely to obtain
external financing than had they not received assistance. (H1-c)
Similarly for the entrepreneur, establishing successful trading partnerships in the supply
chain is just as troublesome as obtaining the capital. In particular, if a venture is perceived as
lacking legitimacy, then it will be less likely for trading partners such as distributors, processors
and retailers to invest time and resources in the new venture. This is because cooperation from
such partners is often based on trust, reliability and reputation, which in turn are based on
familiarity and evidence (Bateson, 1988). In situations where entrepreneurs have little evidence
that their new venture will work out well, and the trading partners do not already have an
established trust relationship, there is little reason for them to engage in the relationship (Aldrich
and Fiol, 1994).
If this is true, then we will be able to find support for the following sub-hypothesis:
Clients from the example EAP will have access to a greater marketing opportunity set
than they otherwise would have without assistance (H1-d).
Logically, if the EAP in our study is effective at assisting entrepreneurs in business
planning, increasing their marketing opportunity set and obtaining external financing, this in
turn should lead to higher overall performance rates for the assisted entrepreneur than they
otherwise would have had without assistance. This begs the question, however, as to what type
of performance are we interested in measuring. In general the previous studies reviewed (see

17

table 2 below, adapted from Kosters and Obschonka, 2010) have focused on measuring the
impact from assistance through an analysis of clients’ subjective assessments of the assistance
provided (57% of studies reviewed), increase in sales or earnings (57%), employment growth
(50%) and venture creature and/or survival (36%). In order to provide convergent validity for
this study, therefore, all four measures will be evaluated in regards to the example EAP.

18

Table 2: Comparison of EAP impact studies (adapted from Kosters and Obschonka, 2010)
Performance Measures Used
Author(s)
and Year of
Study
Roger et al.
(2012)
Yusuf (2010)
Kosters &
Obschonka
(2010)
Mole et al.
(2008)
Stubner et al.
(2007)
Parker &
Belghitar
(2006)
Chrisman et
al. (2005)
Kulicke
(2004)
Wren &
Storey (2002)
Chrisman &
McMullan
(2000)

Method of
Analysis

Control
for
Selection
Bias?

Positive
EAP
impact?

Econometric

Y- PSM

Y-All

Econometric

N

N

Econometric

Y- PSM

N-Emp
YUseful

x

Econometric

N

Y - Emp
N - Sale

x

Econometric

N

Y - All

Econometric

Partially

N

Econometric

N

Y - All

Monitoring

N

Y

Venture
EAP
Emp.
Sales
Survival
Useful Growth growth /Launch
x

x

x

x
x

x
x

x

x
x

x

x

x

Chrisman
(1999)
Barney et al.
(1996)

x

Chrisman &
Lee (1989)

x

Chrisman
(1989)

x

x

x

x

Econometric Y - 2SLS

Y – ME
N –SE.

x

x

x

Mean
Comparison

N

Y - All

x

x

x

Mean
Comparison

N

YLaunch

Econometric

N

Mean
Comparison

N

Monitoring

N

x

Percentage
57%
50%
57%
36%
PSM- propesnity Score Matching; 2SLS - Two Stage Least Squares,
ME-Medium Enterprises; SE – Small Enterprises (author defined)

19

21%

Y - for
less exp.
Y - In
the short
run
Y - Str;
N- Op
&Ad
79%

Whether the example EAP increases the survival rates of its clientele will already be
addressed in H2 (the “planting in” hypothesis) and therefore will not require a separate
hypothesis for testing here. The other three measures will be tested through the evaluation of the
following hypotheses:
Clients from the example EAP will have higher gross annual sales rates than they
otherwise would have without assistance (H1-e).
Clients from the example EAP will higher employment rates than they otherwise would
have without assistance (H1-f).
i.

Distinguishing by clusters of entrepreneurs

Previous research done by both the author and others (e.g. Kosters and Obschonka, 2010;
Rotgers et al, 2012) has also suggested that there are distinct clusters of entrepreneurs who may
receive differential benefits from an EAP. In this regard, a cluster analysis will be done on the
sample obtained to determine if distinct clusters can be identified for entrepreneurs based on the
following opposing distinguishing characteristics: intense versus casual users of the EAP
(through investigating the differential impact of number contact hours for respondents within the
EAP group); and on the basis of entrepreneurial orientation, age, experience, education and goals
for the business. Performance characteristics will then be compared across the different clusters
to determine how different clusters performance varies with and without assistance.
ii.

Distinguishing by the timing of assistance

Up to this point we have presented the assistance process as one where entrepreneurs
have an idea for a new venture, seek assistance to work on that idea, work with the EAP to
develop the idea and take the necessary steps to launch the product, actually launch the product
and then have performance in the market that we can measure. While there are indeed some
20

EAPs who follow this path, there are many who do not come in until their product has already
launched and is having some issue that the entrepreneur wishes to seek assistance for. For this
latter group, it often will involve some specific technical service, such as obtaining nutritional
facts labeling, as opposed to the more in-depth business counseling as they have already
developed their business model and are running with it. Though these services are indeed
valuable, we might not expect to see as much incremental gains in performance from assistance
then we see from the entrepreneurs who are following the more in-depth route detailed at the
beginning of this paragraph. Therefore, we hypothesize the following to be true:
Clients who receive assistance prior to the launch of their new venture will receive
greater impact from assistance than those who receive it only afterwards. (H1-g)
iii.

Distinguishing by size of the venture

Another method in which we can distinguish between new ventures is by their size in
terms of gross annual sales. This will be done because those businesses that are already doing
well enough to be obtaining significant revenues are probably the least likely to need assistance.
Furthermore, until recently, the Product Center’s focus has been on aiding the start-up of small
businesses to go from producing a cottage industry good out of their kitchen to taking the next
step towards becoming a growing commercial product. It helps these businesses at each step of
the way from developing the venture concept, acting as a sounding board to help “weed out”
untenable ideas and “plant in” good ones, to developing a business plan to attract investors, to
obtaining the proper packaging and labeling, to adhering to industry regulations and norms, to
finally gaining access to marketing outlets and developing proper growth strategies. These are all
things that the large business have likely already done and hence do not find that beneficial. For

21

those small business and nascent entrepreneurs without a lot of experience, however, this
assistance can be invaluable.
For the purposes of this study, small businesses will be defined as those entrepreneurs
whose new ventures are grossing under $200,000. While this number may appear a bit arbitrary,
it does allow us to separate out the largest of the sample whose needs will be quite different from
the small start-ups. In addition, it gives us some methodological value as this data set does not
have full information on the sales of the ventures grossing more than $200,000 per year (only
that they are above this limit). Therefore, we can use more efficient estimation techniques such
as OLS (Wooldridge, 2002).
Because the large ventures data is censored with regards to sales data, there is no
variation to evaluate if one wanted to look at the impact of assistance within only this group.
Therefore, this dissertation will look to see if EAP assistance is more pronounced within the
smaller ventures than compared with the full sample that includes the larger ventures through the
testing of the following hypothesis against its null for each of the performance measures:
When restricting the sample to entrepreneurs whose new venture grosses under $200,000 per
year, the impact of the example EAP assistance will be more likely to be significantly positive
than when compared to the full sample.(H1-h)
iv.

Identifying the counterfactual to deal with selection bias

As was mentioned in the introduction, and should be evident in the formulation of the
previous hypotheses with the words, “than they otherwise would have without assistance” as
opposed to “than those who did not seek assistance,” the counterfactual of interest to test
whether we can reject the null hypotheses cannot simply be the average of the non-assisted
population. In fact, early EAP studies did exactly this and utilized a comparison of means of the
22

subjective assessments of assistance within the treated group, or a comparison of means of the
treatment group versus a control population (e.g. Chrisman, 1989; Chrisman & Lee, 1989;
Chrisman, 1999; Chrisman & McMullan, 2000). All found that assistance created a positive
impact on all four measures of performance listed above. However, it is now evident that a
simple comparison of means of assistance entrepreneurs to a general population of entrepreneurs
will not provide a sufficient counterfactual to the treatment group due to the concerns over
selection bias (e.g. Wren & Storey, 2002). This selection bias comes from the notion that the
unobserved factors that determine whether an entrepreneur seeks assistance (e.g. entrepreneurial
skill, experience, information-seeking personality dispositions, etc..,) will also influence the
overall performance of that entrepreneur’s new venture and will hence bias simple mean
comparison tests.
To see why this will cause biased and inconsistent estimations of the average treatment
effect, consider the case of an Ordinary Least Squares (OLS) regression method. To use OLS, we
must make the standard Gauss-Markov assumptions (see Wooldridge, 2002). The first
assumption is that our model is linear in parameters. If so, then we can estimate the impact of
treatment on a continuous performance variable such as sales or employment through estimating
the coefficients in equation (1):
(1)

ܻ௜ = ܺ௜ ߚ + ܼ௜ ߙ + ߝ௜

Where for a set of N individual entrepreneurs, the ith entrepreneur’s performance variable
{Y} is determined by a vector of explanatory variables {W} including the control variables {X}
and a binary treatment variable of assistance {Z}, and an error term {ε} that represents all other
unobserved influences on {Y}. With the key estimator of interest {α} represents the impact of

23

assistance on performance to be solved by minimizing the sum of the squared error terms in
respect to the β and α coefficients.
The selection bias concern would result if some unobserved variables, such as
entrepreneurial ability or the propensity to seek information, influence the participation variable,
as well as the elements of performance not directly controlled for (i.e. the error term). If this
happens, then the estimator of interest (γ) would be correlated with the error term (i.e. with the
omitted variables in the error term) and hence lead to inconsistent estimation. In particular, the
Gauss-Markov assumption of a zero-conditional mean of {ε} in all of these equations, is where
the selection bias issue causes problems. Essentially, the unobservable factors that influences the
treatment variable {Z} as well as influence the uncontrolled elements of performance {ε} cause
E[ε|Z] ≠ 0, which in turn causes the estimation of α to be biased and inconsistent with a
probability limit approaching: α + E{[Z’ε](W’W)-1} (Wooldridge, 2002). This concern is
particularly problematic because, given that the two causes of selection bias act in opposite
directions, not only is the magnitude of the bias uncertain but so too is the sign.
Some authors have attempted to overcome the selection bias problem by focusing on the
incremental benefit provided by various levels of assistance within the assisted population (e.g.
Chrisman et al., 2005), while others have turned to econometric techniques such as a HeckmanLee two-step estimation (Heckman, 1979; Wren & Storey, 2002) or Propensity Score Matching
(Heckman & Navarro-Lozano, 2004; Rotgers et al, 2012; Kosters and Obschonka, 2010). In such
cases where the selection bias has been controlled for, the evidence that EAPs create positive
impacts is much less clear, with some finding no impact (Kosters and Obschonka, 2010), some
finding mixed results (Wren and Storey, 2002) and some still finding positive impacts (Chrisman
et al., 2005; Rotgers et al., 2012).

24

It should be noted, however, {Z} is only endogenous if it is correlated with the
uncontrolled elements of performance {ε}. As discussed above, given the selection biases
inherent in the analysis of any sample population this seems a likely assumption. However, one
must first provide evidence that selection bias is actually occurring and what mechanisms are
causing selection to occur before one can take corrective measures to remove this bias. Then, if
one can model the omitted variables behind the selection bias directly in the control vector {X}
then the resulting {ε} will not be correlated with {Z} and endogeneity will not be a concern
(Wooldridge, 2002).
To model these omitted variables in the performance regressions, requires first answering
research question which states:
What types of entrepreneurs seek assistance?

(R2)

Some of the work done in previous academic journal articles suggests that there are at least
two potential selection forces. Interestingly, these forces have opposite effects in their purported
bias on entrepreneurial performance, thus further complicating the issue.
On the one hand, the fact that an entrepreneur has sought assistance indicates the
entrepreneur is likely having some problems with the launch of their new venture (e.g. Kosters
and Obschonka, 2010). These problems could be due to a lack of business experience or
entrepreneurial ability that in turn will affect the overall performance of the new venture. This
selection force will create an inherent negative bias in performance estimations.
On the other hand, the propensity to seek as much information as possible before making
important strategic business decisions, as opposed to relying on cognitive biases and heuristics,
has also been shown to increase the overall probability of success for a new venture (e.g. Baron,
2004). Given that entrepreneurs often come to EAPs to seek information, thus potentially

25

signaling a higher propensity to seek information than those who do not come to an EAP, some
researchers have suggested that this will cause an upward bias on estimators that compare EAP
clients to a control population that has not sought assistance (e.g. Rotgers et al, 2012).
Therefore, in relation to research question (2), there are two testable hypotheses can be
formed that will ultimately help us answer research question (R1):
Entrepreneurs with less entrepreneurial experience and/or ability will be more likely to seek
assistance than those with more experience and/or ability. (H2)
Entrepreneurs with a higher propensity to seek information will be more likely to seek
assistance than those with a lower propensity to seek information. (H3)
Whether or not clear observable factors can be identified to reject the nulls in relationship to
H2 and H3 will be of prime importance in creating validity for the results found on all other
hypotheses. However, even if one cannot do so for both hypotheses, if one is successful for at
least one of the two, then one can argue that even if measures cannot be taken to entirely remove
the selection bias from the analysis, one can at least identify the direction of the bias. If, for
example, we can remove the upward bias associated with information seeking behavior, but not
the downward bias associated with ability, then knowing the direction of the bias (i.e.
downward) will still allow any positive results to hold as they would therefore be conservative
estimates of the overall impact. This would allow us to show where positive impact has been
made, and give a lower bound for the magnitude of that impact (Wooldridge, 2002).
Furthermore, even if one cannot directly model the omitted variables, other measures can still
be taken to remove selection bias. Surprisingly, one of the most efficient methods of addressing
this issue, namely conducting an Instrumental Variable (IV) regression (Wooldridge, 2002), has
not been attempted in any of the reviewed EAP studies above. This approach, relies on finding

26

an instrument {P} for the endogenous treatment variable {Z} such that cov (P, ε) = 0 and cov
(Z,P) ≠ 0 in order to identify the exogenous element of {Z} such that a consistent estimate of α in
equation (1) can be determined. While finding an appropriate instrument can often be difficult, it
is unclear why this approach has never been attempted. In fact, without such a valid instrument
that can be excluded from the second-stage, the only means to identify the estimator is through
the non-linear estimation of the inverse-mills ratio (Wooldridge, 2002). Relying on the normality
of the first-stage to identify the second-stage, however, does not appear parsimonious nor
strongly credible, whereas a valid instrument that can identify the exogenous element of
treatment would be.
There also does not seem to be clear evidence that the most favored current technique for
dealing with selection bias, Propensity Score Matching, provides a consistent estimate of the
treatment effect. This technique uses a probit or logit regression to predict the likelihood to seek
assistance based on a large number of observable characteristics and then compares the means of
the performance variables of assisted entrepreneurs to those of non-assisted entrepreneurs who
have relatively close “propensity scores” (i.e. the conditional predicted probability of seeking
assistance from the estimated coefficients of the probit regression) (Heckman & NavarroLozano, 2004). However, Rosenbaum and Rubin (1983) in developing this technique pointed out
that in order to use it one must be able to assume what they call, “the strong ignorability of
treatment,” which essentially posits that the outcome variable is independent of the decision to
seek treatment. Heckman et al (2004) later showed this assumption could be weakened to
requiring that only the conditional mean of performance is independent of the decision to seek
treatment. However, in this case, the decision to seek treatment is exactly what we are positing is
causing the selection bias and therefore this assumption appears completely invalid for use in

27

EAP studies. Not to mention, if such an assumption could be made then conducting a standard
ordinary least squares (OLS) regression would also be valid and would be the best unbiased
linear estimator (Wooldridge, 2002) that could also more effectively control for the impact of
other exogenous variables on performance as well. Therefore, this technique will not be used in
this study.
Instead, this paper will attempt to directly model the omitted variables if they can be
identified through the testing of H8 and H9, using as a fall back option of bounding the results
either on the upwards or downwards side if only one of the selection forces can be observed
directly. We will also attempt to find valid instruments to conduct a Heckman double-hurdle
model. This method involves two-stages, one to predict selection and the one that measures
performance. In order to identify the first stage, an instrument is required (Wooldridge, 2002)
unless one wishes to obtain identification through the non-linear properties of the inverse-mills
ratio, which is undesirable as it relies on the normality of the selection equation. Therefore, to
determine if the 2SLS or the Heckman Two-Step procedure can be used, the following
instrumental variables will be tested for use:
Distance away from the example EAP (IV-1)
Having a social tie to the example EAP’s institutional home (IV-2)
These will be tested through determining first if the instruments are significant first stage
predictors in the probit analysis done for testing H2 and H3, then if so, will be used to conduct a
Hausman test to see if the instrument can reject the no endogeneity null in the second stage. If
the instrument cannot do satisfy both of these conditions, it will not be considered valid. If, in
that case, we are not able to reject H଴ଶ for either instrument, then an alternative approach must be
used. Since we have already rejected the use of Propensity Score Matching on the basis of its
28

invalidity under the presence of selection bias, the only other option would be including the
proxies for the unobserved variables that are causing the selection bias. The evaluation of the
validity of this method will depend on the face-value validity of the proxy as well as its ability to
reject the null hypothesis in the selection equation
B. Different differences are affected differently
Finally, after addressing the first two research questions, this dissertation will look to
answer the third stated research question which states:
What types of entrepreneurs find assistance useful and why?

(R3)

This will be done in order to provide greater insight into the value the example EAP has
created for different types of entrepreneurs. This will be done because the types of entrepreneurs
who come to the Product Center are very diverse. For example, one Product Center client
described her venture in the following way, “my business is a small cottage foods business that
will never be big due to the nature of the products (wild harvested). My goal is to educate people
about wild foods, good health, and get people out-of-doors.” For this type of entrepreneur the
goals and needs will be quite different from the one who said the following, “Because of (the
Product Center’s trade show) Making it in Michigan 2011 (we were) able to meet with Kroger
Michigan and Meijer Corporate. We are working with Meijer to provide a (product) for the 2013
season in their 199 bakeries.” For this reason, while we can gain some insight through the
evaluating the average impact the EAP is having on the performance measures used in this
analysis, this might not tell the whole story.
Furthermore, the existing skill set of the entrepreneur who said, “I never worked in this
industry prior to starting this business - previously a scientist employed in the automotive
industry,” will be quite different from the one who tells the following tale, “Farming has been
29

our family business since 1854. I am the 5th generation on our farm. My sons are the 6th
generation. Upon the retirement of my father & uncle I took over operation of our farm in 1991.
In 1994 I initiated a shift to organic production--which placed me closer to the end-user, and
opened up inquiries from consumers for direct purchase of products. My initial product was a
wheat free pancake/waffle mix. I made it for personal consumption after I discovered that I am
allergic to wheat. Friends commented on how good it tasted--asked for mix to take home--and a
literal cottage/kitchen business was born!!” It is quite likely that the value that assistance
provides to these two different types of entrepreneurs is also quite different. Therefore, we
should attempt to distinguish what the different types of entrepreneurs are and how assistance
has impacted these different types differently.
i.

Distinguishing by the timing of assistance

Given the intense focus that EAPs have on providing pre-launch assistance in order improve
the business strategy before committing resources to the idea, we also wish to investigate in this
regard is whether those who do receive pre-launch assistance will find it more useful than those
who receive only post-launch assistance. We can see if there is evidence for this through the
testing of the next hypothesis against its null:
Entrepreneurs who seek assistance prior to launch will be more likely to find that assistance
useful than those who seek assistance only after launch. (H4)
If this is shown to be true, we may wonder about whether those who seek assistance prior to
launch are categorically different from those who seek it after launch, and therefore will have
different perceptions of the perceived usefulness of assistance. In this case, it might be
reasonable to assume that the propensity to seek information is driving the decision to seek

30

assistance prior to launch, as by our definition, proactive information seekers will seek out as
many sources of information as is necessary to make an informed decision.
On the other hand, the lack of ability might be assumed to be the more important driving
factor in the decision to seek it after launch. This is because, as Baron (2004) has shown, most
entrepreneurs are often subject to cognitive biases prior to launch that cause them to overestimate
their own abilities and/or the value of the perceived entrepreneurial opportunity.
However, after the launch of the new venture and the subsequent feedback from the market,
the extent of the overestimation can be brought to light. This is likely to be greater for those with
less ability, who then might finally decide to seek assistance. In this case we would not find any
significance difference between the pre-launch assisted group and the post-launch assisted group
based on the propensity seek information, but would in regards to a proxy of entrepreneurial
ability (such as experience). If so, this might provide additional insight into how distinguish
between the entrepreneurs who seek assistance because of their higher propensity to seek
information from those that seek assistance because of their lack of ability. This can be
investigated further through testing the following sub-hypotheses:
Entrepreneurs who seek assistance prior to launch will be more likely to have a higher
propensity to seek information than those who seek assistance only after launch. (H4-a)
Entrepreneurs who seek assistance only after launch will have significantly less
entrepreneurial ability than those who seek assistance prior to launch. (H4-b)
Furthermore, given the notion that EAP assistance helps entrepreneurs through providing
knowledge and resources to help fill gaps that the entrepreneur would otherwise have in
achieving a successful venture, and that those gaps are more likely to occur with less experienced

31

ventures as well as those who have smaller firms, we also propose the following two hypotheses
to be tested:
Entrepreneurs who have launched products and have less entrepreneurial experience will be
more likely to find assistance useful. (H5)
Entrepreneurs who have launched products with firms whose gross annual sales is less than
$200,000 per year will be more likely to find assistance useful. (H4-6)
These will be tested through conducting a probit analysis on the whether an assisted
entrepreneur sough assistance either prior to launch of their new venture or not. In particular, we
will wish to see whether any of the proxies for entrepreneurial ability are significant in this
decision.
Next, though the literature has not dealt specifically with this issue, there is some concern
when estimating treatment effects that many experienced or skilled entrepreneurs do not find the
assistance all that beneficial. This would be different from the initial selection bias, as this effect
would occur only within the group that has sought assistance, but would give us an indication if
the type of entrepreneur who persists through the process might be different than the one who
drops out from assistance, but still goes on to launch a product. This might indicate that the
counseling and services provided are a substitute, rather than a complement, to the
entrepreneur’s own knowledge and skill set. In this case, the entrepreneurs who drop out from
assistance, but still go on to launch their new venture, will be likely to have higher performance
and survival rates than those who persist in the process due to the higher degree of
entrepreneurial skill and ability the ones who drop out possess. In this case, if these
entrepreneurs who drop out because of their higher ability levels were excluded from the assisted
group it would certainly cause bias. However, most analyses, including this one, will often still

32

categorize them as “treated” as they have been impacted by the EAP despite whether they
consciously found it to be useful and therefore not cause bias. Nevertheless, if it is evident that
the higher ability entrepreneurs who seek assistance are not finding it useful, then it is likely that
there are similar entrepreneurs who make the decision not to even seek assistance on the
assumption that they too will not find it useful. In this case, it would point out that the concern
over selection bias is genuine and should be accounted for.
Therefore, the following hypotheses can be developed in regards to research question (3):
Entrepreneurs who seek assistance and then decide to not launch a new venture will have been
less likely to be successful than those who have persisted with the assistance process. (H7)

33

C. Summary of Research Questions and Hypotheses
In summary, the three stated research questions of interest for this dissertation will be
addressed through the testing of these five hypotheses (summarized below in table 3).
Table 3a: Summary of Research Question (R1) and Associated Hypothesis and Sub-Hypotheses.
R1

How much value has the Product Center, as an example EAP, created for its
clients?
H1
Entrepreneurs who receive assistance from an EAP will be more likely to have higher
performance, survival rates and legitimacy than had they not sought assistance.
H1-a Entrepreneurs who receive assistance from the example EAP prior to the launch of a new
venture will be less likely to launch that new venture than those who do not (weeding out
hypothesis).
H1-b Entrepreneurs who receive assistance from the example EAP prior to the launch of a new
venture and decide to launch that new venture will be more likely to stay in business than
those who launched a new venture but did not receive assistance (planting in hypothesis).
H1-c

Entrepreneurs who receive assistance from the example EAP will be more likely to
obtain external financing than had they not received assistance.

H1-d Clients from the example EAP will have access to a greater marketing opportunity set
than they otherwise would have without assistance.
H1-e

Clients from the example EAP will have higher gross annual sales rates than they
otherwise would have without assistance.

H1-f

Clients from the example EAP will higher employment rates than they otherwise would
have without assistance.

H1-g Clients who receive assistance prior to the launch of their new venture will receive
greater impact from assistance than those who receive it only afterwards.
H1-h When restricting the sample to entrepreneurs whose new venture grosses under $200,000
per year, the impact of EAP assistance will be more likely to be significantly positive.

34

Table 3b: Summary of Research Question (R2) and Associated Hypotheses.
R2
H2

What types of entrepreneurs seek assistance and why?
Entrepreneurs with less entrepreneurial experience and/or ability will be more likely to
seek assistance than those with more experience and/or ability.

H3

Entrepreneurs with a higher propensity to seek information will be more likely to seek
assistance than those with a lower propensity to seek information.

Table 3c: Summary of Research Question (R3) and Associated Hypotheses and SubHypotheses.
R3
H4

What types of entrepreneurs find assistance useful and why?
Entrepreneurs who seek assistance prior to launch will be more likely to find that
assistance useful than those who seek assistance only after launch

H4-a

Entrepreneurs who seek assistance prior to launch will be more likely to have a higher
propensity to seek information than those who seek assistance only after launch.

H4-b

Entrepreneurs who seek assistance only after launch will have significantly less
entrepreneurial ability than those who seek assistance prior to launch.

H5

Entrepreneurs who have launched products and have less entrepreneurial experience
and probabilities of success without assistance will be more likely to find assistance
useful.

H6

Entrepreneurs from firms whose gross annual sales is less than $200,000 per year will be
more likely to find assistance useful.

H7

Entrepreneurs who seek assistance and then decide to not launch a new venture will have
been less likely to be successful than those who have persisted with the assistance
process.

The next chapter will develop a conceptual framework of how assistance can impact the
entrepreneurial process and will place the research questions and related hypotheses within the
context of this model. After this model is developed, we will present a chapter on the data used
to test the hypotheses, followed by individual chapters for each research question and
corresponding hypotheses.
35

III.

CONCEPTUAL FRAMEWORK
In order to better conceptualize how we can measure the impact an EAP such as the

Product Center is having on an entrepreneur, consider the decision-tree framework for each
entrepreneur (i) who has a business idea in regards seeking assistance and launching their
product is presented below in figure 1.
Figure 1: Decision-tree framework for an entrepreneur with a business idea
a

b

c

d

Seeks
Assistancec1

Performanced1

Does not
seek
Assistancec2

Performanced2

Seeks
Assistancec3

Performanced3

Does not
seek
Assistancec4

Performanced4

Launchb1
Seeks
Assistancea1
No
Launchb2

Entrepreneur
i
with a
business idea

Launchb3
Does not
seek
Assistancea2

No
Launchb4

Essentially, if an individual entrepreneur is considering turning a business idea into a
new venture, they make a decision (whether consciously or not) at point (a) whether to seek
assistance from an EAP. Next, prior to the launch of their new venture they must develop the
business idea and make a decision at point (b) whether they feel the expected utility of launching
the new venture is greater than the expected utility they will receive if they do not launch. Once
the decision to launch the venture is made, then the characteristics of the new venture, the

36

entrepreneur and the market will determine its performance. At any time post-launch at point (c),
however, the entrepreneur can also seek assistance prior to when data is collected at point (d).
Therefore, there are four different paths possible (d1-d4) an entrepreneur who has provided
performance data can take in this framework at the time of our observation. In addition, there are
two other paths respondents could have taken (b2) & (b4) that would lead them to still enter the
database, but without performance data.
For the purposes of this study, and most other studies, any of the performance data observed
at points (d1-d3) would be considered within the treatment group, where only the entrepreneurs
at point (d4) would be considered to be in the non-treated group. However, as was mentioned
before, the selection bias concern manifests itself if the underlying factors that drive the
decisions made at points (a) & (c) also influence the performance observed at point (d), then
comparing the performance of entrepreneurs at points (d1-3) to point (d4) would be biased
without controlling for this difference. Furthermore, the incremental impact of assistance given
at point (a), that is prior to the launch of the new venture, is likely to be greater in magnitude
than when given after the launch, at point (c). This is because, prior to launch, the EAP can
directly influence the overall strategy and even decision to launch more than after launch when
resources have already been committed.
A. Dealing with Selection Bias
To begin this study, therefore, we will first need to determine what the differences are
between those who seek assistance at all versus those who do not through the testing of H8-H11
using those respondents who fall under category a1 and c3 as the treated group and those who
fall under category b4 and d4 as the control. This will allow for the most unrestricted testing of
H8-H11 to give us the broadest perspective of the differences between these groups. Next we can
37

see if there are any distinct differences in observable characteristics between the timing of the
assistance through modifying the hypotheses of H8-H9 to add the words “prior to the launch of
their products” after the words “to seek assistance” and then use those respondents who fall
under category a1 as the treated group and those under category c3 as the control. The
instrumental variable hypotheses will suffice as written so long as one bears in mind the
instrument will always refer to the treated group as stated. Finally we can test whether there are
differences in the differences between the treated and control groups at the (a1, a2) decision
node and the (c3, c4) decision node. These series of hypothesis testing will give us a better
understanding of the differences between those entrepreneurs who seek assistance versus those
who do not, as well as the differences between those who seek assistance prior to the launch of
their product and those who seek it only afterwards, so that we can have a more complete picture
of how well suited our control group is for testing the hypotheses for R1 and R3. This will also
help to guide us on the appropriate methods we can use to overcome or at least mitigate any
selection bias concerns.
B. Determining the Average Treatment Effect on the Treated (ATET)
Ideally, as with any impact study, the goal of this paper is to identify how much value the
example EAP has created for its clients above what would have otherwise occurred. Of course,
the actual counterfactual can never be 100% determined as it only occurs within the realm of
possible worlds that might have resulted from individuals making different decisions at nodes (a)
and (c) in our conceptual model. Therefore, we must estimate what could have happened through
measuring the performance of the assisted entrepreneur versus similar non-assisted
entrepreneurs. Determining how similar the control group of respondents is to the treated group
has been discussed in the previous section, and will be used to create the validity of the
38

econometric models employed to address R1 & R3. Testing hypothes1s H1 will give us various
measures of the ATET. H1 will first compare the probability of the treated group to make the
launch decision at node (b) versus the non-treated group’s probability.
The fact that the hypothesis states the benefit provided by the example EAP is to decrease the
probability of launch might seem strange at first glance, and therefore requires some further
explanation. We know from Shane’s (2008) work and others that the probability of success for
new ventures is in fact quite low (under 50% survival rates after 5 years). One can deduce from
this knowledge that many of the entrepreneurs who launch new ventures, therefore, were either
not appropriately identifying an entrepreneurial opportunity to obtain greater returns on the
utilization of resources than their underlying economic rents, or were not appropriately skilled
enough to capitalize on this opportunity. Therefore, by bringing this to those entrepreneurs’
attention prior to tying up those resources in a failed venture, the EAP can create value for the
entrepreneur through the prevention of a loss. The mere fact that less assisted entrepreneurs
launch products (the weeding out hypothesis) is, by itself, not sufficient evidence to support this
claim, however. Which is why, H1-b (the planting in hypothesis), that those who do launch with
assistance (evidence at nodes d1 & d2) will be more likely to survive than those who do not (d4).
In this case, d3 is left out because assistance after launch is also hypothesized to increase the
probability of survival through different means, since the EAP was not involved in the decision
to launch, and therefore would confuse the analysis.

39

IV.

DATA
The data that will be used to conduct the analysis of the stated research questions was

collected through online and in-person surveys in the months of August, September and October
of 2012. Surveys were sent out online to roughly 2200 Michigan entrepreneurs that were listed in
the Product Center’s database or had applied for a new food license to the Michigan Department
of Agriculture in the past seven years. Recipients of new food licenses were chosen to obtain a
sample of respondents that may be similar to the Product Center’s clientele but may have not had
any contact as of yet with the Product Center. These respondents could potentially be used for
control purposes, though how to do so will be developed below in the discussion related to R2.
From this sample, 617 respondents participated in the survey for a response rate of
approximately 28%. However, respondents who had not launched a new product in the past eight
years, which is the amount of time the Product Center has been in existence, or had incomplete
surveys were screened out leaving only 467 usable observations. In addition, depending on the
particular regression, the amount of usable observations will typically be smaller due to the
tendency for many respondents to leave one or more key explanatory variables blank. For the
purposes of this study, and given the relatively small sample size of complete data, the missing
values will be assumed to be missing at random, and therefore not imputed so as not to further
complicate the analysis.
Furthermore, some of the regressions used below have been done on a particular subsample based on observable characteristics that will also reduce the total number of observations.
For each regression, therefore, the number of respondents used will be explicitly stated.

40

A. Determination of the Treatment Variable – pcassist
In order to understand the impact that assistance has, we must first distinguish between
those who sought assistance and those who did not. In order to do this, the binary variable
pcassist will be used as our primary indicator of treatment. This was done as follows: After
obtaining informed consent for the survey and screening out those who had not launched a new
venture in the past eight years, respondents were asked if they had received assistance from a list
of EAPs within Michigan that included the MSU Product Center. Respondents who indicated the
Product Center had provided them assistance were coded as a one in the binary discrete variable
“pcassist”.
However, after the survey data was collected, all respondents were then compared to the
Product Center’s database to supplement their responses with: the year of first contact with the
Product Center “firstcontact”; the number and type of contact hours they had “contact”; who
their counselor or counselors were and whether they had participated in any Product Center
events such as the annual Making it in Michigan trade show. While this was being done, it
became apparent that many of the respondents who did not indicate that they had received
assistance from the Product Center were in fact in the Product Center’s database. This caused
some concern about whether to code these respondents as a “1” or a “0” for treatment. In the end,
for each respondent who was present in the database, the records of their interactions with the
Product Center were reviewed qualitatively to determine whether or not treatment had been had.
If there was an indication that counseling or services were provided, then they were marked as a
“1” for treatment. On the other hand, if the only interactions were a few phone calls or emails
made to try and set up an appointment but nothing further, then they remained a “0” indicating
no treatment.

41

Additionally, many of these respondents were also in the database for only having
attended the Product Center’s trade show, these were also listed as not having received treatment
as we are defining treatment as receiving counseling and/or services from the Product Center.
However, another variable “pccontact” was created to indicate a “1” if the respondent was in the
database at all, regardless of treatment status, and a “0” otherwise. The results of this procedure
are presented below in Table 4.
Table 4: Summary of the Treatment Variable
Description
Were present in the Product Center’s database
(i.e. pccontact=1)

Respondents

%of Total Sample

362 (out of 467)

76%

Indicated they received assistance from the
MSU Product Center

252 (out of 467)

Were qualitatively determined to have received
assistance despite not indicating so.

60 (out of 110)

13%

Total Treated respondents (i.e. pcassist=1)

312 (out of 467)

67%

54%

There were also a number of respondents (around sixteen) who received assistance from
an EAP other than the Product Center in our sample. For reference, the other EAPs that were
present in the sample were: Starting Block, the Michigan Small Business and Technology
Development Center (SBTDC) and TechTown. These sixteen respondents will be included in the
regression analysis when testing the hypotheses related to who seeks assistance but will be
excluded from the treatment group in the impact regressions in order to focus solely on the
impact that the case EAP has provided. Therefore, the when answering the “Who seeks
assistance question,” the variable (seekassist) that identifies any entrepreneur who has received
assistance from an EAP will be used.

42

B. Performance Variables {Y} of Interest
To maintain consistency with past research, this dissertation will evaluate the MSU
Product Center on the basis of: venture creature, survival, increase in sales or earnings,
employment growth and subjective assessments of the assistance provided. Furthermore, this
research will also extend the measures of performance evaluated to include two measures of
perceived legitimacy: ability to obtain external financing and volume of sales sold through a
formal distribution channel. This section will discuss the rationale behind each of these variables
and how it was recorded in the database.
i.

Measure of Survival: Still in operation – “stillalive”

The first measure of performance we are interested is simply whether or not the new
venture is still in operation. Given the low survival rates of new ventures presented by Shane
(2008) and others, simply increasing survival should be considered a successful influence of an
EAP. To measure this, respondents who indicated that they launched a new venture were also
asked whether that venture was still in operation, those who responded yes were coded as a “1”
for the binary variable “stillalive” and coded as a “0” otherwise. For those indicated no to this
question, they were asked to indicated why this was so from a list of available options.
ii.

Measures of Performance: Current gross annual sales and total employment

In addition, following the work of other EAP researchers (e.g. Robinson, 1982;
Chrisman, Nelson, Hoy and Robinson 1983; and Chrisman, Hoy and Robinson 1987), two
measures of growth: current gross annual sales and current total employment; will also be
compared across the assisted and non-assisted group. For those respondents who indicated that
they had launched a venture, data was collected on their first year as well as the last two years of
gross annual sales and employment levels. While this was initially done to get an indication of
43

percent growth, many respondents only provided the first or last year of sales or employment.
Therefore, these variables were created to indicate that last full year of gross annual sales or
employment provided; age of the venture will then be included within the regression to account
for growth over time. For those respondents who launched in 2012 and only had nine months’
worth of sales data, this value was projected for the full year. Due to the nature of the survey
instrument both of these variables are continuous from 0 to an upper limit ($200,000 for sales
and 250 for employment). When including the upper limits in the regressions, therefore, a righthand censored Tobit model will be used.
iii.

Perceived Legitimacy Measures

In order to measure legitimacy, two different observable variables were chosen based on
their likely correlation with the unobserved perceived legitimacy of the entrepreneur. These two
variables are: ability to obtain external financing and volume of sales sold through a formal
channel (i.e. retail or wholesale). The idea behind these choices is that in order to obtain
financing or gain access to a formal channel, the new venture must have obtained the threshold
level of legitimacy Zimmerman and Zeitz (2002) mention as a necessary prerequisite.
iv.

Ability to Obtain External Financing – “externfinance”

As the author’s master’s work and that of others (e.g. Rotgers et al, 2012) have suggested
that one of the mechanisms by which an EAP can create value for an entrepreneur is through
signaling to investors the viability of the venture concept, and hence increasing the probability of
external financing, examining the differences between the treatment and control group in their
ability to obtain external financing can help determine if this has truly been the case ceteris
paribus. In the case of the ability to obtain external financing, this variable will measure directly
whether or not that threshold has been obtained with external financiers. This variable is
44

measured as a binary variable where “1” indicates that external financing was obtained and “0”
indicates that it was not.
v.

Total Sales Sold through Formal Distribution Channels

Similar to how an EAP can “badge” the worthiness of a new venture to investors
(Rotgers et al, 2012), it can also help to provide this badge to trading partners. One of the key
trading partners for food processors is a distributor who can sell their product to retailers or a
direct relationship with the retailer themselves. In the cases of volume sold through retail and
wholesale, these variables will measure both whether the threshold was obtained with the
retailers or wholesalers themselves, and then how much incremental legitimacy was achieved
with the retail or wholesale consumers.
Respondents who indicated that they had launched a product were also asked what
percentage of their product they sold through retail outlets, wholesale to a commercial buyer and
direct to consumers with the caveat that the three must add to 100%. Using this data, the
percentage sold through each distribution channel was multiplied by grosssales to obtain the
volume sold through those channels. It is important to note here that because grosssales is
censored, those who indicated sales above the maximum had to be excluded from the analysis on
this variable as the percentage sold through that channel multiplied by $200,000 is not an
accurate representation of the volume for that respondent to be compared to other respondents.
For example, a respondent making over $2 million in gross annual sales could indicate 10% sold
through wholesale which would indicate a proper value $200,000, but in this case would appear
as only $20,000 and might be incorrectly categorized lower than a respondent who is grossing
$100,000 annually and sells 50% through wholesale. Therefore, though unfortunate, analyses
using these variables will only be accurate for those entrepreneurs grossing under $200,000
45

annually. Fortunately, this is the majority (248 out of 281 who provided sales data or 88%) of the
sample data collected.
vi.

Subjective Assessments of the Assistance Provided

Any of the respondents who indicated that they had received assistance from an EAP
were then asked to rate, on a scale from 1 to 7, how satisfied they were with that assistance
where 1 indicated very dissatisfied and 7 indicated very satisfied. This variable - “eapsatis” was
collected as a primary source indicator that the EAP had provided some value (or not) to the
respondent.
Similarly, respondents who indicated that they had received assistance from an EAP were
asked to rate, on a scale from 1 to 7, how useful they found that assistance for their new venture
where 1 indicated very useless and 7 indicated very useful. This variable - “eapuseful” was also
collected as a primary source indicator of the EAP value.
As with the previous two variables, a 7-point likert scale was given to respondents who
indicating they had received assistance from an EAP to rate the influence that the assistance had
on their decision to launch or not launch their new venture “eapinfluence, where 1 indicated not
very influential and 7 indicated very influential.
C. Control Variables to be used
In addition to the performance variables, instruments and treatment variable mentioned
above, there are other factors that we would like to hold constant when evaluating the impact of
the EAP on new venture performance. A summary of these variables and the reason for the
inclusion will be presented below.

46

i.

Food Processor Industry Indicator – foodproc

One of the primary industries that the Product Center serves is the food processing
industry. Clients in this industry range from small cottage industry operators who may be selling
cookies or salsa they make out of their home kitchens to large commercial suppliers to retail
chains. In addition, given that in this industry there does not appear to be a dominant design and
there are very low barriers to entry, this industry should allow for strong returns to
entrepreneurial activity (Shane, 2003) and hence EAPs can potentially have a significant impact.
Furthermore, the barriers to entry have recently become even lower with the passage of the
Michigan Cottage Food law, which according to the Michigan Department of Agriculture, allows
entrepreneurs to manufacture and store certain types of “non-potentially hazardous foods that do
not require time and/or temperature control for safety” in an unlicensed home kitchen (Michigan
Department of Agriculture, 2010). Therefore, it is not surprising that the majority of respondents
to this survey (around 70%) are food processors, which the binary variable “foodproc” identifies
with a “1” indicates that the description of their product given was deemed to classify them as a
food processor, and a “0” otherwise. For reference, the industries referenced in the “0” category
consist primarily of agricultural producers, biofuels and non-food products such as soaps and
perfumes.
ii.

Pushed/pulled into entrepreneurship – pushed

Poschke (2008) has found that there is a “U-shaped relationship between
entrepreneurship and ability” indicating that entrepreneurs are typically either pushed (because
of no better option) or pulled (because of superior ability and opportunity recognition) in
entrepreneurship and using education, experience and asking directly why one decided to
become self-employed can help identify these groups. This was done directly and is
47

characterized by the binary pushed variable where “1” indicates an entrepreneur became selfemployed because of “no better options for employment” and “0” indicates an entrepreneur
became self-employed in order to “pursue a profitable opportunity.”
iii.

Contact hours – contact, contact2

As was mentioned above, all respondents were cross-checked with the Product Center
database for when they had contact with the Center, who their counselor or counselors were, and
to determine how much contact hours of assistance was provided. This last variable was
recorded as the continuous variable “contact” and as Chrisman and McMullan (2005) found a
curvilinear relationship between contact hours and performance, a squared term was generated
from the number of contact hours and is labeled “contact2”.
iv.

Venture Age – ventureage

Respondents were asked in which year they launched their new venture, where launch
was defined as collecting sales on the product. This year was subtracted from 2012.75 to give the
venture it’s age. The 0.75 was added to 2012 as data was primarily collected in the 9th month of
the year and so those who launched in 2012 would not have a zero for venture age, which could
cause some difficulties when using this variable to account for growth over time.
v.

Other Controls – inherit, familyentrep, male, white

Blanchflower and Oswald (1998) have also found that receiving a significant inheritance
can in part identify whether an entrepreneur is likely to not have capital constraints, which is
expected to correlate with take-up of assistance (negatively) and performance (positively).
Respondents were asked if they had received such an inheritance and the binary variable
“inherit” is coded as a “1” if so. Having a parent or other close family member entrepreneur,

48

prior entrepreneurial activities and gender (Tamasy, 2011; Yusuf, 2009 & 2012) have also all
been shown to correlate with both the take-up of assistance and overall performance, and so are
being collected and coded as binary response variables “familyentrep” and “male” in which
cases a “1” indicates having an entrepreneur as a close family member or if one is a male,
respectively and a “0” indicates otherwise. Ethinicity will also be controlled for with the binary
response variable “white” which codes a “1” for entrepreneurs who indicate they are Caucasian,
and “0” otherwise.
D. Instrumental Variables (IV) to be tested
Following the discussion on selection bias, it would appear an instrumental variable (IV)
two-staged least squares (2SLS) regression would be the most efficient method for removing
selection bias, with the Heckman two-step correction procedure as an alternative measure that
could provide additional support. However, in order to use these procedures, one must have a
valid instrument.

This study proposes to test two such instrumental variables: proximity to a Product
Center office, and fan of Spartan Athletics. The underlying hypothesis for the first instrument is
that by being located nearer to a Product Center office, than other respondents, one might have a
higher probability of hearing about the assistance offered and come to use the EAP’s services
thus satisfying the cov (Z,P) ≠ 0 condition. In order to satisfy the cov (P,ε) = 0 we would have to
assume that this proximity is not correlated with the entrepreneur’s venture performance, which
on face value might seem reasonable enough unless the location of the offices are also placed in
regions with higher economic activity. Since the EAP in question’s offices are either on MSU’s
campus or placed in extension centers (which are often county seats), this second condition
might prove problematic, but should be tested empirically nonetheless.
49

Given this concern, a second instrument will also be tested: having a friend or family
member who is an alumni, student, faculty or staff of MSU. It is hypothesized that having a
social connection to MSU might induce a higher probability of knowing about MSU services and
hence the Product Center, thus satisfying the cov (Z,P) ≠ 0 condition. And while being connected
to MSU in combination with having attended MSU might both influence performance of one’s
new ventures, if education level is controlled for in the {X} vector, then uncontrolled element of
performance {ε}, will hopefully satisfy the cov (P,ε) = 0 condition. The first condition (cov (Z,P)
≠ 0 ) can be empirically tested by regressing P on Z and the relevant control variables in order to
test the resulting coefficient for significance. The second condition cov (P,ε) = 0 needs have exante validity. If so, it can be used in a Hausman (Wooldridge, 2002) test for the endogeneity of
the treatment where the residuals from the regression of P on Z and X are predicted and then
included in the performance regression. If the residuals are found to have significant predictive
power on the performance variable when included in the second stage regression along with the
endogenous variable, then we can reject the null that there is no endogeneity (i.e. selection bias)
of treatment. These instruments were collected as is described below.
i.

Distance away from the Product Center – disttopc

All respondents were asked for the zip code of their primary residence. Then using an
online mapping service and the zip codes of the 14 Product Center offices around the state, the
minimum distance between the residence zip code and nearest Product Center office were
calculated and recorded as a continuous variable.

50

ii.

Social tie to Michigan State University – msuconnect

All respondents were also asked if they had a close friend or family member who was an
alumni, student, faculty or staff member of Michigan State University. If the answer was yes,
then the binary variable “msuconnect” was coded as a “1” and “0” otherwise.
E. Proxies for Omitted Selection Bias Variables
While it appears to be troublesome to directly observe entrepreneurial ability, though an
attempt will be made, there is perhaps some hope that one can ask respondents directly about
their propensity to seek information and expect honest answers. If any of the proxies are found to
be significant predictors of the probability to seek assistance, they can then be included in the
impact evaluation regressions to remove the selection bias associated with the variables they are
a proxy for. If, however, only one of the two omitted variables can be proxied, then this will act
to provide an upper or lower bound on the treatment effect’s estimator. To be more specific, if
the propensity to seek information proxy is found to be a significant predictor of the probability
to seek assistance, then we can include it in the regressions that evaluate the impact of the EAP
to remove the upward selection bias, but cannot remove the downward selection bias associated
with entrepreneurial ability. While this will not be the most efficient method, it will at least
provide a bounded result. In this case, a lower bound, indicating that the estimates obtained are
likely more conservative than the actual impacts, and vice versa if the entrepreneurial proxy can
be used but not the propensity to seek information one.
The variables that will be measured in order to attempt this will be discussed below.
i.

Propensity to Seek Information – propinfoseek

All respondents were asked to respond what they do when faced with making a decision
under uncertainty on a scale from 1 to 7 where 1 indicated “rely mainly on intuition and trail and
51

error” and 7 indicated “seek out as much information as possible before making an informed
decision.” It is hoped that variable will correlate with the unobserved propensity to seek
information to be used as a control for this factor in order to remove any potential upward bias
on the selection into treatment.
ii.

Propensity to Grow the Business – propgrowbiz

In addition to distinguishing between information seekers and those who rely on
intuition, we were also interested if there was an observable difference in the goals the
entrepreneur had for the business. To be more specific, in terms of the ultimate goal for one’s
business, the literature often distinguishes between “lifestyle” entrepreneurs who are interested in
starting a new venture primarily in order to make a living doing something they have a passion
for and “innovative” entrepreneurs who are interested in trying to grow the business as much as
possible. In order to measure this, all respondents were asked to respond what their primary goal
was for their business on a scale from 1 to 7 where 1 indicated “Provide a fair income for my
family or self “ and 7 indicated “Grow a multi-million dollar business.” This variable was coded
as “propgrowbiz” and is hoped can be used to control for the underlying drive for business
growth that the entrepreneur has.
iii.

Entrepreneurial ability - Education and Experience –baormore and indusexp,

Shane (2003) states the non-psychological factors that influence the expected returns of the
opportunity will be greater for those entrepreneurs who have greater education levels (Casson,
1995), experience in the relevant industry (Knight, 1921; Von Mises, 1949) and previous
experience starting up a business (Jovanovic, 1982). Therefore to proxy for entrepreneurial
ability respondents were asked their education level for which a binary indicator of having

52

achieved a bachelor’s or higher was created (baormore), and the amount of experience in years
they had in an industry relevant to their new venture prior to the launch of their new venture
(indusexp).
F. Entrepreneurial Orientation Proxies
We might wonder if there are not more omitted variables we should be concerned about,
in particular with the differences in the “entrepreneurialness” of the entrepreneurs themselves or
what researchers commonly refer to as an individual’s entrepreneurial orientation (EO).
Entrepreneurial Orientation (EO), as defined by Lumpkin and Dess (1996), refers to the
processes, practices, and decision-making activities that lead to new entry. However, simply
engaging in new entry is not a guarantee of success. Thus, numerous scholarly articles have been
written on examining the relationship between the processes and practices that lead to new entry
and the subsequent success and performance of the relevant firms. In their 1996 paper, Lumpkin
and Dess suggest that EO is composed of five principle components: proactiveness,
innovativeness, risk-taking, competitive aggressiveness and autonomy.
Their model indicates that a higher EO will lead to higher performance levels, but is
moderated by both environmental and organizational factors. In the context of this study,
environmental factors were controlled as much as possible by selecting firms all in similar
industries (i.e. food and agriculture) and in the same state. Furthermore, EO measures are often
applied to businesses and not entrepreneurs themselves, and hence when the survey was
designed, we did not think a measure of autonomy was needed, as all respondents are selfemployed and would be assumed to be fairly autonomous. In reflection, some measure of the
willingness to accept help from others might have been beneficial, however, it terms of its
relationship to both seeking assistance and overall performance.
53

In any case, we will test the whether the other four components are also significant in the
decision to seek assistance through the use of the commonly used measures developed by Covin
and Slevin (1989). What measures will be used and how they were collected will be described
below.
i.

Innovativeness - propinnovate

All respondents were asked to respond on a scale from 1 to 7 in general what they favor,
where 1 indicated “A strong emphasis on the marketing of tried and true products or services”
and 7 indicated “A strong emphasis on R&D, technological leadership, and innovations.” The
recorded response is captured in the discrete variable “propinnovate”.
ii.

Proactiveness – numbernewlines; changesinlines

All respondents were also asked to respond on a scale from 1 to 7 how many new lines of
products or services had their firm marketed in the past 5 years, where 1 indicated “no new lines
of products or services” and 7 indicated “very many newlines of products or services” and this
variable was recorded as “numbernewlines.” Following this question, respondents were also
asked how dramatic the changes in product or service lines have been where 1 indicated “mostly
of a minor nature” and 7 indicated “quite dramatic” and this variable was recorded as
“changesinlines.” However, these variables cannot be considered exogenous as assistance may
influence the number of changes to one’s product lines one makes, therefore they will not be
used as control variables in the regression analysis, but may still be used in the SEM comparison
for identifying underlying latent EO factors.

54

iii.

Risk-Taking – proprisktake; propaggressiveposture

In regards to their risk preferences, all respondents were also asked to respond on a scale
from 1 to 7 in general what they favor, where1 indicated “Low-risk projects (with normal and
certain rates of return)” and 7 indicated “High-risk projects (with chances of very high returns).”
This variable was recorded as “proprisktake.” In addition, respondents were also asked what
their firm does when confronted with decision-making situations involving uncertainty, where 1
indicated “Typically adopts a cautious, 'wait-and-see’ posture in order to minimize the
probability of making costly decisions” and 7 indicated, “Typically adopts a bold, aggressive
posture in order to maximize the probability of exploiting potential opportunities.” This variable
was recorded as “propaggressiveposture.”
iv.

Competitive Aggressiveness

Finally, respondents were asked three questions, on a scale from 1 to 7, regarding how
their firm acted when dealing with their competitors. In the first question, a 1 indicated they
“typically respond to actions which competitors initiate” and 7 indicated they “typically initiate
actions which competitors then respond to,” this variable was recorded as “propinitiatechange.”
In the next question, a 1 indicated they were “very seldom the first business to introduce new
products/services, administrative techniques, operating technologies, etc..,” whereas a 7
indicated they were very often the first business to do such things. This variable was recorded as
“propfirstmove.” While the last variable asked whether they 1 – “typically seek to avoid
competitive clashes, preferring a ‘live and let live” posture” or 7 – “typically adopt a very
competitive, ‘undo-the-competitors’ posture,” this variable was recorded as
“propcompaggressive.” Again, these variables are not completely exogenous so will not be used
as control variables in the regression analysis.
55

G. Summary Statistics
i.

Performance Variables

Presented below in table 5, is a summary of: the names of the performance variables
collected; the description of each variable; the mean, standard deviation and number of
observations tabulated by treatment status; the range of observed responses; and the differences
between the means of the treated and non-treated respondents.
There are a number of interesting results that come from looking at these statistics. To begin
with, one can see by the means subjective feedback on the assistance that respondents received
(eapsatis, eapuseful, eapinfluence) that the average of the Product Center clients is higher on all
three measures when compared to the sixteen receiving assistance from other EAPs. However,
sixteen is not really a large enough number for strong statistical significance in terms of
estimating the actual population mean (Wooldridge, 2002).
Next, as was hypothesized, we can see that indeed Product Center clients have a lower
average of overall entrepreneurs who have actually launched their new venture when compared
to non-clients (73% to 81%), but for those who have launched they are experiencing higher
survival rates (95% to 89%). Nonetheless, it is quite surprising how high the overall survival rate
is for both groups, especially given the research done by Shane (2008), Cooper et al (1988) and
others that have often found very low survival rates, especially given the average venture ages
(as shown in table 5, below) of 3.84 and 3.93 for Product Center clients and non-clients
respectively. This likely is due to a survey response bias where those entrepreneurs who have
dropped their business idea are not very interested in talking about it. For comparison purposes
between the treatment and non-treatment group, however, as long as this response bias is
equivalently present on both sides, it should not affect our analysis of the differences.
56

Table 5: Performance Variable summary statistics tabulated by treatment status

Variable
name
eapsatis

Description
Satisfaction
with EAP
Scale

Product Center
Non-Product
Clients
Clients
Mean
Obs Mean
Obs
(Std.Dev) (N) (Std. Dev) (N)
3.96
247
3.75
16
(1.41)
(1.21)

Range
1-7

Difference
(PCControl)
0.21

eapuseful

EAP
Perceived
Usefulness
Scale

4.16
(1.32)

250

3.81
(1.11)

16

1-7

0.35

eapinfluence

EAP’s
Influence on
Launch Scale

3.34
(1.35)

222

2.88
(1.02)

16

1-5

0.46

launched

Launch Status

73%

310

81%

154

0-1

-9%

stillalive

Indicates if
still in
Operation

95%

219

89%

122

0-1

7%

grosssales

Current Gross
Annual Sales

43,696
(60,620)

173

52,645
(73,715)

108

0–
200,000

-8949.11

Totemp

Current
employment

7.26
(27.82)

194

8.69
(28.01)

104

0 - 250

-1.43

voldirect

Sales sold
direct

11,018
(20195)

155

17,244
(31022)

116

0–
164,541

-6226.28

volretail

Sales sold
through retail

5,854
(15,682)

155

2,959
(8670)

116

0–
62,911

2894.88

volwhole

Sales of
product sold
wholesale

11,893
(23,596)

155

7,658
(18,231)

116

0–
107,324

4234.52

totalinvest

Total Amount
Invested

76,558
(106,402)

221

74,481
(97,709)

124

0–
400,000

2076.71

28%

202

30%

111

0-1

-2%

externfinance Externally
financed

57

Next we can see that total gross annual sales and total employment are both lower for the
treated group than the non-treated group, though there is a lot of variability in both measures as
the standard deviations are in both cases larger than means. This, if nowhere else, shows us that
there is likely some selection bias we must be concerned with when comparing the two groups as
it seems unlikely that providing assistance to entrepreneurs would in fact cause them to have
lower sales and employment. Instead, the factors that are causing those entrepreneurs’ lower
sales and employment data is likely what is causing them to seek assistance.
In terms of the volume of sales through the different marketing outlet, a very interesting
pattern emerges that coincides well with the hypothesis that EAPs help increase the legitimacy of
entrepreneur’s new ventures. Namely, despite the fact that the overall sales level is lower in the
treated group, the volume sold through formal distribution channels, on average, is higher.
Whereas the sales sold direct to consumers is higher in the non-treated group. This could be in
part because we don’t have the volume sold through the different marketing channels for those
grossing over $200,000 due to the inability to interact percentages with upper limits, but it could
also indicate that our hypothesis is correct.
Lastly, we see that the ability to obtain financing is slightly lower for the treated group,
but the amount obtained is slightly higher. Given the small differences, these two findings are
probably best viewed as not significantly different from each other, but as with all of these
factors, more sophisticated data analysis than comparing means is required.
ii.

Instrumental Variables, Omitted Variable Proxies and EO measures

Next, if we turn our attention to the summary statistics for the instrumental variables,
omitted variable proxies and entrepreneurial orientation (EO) measures, we can see there are
58

some potential areas of distinction between Non-PC and PC clients (see table 6, below). In
particular, it appears that PC clients are likely to score higher on almost all the EO measures
except the propensity to take risks (proprisktake) and the propensity towards an aggressive
competitive posture (propaggresiveposture). However, the difference in most of these measures
is quite small, with the exception of the propensity to seek information (propinfoseek) and the
propensity to grow the business (propgrowbiz) measure. This gives some indication that those
who are seeking assistance are likely to have a higher propensity to seek information as
expected, be slightly more risk averse, and a higher propensity to grow their business. The last of
these findings is interesting as it was uncertain whether the Product Center was attracting more
lifestyle or more innovative entrepreneurs. Though it is not surprising that those with a higher
drive to grow their business are also more likely to seek assistance in doing so.
In terms of the other proxy variables, it appears that those seeking assistance are slightly
more likely to have received a bachelor’s or higher in education, while the difference between
the two groups in the amount of industry experience appears to be negligible. Again, the latter of
these two findings is surprising as one would expect that those seeking assistance to have
significantly less experience, but the summary statistics provide no evidence that this is the case.
Finally, in terms of the summary statistics, both instrumental variables appear to have
some potential for satisfying the first condition necessary to be an instrument, as those seeking
assistance are closer in proximity on average to the Product Center and more likely to have a
connection to the EAP’s institutional home. However, as with all of these variables, to determine
if these values are truly significant while holding all other relevant observed control variables
constant, will require regression analysis.

59

Table 6: Summary of Instrumental, Proxy and EO variables

Values
Average of propinnovate
Average of changesinlines
Average of numbernewlines
Average of proprisktake
Average of propinfoseek
Average of propfirstmove
Average of propcompaggresive
Average of propaggresiveposture
Average of propgrowbiz
Average of disttopc
Average of msuconnect
Average of indusexp
Average of baormore

iii.

Non-PC
Client

PC
Client

3.66

3.75

2.94

3.06

3.30

3.30

3.82

3.64

4.50

5.05

4.47

4.62

3.45

3.61

4.06

3.94

3.24

3.85

52.58

47.32

47%

56%

12.07

12.10

60%

65%

Other Control Variables

In addition to examining the differences between the performance variables, it is also
worthwhile to investigate the differences in the summary statistics for the control variables to be
used in this analysis as well.
Except for contact hours, the preceding control and IVs’ name, description, means,
standard deviations (std. dev) if applicable, observations (obv), range observed and difference
between treated and non-treated respondents are summarized below in table 7.

60

Table 7: Summary Statistics for Control and Instrumental Variables Tabulated by Treatment

Variable
name
indusexp

Description
Years of relevant work
experience

male

Gender is male

baormore

Highest education is a
bachelor's degree or more

Non-PC
PC Clients
Clients
Mean Obs Mean Obs
(s.d) (N) (s.d)
(N)
12.1
12.07
(12.6 272 (11.1 132
4)
9)

Range

PCControl

0 - 42

0.03

48%

293

46%

138

0-1

1%

65%

294

60%

139

0-1

6%

84%

294

90%

139

0-1

-6%

white

Ethnicity is Caucasian

familyentrep

Respondent has an
entrepreneur in their family

68%

288

54%

138

0-1

14%

Respondent has received a
significant inheritance

9%

288

7%

138

0-1

2%

Self-employed because of "no
better choices for work"

14%

288

18%

138

0-1

-4%

56%

293

47%

139

0-1

8%

47.32
(52.3
1)

290

52.58
(52.3)

137

5 - 500

-4.99

5.05
(1.73)

281

4.5
(1.8)

130

1-7

0.55

48.57
(11.5
8)

292

49.3
(11.0
5)

139

20 - 74

-0.74

3.84
(2.33)

221

3.93
(2.30)

120

0.75 8.75

-0.10

Key Individuals involved with 4.26
Launch
(5.28)

189

3.18
(3.16)

104

0 - 30

1.08

inherit

pushed

msuconnect

Social tie to MSU

disttopc

Approximate distance in
miles from the nearest PC
office

propinfoseek

age

ventureage

Startupnetwork

Propensity to Seek
Information Scale
Respondent's age

Time (in years) since the
launch of the new venture

61

A comparison of the other relevant control variables reveals some more interesting differences
between the two groups (see table 8, below). It appears that the distribution amongst clients and
non-clients based on their venture age is fairly similar, as is the actual age of the entrepreneur
and the percentage of entrepreneurs who have received a significant inheritance. However, those
who are seeking assistance appear more likely to be of non-White ethnicities, have lower
incomes, a history of family entrepreneurship and a larger start-up network.
Table 8: Comparison of other relevant control variables between PC and Non-PC clients
Values
Non-PC Client
Average of ventureage
3.93
Average of inherit
Average of white
Average of age
Average of hhincome
Average of pushed
Average of familyentrep
Startup-network

PC Client
3.84

7%

9%

90%

84%

49.30

48.57

$ 92,166.65

$84,066.35

18%

14%

54%

68%

3.18

4.26

H. Cluster Analysis on Entrepreneur Types
A cluster analysis was performed on the data on the basis of the control variables to
determine if there were distinct groups of entrepreneurs that could be identified. The results of
this analysis are presented below in table 9. The most insightful results came from when it was
posited there were three distinct clusters. In this case, the main distinguishing characteristics of
the different clusters were: age, experience, whether the entrepreneur was starting-up a new
business and to a lesser extent, education.
62

Table 9: Results of Cluster analysis on control variables
Seasoned
Entrepreneurs

Young Bucks

Late Bloomers

77

117

126

49%

87%

63%

Average of launched

78%

79%

74%

Average of disttopc

39.66

55.35

43.81

Average of msuconnect

57%

55%

55%

Average of startup

56%

80%

76%

Average of age

55.61

35.49

54.65

Average of pushed

18%

17%

13%

Average of eapuseful

4.33

4.10

3.99

Average of eapsatis

4.24

3.81

3.96

Average of familyentrep

66%

65%

65%

Average of white

90%

85%

90%

Average of indusexp

29.48

5.90

6.20

Average of yrsownbiz

16.58

4.15

11.43

Average of male

51%

49%

49%

Average of baormore

65%

74%

58%

Average of pcassist

64%

71%

67%

Cluster
Number of observations
(N)
Average of foodproc

63

i.

Cluster characteristics

The first group that emerges out of the cluster analysis is that of the “seasoned
entrepreneur.” Respondents in this group can be characterized as the older, more experienced
entrepreneurs. They have an average age of 56 years, have worked in an industry relevant to their
new venture for an average of 29 years, have owned their own business for an average of 17
years and are the most likely to have been “pushed” into entrepreneurship. All of these
characteristics are higher than the other two clusters. In addition, this cluster is the most likely to
be expanding their business with new product lines as opposed to starting up a new business
(44% of respondents) and is this least likely to identify themselves as a food processor (49%).
Overall, the seasoned entrepreneurs are the most likely to have an established venture and the
least likely to be seeking out assistance from the Product center. Interestingly, of those that do
seek assistance, they are also the most likely to find that assistance useful and satisfactory, but
the difference from other clusters is small.
By contrast, the second group – the “young bucks,” can be characterized as the younger,
well-educated but inexperienced food entrepreneurs. This group has the youngest average age
(35years), the least relevant industry experience (6 years) and the least amount of time owning
their own business (4 years). They are, however, the most likely group to be starting up a new
business (80%) and have the highest education status with 74% of respondents who have
graduated from a 4-year institution, as compared to 65% with cluster 1 and 58% with cluster 3.
This group also has the highest percentage of food processors (87%). This cluster, therefore,
appears to be identifying the younger, less-seasoned entrepreneurs who are attempting to start-up
a new food business.

64

This brings us to the final group – the “late bloomers”, which is best characterized as
being older less educated entrepreneurs with less relevant industry experience who were “pulled”
into entrepreneurship. This group’s average age is closer to that of the seasoned entrepreneur,
but experience levels closer to that of the young bucks. They are however, more likely to be
starting up a new business and have the lowest percentage of college graduates (58%) but highest
percentage of those who felt “pulled” into becoming self-employed (87%). This cluster appears
to be identifying those entrepreneurs who have likely been working in a traditional job for many
years and are transitioning over to self-employment later in life but do not have the same level of
experience with doing so as the seasoned entrepreneurs.
Interestingly, there does not appear to be much difference on the basis of gender, history
of family entrepreneurship or likelihood to seek assistance between these three groups, but the
older experienced cluster appears to be the group that finds assistance most satisfying and useful.
This is surprising, since one of our hypotheses was that those with less experience would find
assistance more useful, but that does not appear to be the case here.
ii.

Comparison of performance statistics between clusters

When comparing the differences in the averages and medians of the performance
statistics between the three clusters (see table 10, below), we can see that there are some evident
distinguishing characteristics. Most prominently, the seasoned entrepreneurs group, not
surprisingly, has the highest average of overall sales, employment levels and financing. This fits
with the notion of this group consisting of the more established businesses. Interestingly, the
average age of the venture is actually less than the older inexperienced cluster, but it appears this
group has achieved higher levels of legitimacy on the basis of volume in retail and especially
with the ability to obtain external financing.
65

Table 10: Comparison of performance statistics between clusters

Number of Observations

Seasoned
Entrepreneurs
77

117

Late
Bloomers
126

Young Bucks

Average Still in Operation

88%

92%

97%

Average Total Employment

15.30

3.15

8.35

$20,329

$11,737

$19,249

% Obtained External Financing

40%

30%

21%

Average Age of Venture

4.00

3.24

4.51

% Start-ups (as opposed to expansions)

56%

80%

76%

Median Gross Sales

However, there is one performance characteristic by which the seasoned entrepreneurs
are underperforming in, compared to the other two clusters, and that is with survival, as they
have the lowest overall percentage of being still in operations (88% compared to 92% and 97%
for clusters 2 and 3, respectively). This may indicate that the larger operations carry with them a
higher risk during the economic downturn of the past few years, as larger operations also have
larger commitments to creditors and employees that are harder to meet.
For the two less experienced clusters, on the other hand, we can see that while both have
lower sales and employment than the seasoned entrepreneurs, the late bloomers have on average
higher sales and employment than the young bucks but are less likely to obtain external
financing. It may be that the late bloomers are less appealing to investors, than a younger more
educated entrepreneur, but in fact are actually more successful at sustaining and growing their
business.
When drilling down to the differences between Product Center clients and non-clients in
amongst these clusters some very striking differences appear (see table 11, below). To begin
66

with, when we don’t control for the size of the firm, we can really see the impact of selection
bias with the seasoned entrepreneurs. In this case, as we saw earlier, non-clients outperform the
Product Center clients on total sales and employment, though Product Center clients were more
likely to still be in operation with older ventures and have obtained external financing. Amongst
the inexperienced clusters, however, there doesn’t appear to be a significant difference between
the PC-assisted and non-assisted groups, except perhaps with late bloomers having slightly
smaller levels of total employment, but slightly higher levels on the other performance statistics,
whereas with the young bucks have slightly lower performance statistics on all the categories
except % still in operation.
Table 11: Comparison of performance statistics between PC clients and non-clients by cluster.
Seasoned
Entrepreneurs

Young Bucks

Late Bloomers

Non-Client

PC Client

NonClient

PC
Client

NonClient

PC
Client

28

49

34

83

42

84

% Still in Operation

77%

95%

91%

93%

94%

98%

Avg. Total Employment

21.38

11.66

3.21

3.11

9.00

8.04

$32,864

$15,024

$11,737

$11,573

$18,310

$20,828

Avg. Volume in Retail

N/A

N/A

N/A

N/A

N/A

N/A

% External Financing

37%

42%

33%

28%

17%

23%

Average Age of Venture

3.65

4.21

3.53

3.08

4.68

4.42

% Start-up

42%

63%

93%

73%

79%

75%

Number of Observations

Median Gross Sales

When, however, we focus on only those firms that have gross annual sales under
$200,000, really striking differences can be observed (see table 12, below). For the older
experienced cluster, we can see that there is less of a performance gap in sales and employment,
67

though Product Center clients are still smaller on average than the non-clients, but are achieving
higher levels of legitimacy and survival than their non-client counterpart.
Table 12: Comparison of performance statistics between Product Center clients and non-clients
for firms grossing under $200,000
Seasoned
Entrepreneurs

Young Bucks

Late Bloomers

Gross Sales under
$200,000
# of Observations

NonClient

PC Client

NonClient

PC
Client

NonClient

PC
Client

23

44

30

80

37

80

% Still in Operation

71%

94%

89%

93%

92%

98%

Avg. Employment

4.25

3.67

2.00

3.02

2.62

3.78

Median Gross Sales

$14,381

$13,224

$4,695

$7,905

$9,450

$19,329

Avg. Volume in Retail

$2,250

$5,387

$1,979

$4,459

$2,939

$8,527

% External Financing

29%

41%

27%

26%

13%

19%

Avg. Age of Venture

3.94

4.23

3.39

3.08

4.71

4.26

% Start-up

57%

63%

92%

75%

83%

75%

For both the inexperienced clusters, on the other hand, Product Center clients are
outperforming their non-client counterparts in terms of survival rates, employment, sales, and the
legitimacy measures of volume in retail and obtainment of external financing (except for the
younger inexperienced group where they are slightly lower but practically tied with the nonclient. Potentially indicating that the Product Center’s services have helped the older experienced
clients with smaller businesses achieve greater legitimacy, while also helping the more
inexperienced clients achieve greater performance.

68

V.

WHO SEEKS ASSISTANCE?
At this point we will begin to provide answers to the research questions stated at the

beginning of this document by testing the hypotheses posed above. As has been previously
discussed, before we can answer research questions (1) or (3) via regression analysis, we must
determine if we can control for the inherent selection bias in the EAP-process that would
invalidate any regression results. To accomplish this, therefore, we need to determine if we can
empirically distinguish between those entrepreneurs who seek assistance and those who do not
through the answering of research question (2), and its associated hypotheses (see table 3b, page
35). This is done because, as Heckman (1979) has shown, selection bias is only a concern when
the cause of the bias is omitted from the regression and hence causes the error term to correlate
with the explanatory variable of interest. Therefore if this omitted variable can be observed
through the testing of (H2) & (H3) or instrumented away, then we can effectively control the
selection bias on the treatment variable (pcassist).
A. Fit within the decision-tree framework
In particular, we will wish to determine whether those observable differences provide
empirical support for the aforementioned causes of selection bias: higher propensity to seek
information and lower entrepreneurial ability. However, it is important to distinguish what point
in the conceptual model (see figure 1, page 36) that we are testing when an entrepreneur decides
to seek assistance (or not).
B. Methods
To first test these hypotheses, any respondent who has sought assistance at points (a1) or
(c3) will be compared to the set of respondents that has never sought assistance (b4) & (c4) via

69

probit regression analysis of the treatment variable (pcassist) on the control variables listed
above in the data section plus the omitted variable proxies for the propensity to seek information
(propinfoseek) and entrepreneurial experience or ability (indusexp and baormore).
i.

Probit regression analysis

To conduct a probit regression we can postulate there exists a latent factor Y*, which
represents the perceived value of assistance to the entrepreneur, that will determine whether or
not that entrepreneur decides to seek assistance, such that if Y* > 0 the entrepreneur will seek
assistance and if Yi* ≤ 0, the entrepreneur will not seek assistance. Furthermore, let us assume
that Yi* is influenced both by observable characteristics represented by a control vector {Xi}
such as the venture’s age, the entrepreneur’s experience, education and so on as well as the
influence of the omitted variable proxies{Zi} has had. It is likely that the perceived value of
assistance is also influenced by unobservable characteristics (to us) as well such as the
entrepreneur’s willingness to accept help, perception of his or her own entrepreneurial ability,
trust of the EAP and so on, that will be represented by the error term {εi}. Then we can
characterize the perceived value of assistance as (4):
Y௜∗ = ܺ௜ ߚ + ܼ௜ ߙ + ߝ௜ 	(݅ = 1,2 … , ܰ)

(4)

However, as Yi* is a latent variable we only observe its sign indicated by the treatment
dummy variable {Di} where Di = 1 (if the entrepreneur sough assistance) if Yi*>0 and Di =0 if
Yi* ≤ 0. Then we can rewrite equation (4) as follows:
ଵ:		௜௙	௑ ఉା௓ ఈାఌ 	வ	଴

	‫ܦ‬௜ = {଴:	௜௙	௑೔೔ఉା௓೔೔ఈାఌ೔೔		ஸ	଴ 	

(4a)

Then, using equation (4a), then we can see that the probability that an entrepreneur (i)
seeks assistance (i.e. Di =1) can be written as follows in equation (4b):
(4b)

ܲ‫ߝ(ܾ݋ݎ‬௜ ) 	 > −ܺ௜ ߚ−ܼ௜ ߙ	
70

Next, if we assume that the error term {εi} follows an independently and identically
distributed normal distribution with mean (0) and variance σε2, then the probability that εi is
larger −ܺ௜ ߚ−ܼ௜ ߙ is simply 1 – Φ(−ܺ௜ ߚ−ܼ௜ ߙ) where Φ represents the normal cumulative
density function (CDF). Furthermore, since the normal distribution is symmetrical in nature, 1 –
Φ(−ܺ௜ ߚ−ܼ௜ ߙ) is equivalent to Φ(ܺ௜ ߚ+ܼ௜ ߙ). Therefore, the probability that an entrepreneur’s
new venture is still alive, given the above assumptions, can be characterized by equation (4c):
(4c)

ܲ‫	ܦ(ܾ݋ݎ‬௜ = 1|ܺ) = ߔ(ܺ௜ ߚ + ܼ௜ ߙ)

This equation can be solved using Maximum Likelihood Estimation (MLE) to obtain an estimate
of the impact the omitted variable proxies have on the decision to seek assistance represented by
ߙො (Wooldridge, 2002).
In addition, we can test (IV-1) & (IV-2) through determining whether the proposed
instruments (disttopc) and (msuconnect) respectively are significant predictors of the decision to
seek assistance, providing a necessary but not sufficient step in validating their use as an
instrumental variable (IV). As was stated earlier, if the instruments are shown to be significant
first stage predictors, then a Hausman test will be used to see if the instruments can reject the no
endogeneity null in the second stage (see Wooldridge, 2002). If the instruments fail to satisfy
both of these conditions, they will not be considered valid. Since we have also already rejected
the use of Propensity Score Matching, the only other option would be including the unobserved
selection bias variables proxies.
We might also expect the selection bias forces, namely the propensity to seek information
and entrepreneurial experience and/or ability to be more significant in those who seek assistance
prior to the launch of their products than those who seek it only afterwards. To determine if there
is any difference between those who seek assistance before or after the launch of their product,

71

the same regressions will be run comparing those at point (a1) to those at point (c3) through
restricting the sample to only those who sought assistance (i.e. pcassist = 1) and using the
variable prepcassist (for which a 1 denotes that the entrepreneur sought assistance prior to the
launch of their product) as the regressand. This will also help to provide the foundation for some
of the insights obtained in Chapter 7 when we drill down into how assistance affects different
entrepreneurs differently.
C. Factors that influence the decision to seek assistance
In order to answer the research question of what types of entrepreneurs seek assistance an
unrestricted probit of treatments status (evidenced at points a1 & c3) was regressed on the
control vector, omitted variable proxies and the instruments to test H2 & H3 provides us with the
following results (see table 13, below):
Table 13: Unrestricted probit regression of treatment on controls and instruments (statistically
significant variables in bold, N=346, pseudo R2=0.0717)
Y=pcassist

Coef.

Std. Err.

z

P>|z|

propinfoseek
proprisktake
propinnovate
propgrowbiz
age
indusexp
baormore
male
white
familyentrep
inherit
pushed
msuconnect
disttopc

0.11
-0.05
0.03
0.08
0.00
0.00
0.26
0.05
-0.45
0.45
0.11
0.00
0.03
0.00

0.04
0.05
0.04
0.04
0.01
0.01
0.16
0.15
0.25
0.15
0.31
0.20
0.15
0.00

2.53
-0.96
0.64
1.88
-0.25
-0.04
1.66
0.36
-1.84
2.96
0.37
0.02
0.18
-1.50

0.01
0.34
0.52
0.06
0.80
0.97
0.10
0.72
0.07
0.00
0.71
0.98
0.86
0.13

72

[95%
Conf.
0.02
-0.15
-0.06
0.00
-0.02
-0.01
-0.05
-0.24
-0.94
0.15
-0.49
-0.40
-0.27
-0.01

Interval]
0.19
0.05
0.11
0.15
0.01
0.01
0.57
0.35
0.03
0.75
0.71
0.40
0.32
0.00

The results of this probit provide some interesting insights. To begin with, it appears that
we can reject the null for H3 as the variable propinfoseek is a strong predictor of seeking
treatment. This is good news, as it is hoped that this variable can proxy for the unobserved
propensity to seek information on the part of the respondents, such that when included in the
performance regressions it controls for what would otherwise be an upward bias on the treatment
effect. Given the construction of this variable and confirmation of its expected positive
prediction of treatment, there appears to be support for including it as such a control. There was
not strong evidence, however, that the propensity towards taking risks or innovating new
products was a significant influence in the decision to seek assistance from the Product Center.
However, in conformation with what was seen in the summary statistics, there was a significant
correlation uncovered with the propensity to grow the business, with seeking assistance. The
combination of the propensity to seek information and grow the business confirms our suspicions
that there is some concern of selection bias on the basis of the proactiveness and motivations of
the entrepreneur. However, since we appear to be able to observe this element of selection bias,
we can include these two factors as controls in the performance evaluation to remove this
concern.
In contrast, if we believe education is an adequate proxy for entrepreneurial ability, we
are finding evidence that contradicts our hypothesis that those with lower entrepreneurial
abilities will be more likely to seek assistance. However, whether education level is a good
measure of ability is uncertain as the intellectual abilities and skills set required to do well in an
academic environment are not the same as the ones required to succeed as the manager of one’s
own business. While this finding still remains an important observation, there are other reasons
why it might be occurring. First, the fact that the Product Center resides within a university

73

setting could indicate that those more familiar and comfortable with institutions of higher
education (as indicated by their ability to have earned a Bachelor’s or higher) are more likely to
use the services provided by a University. In addition, though propensity to seek information and
a connection to MSU are also being controlled for, there could be an additional element at play
in the philosophical value that higher education provides. Namely, it could be the obtaining of
bachelor’s degree or higher has helped individuals to develop greater wisdom in the Socratic
sense that, “True wisdom comes to each of us when we realize how little we understand about
life, ourselves, and the world around us.” (Plato, Republic) and thus has made those individuals
less reliant on their own cognitive biases even above their innate propensity to seek information.
In any case, the education control variable will be kept in the performance evaluations, but it
does not appear that it is adequately controlling for entrepreneurial ability.
The other variable used to proxy for entrepreneurial experience, is the variable
(indusexp), does not appear to have any significant predictive powers. Though industry
experience is not perfectly correlated with entrepreneurial experience, it is hope that there would
be some correlation to indicate its use as a proxy for this selection bias force. The fact that it is
not, and that the education variable had the opposite sign then expected leads us to not be able to
reject the null in relation to H2.
Most of the other controls do not appear to be significant predictors of treatment except
for being non-white and having an entrepreneur in the family. The vast majority of respondents
in both groups are Caucasian, but it would appear amongst the set of entrepreneurs that are not
Caucasian, significantly more are seeking assistance. It is also interesting that the presence of an
entrepreneur in one’s family is a strong predictor of seeking assistance, as one might initially
expect the reverse to be true given that having an entrepreneur in the family might act as a

74

substitute for the assistance counseling provides. Perhaps, however, having first-hand knowledge
of the many difficulties associated with entrepreneurship causes one to be more attuned to the
support networks out there available to overcome such difficulties. More research is therefore
needed on both of these factors.
D. Instrumental Variables
Also, it appears neither the distance to the Product Center (disttopc) nor having a social
tie to MSU (msuconect), the two proposed instrumental variables, are showing any promise for
being a valid instrument given the results presented above in table 13. This is true even if we
restrict the sample to the more homogenous group of only food processors as can be seen below
in table 14.
Table 14: Probit regression of treatment on controls and instruments restricted to food
processors (statistically significant variables in bold) (N=237, Psuedo R2=0.0813).
Y=pcassist
propinfoseek
proprisktake
propinnovate
propgrowbiz
age
indusexp
baormore
male
white
familyentrep
inherit
pushed
msuconnect
disttopc
_cons

Coef.
0.06
-0.05
0.02
0.12
-0.01
0.00
0.14
0.13
-0.33
0.45
-0.25
-0.09
0.19
0.00
-0.07

Std. Err.
0.05
0.06
0.05
0.05
0.01
0.01
0.19
0.18
0.29
0.19
0.38
0.25
0.19
0.00
0.64

z
1.11
-0.88
0.41
2.52
-0.59
0.24
0.74
0.74
-1.17
2.38
-0.66
-0.36
1.04
-1.31
-0.11

75

P>|z|
0.27
0.38
0.68
0.01
0.55
0.81
0.46
0.46
0.24
0.02
0.51
0.72
0.30
0.19
0.91

95%
Conf.
-0.04
-0.17
-0.08
0.03
-0.02
-0.02
-0.23
-0.22
-0.89
0.08
-1.00
-0.58
-0.17
-0.01
-1.33

Interval
0.16
0.07
0.13
0.22
0.01
0.02
0.51
0.49
0.23
0.82
0.50
0.40
0.56
0.00
1.19

In this case, the family history of entrepreneurship and propensity to grow the business
remain significant predictors of the decision to seek assistance, but the education variable and
propensity to seek information variables have both dropped slightly out of significance. This
might indicate there is less heterogeneity amongst this particular subset of respondents than the
entire sample, as one can see age, industry experience, education, gender, reason for becoming
self-employed, and the proxy of capital constraints (inherit) are also not significant. Also, it
appears for this group that the social tie to MSU is also significant, though just barely so. This
gives some hope that either: 1) selection bias is not that strong amongst food processors and/or 2)
the MSU connection IV can be used to moderate the impact of the selection bias that is occurring
within this industry.
Furthermore, if one restricts the subsample to the respondents that have provided sales,
neither of the instrumental variables are significant. This is shown below in table 15 in
relationship to the subsample of respondents that are both food processors and have provided
sales data. This holds true for the subsamples that have provided employment and investment
data as well. It should be noted that in order to provide this data a respondent must have already
launched a new venture, and of course made the decision to reply to those questions, which many
often choose to leave blank. With this subgroup, we also see that the ethnicity variable becomes
significant again. The propensity to seek information also remains significant, so not all hope is
lost in regards to dealing with selection bias as bounded results of the estimates will still be
feasible.

76

Table 15: Probit regression of treatment on controls and instruments (statistically significant
variables in bold) restricted to those providing sales data (N=328, Pseudo R2 = 0.0654)
Y=pcassist
propinfoseek
proprisktake
propinnovate
propgrowbiz
age
indusexp
baormore
male
white
familyentrep
inherit
pushed
msuconnect
disttopc
_cons

Coef.
0.09
-0.04
0.02
0.05
0.00
0.00
0.25
0.02
-0.48
0.47
0.24
0.03
0.05
0.00
-0.05

Std. Err.
0.04
0.05
0.05
0.04
0.01
0.01
0.16
0.15
0.26
0.16
0.34
0.21
0.16
0.00
0.58

z
2.09
-0.68
0.36
1.22
0.01
0.28
1.57
0.13
-1.84
2.96
0.69
0.14
0.30
-1.36
-0.08

P>|z|
0.04
0.49
0.72
0.22
0.99
0.78
0.12
0.89
0.07
0.00
0.49
0.89
0.77
0.17
0.94

[95%
Conf.
0.01
-0.14
-0.07
-0.03
-0.01
-0.01
-0.06
-0.28
-0.99
0.16
-0.43
-0.39
-0.26
0.00
-1.18

Interval]
0.18
0.07
0.11
0.13
0.01
0.02
0.57
0.32
0.03
0.77
0.90
0.45
0.35
0.00
1.09

In order to understand why these IVs were not significant, a power analysis can be done
on the distance variable, for which a mean and standard deviation is available, to determine what
an appropriate sample size should be in order to obtain have confidence that we are not
committing a Type II error of failing to reject the null of no significance difference between the
means of the treated and non-treated when we should. In this case, given the means and standard
deviations reported above in table 5, a significance level of 0.1 and a power level of 0.8, then one
would need a sample size of 1,223 treated entrepreneurs and 1,223 non-treated entrepreneurs.
This study’s response rate is well below those numbers, so we cannot reject the possibility that
these are valid IVs, just given the data we have they do not appear to be effective for our needs
and therefore we also cannot use either as an instrument.

77

E. Who seeks assistance prior to the launch of their venture?
Next, the analysis on the difference between those who sought assistance prior to the launch
of their product versus those who only sought assistance after the launch of their product was
conducted with the results presented below in table 16. To test for this difference, another
variable “prelaunchpcassist” was created to determine if the entrepreneur first came in contact
with the Product Center prior to the launch of their product had any effect on perceived
usefulness. This binary variable was created by first subtracting the year of the venture’s launch
from the year of first contact with the Product Center. If this was value was zero or less, then it
was coded as a “1” if it was greater than zero than it was coded as a “0”. Then it was multiplied
by treatment status to identify those that received treatment versus those that merely had contact.
Table 16: Probit Regression of the decision to seek assistance prior to launch (statistically
significant variables in bold N=234, Pseudo R2=0.0422).
Y=
prelaunchpcassist
propinfoseek
proprisktake
propinnovate
propgrowbiz
Age
indusexp
baormore
Male
White
familyentrep
Inherit
Pushed
msuconnect
Disttopc
_cons

Coef.
0.0870133
-0.051688
0.0754134
-0.011231
0.0028861
-0.012453
-0.218399
0.1519751
-0.277568
-0.159938
0.1614859
-0.094056
-0.053271
0.002268
0.3493327

Std. Err.
0.0531446
0.0617804
0.0529411
0.0470146
0.0084428
0.0074234
0.1921471
0.1838781
0.2591673
0.1982729
0.3626521
0.2602949
0.1835496
0.0019084
0.6697895

z
1.64
-0.84
1.42
-0.24
0.34
-1.68
-1.14
0.83
-1.07
-0.81
0.45
-0.36
-0.29
1.19
0.52

78

P>|z|
0.102
0.403
0.154
0.811
0.732
0.093
0.256
0.409
0.284
0.42
0.656
0.718
0.772
0.235
0.602

[95%
Conf.
-0.017148
-0.172776
-0.028349
-0.103378
-0.013661
-0.027003
-0.595
-0.208419
-0.785527
-0.548546
-0.549299
-0.604224
-0.413022
-0.001472
-0.963431

Interval]
0.1911748
0.069399
0.1791761
0.080916
0.0194336
0.0020963
0.1582024
0.5123695
0.2303901
0.22867
0.8722709
0.4161133
0.3064792
0.0060084
1.662096

Interestingly, in this case we find support for rejecting the nulls for H2 & H3 in regards to
how the selection bias forces act based on the timing of assistance. In other words, the propensity
to seek information variable (propinfoseek) and the industry experience variable (indusexp)1 both
are significant (though only at the 90% confidence level) of the decision to seek assistance prior
to the launch of the entrepreneurs’ product, while no other controls or instruments are significant.
This would imply that those who seek assistance prior to the launch of their product are more
likely to have both characteristics associated with the two forces of selection bias – high
propensity to seek information and lack of experience. This will be important to bear in mind
when conducting the analysis on how assistance affects entrepreneurs differently based on when
they received assistance.
Taking this all into account it appears we can reject the null in regards to H2 though we
cannot reject the null for H3. In this case, there does appear to be evidence that those with a
higher propensity to seek information are more likely to seek assistance. It also appears that this
can in part be observed and controlled for and so can be left out of the selection bias concern.
Furthermore, it also appears the assisted group has other significant observable differences as
well. Though both the treated and non-treated samples contain primarily Caucasians, the nonCaucasians appear to be more likely to seek assistance. In addition, those with a history of
entrepreneurship in their family also are more likely to seek assistance, so the two populations
cannot be deemed to be exactly the same. This is why regression analysis that can hold these
factors constant will be important in determining what the overall impact effect should be, and
why one cannot rely on a simple comparison of means.

1

The negative coefficient on industry experience indicates that those with less experience are more likely to seek
assistance prior to the launch of their product.

79

F. Implications of these Results
One of the most important implications to arise out of these results is the indication that
the entrepreneur who seeks assistance is categorically different than the one who does not seek
assistance. This has been shown on the basis of family history of entrepreneurship, ethnicity, the
propensity to seek information and entrepreneurial orientation. Given that these observable
characteristics have been found to have an impact on overall performance of a new venture, one
cannot assume that strong ignorability of treatment (from Rosenbaum and Rubin, 1983) or even
the weaker conditional mean equivalence condition (from Heckman et al, 1998) to be satisfied in
order to conduct propensity score matching techniques to uncover the average treatment effects
of EAPs. That others have tried this (e.g. Rotgers and Story, 2012; Kobsters and Obschonka,
2012) and have not found significant results is therefore not surprising given the lack of
foundation for the model’s effectiveness.
Instead, what is needed is a valid instrumental variable in order to identify the exogenous
component of treatment in a two-stage least squares model or Heckman control function.
Unfortunately, this research did not uncover such an instrument, so this is an area where further
work will be required. In this case, distance did not prove all that helpful, which may be due to
significant coverage provided by Product Center offices throughout the state, as well as the
mobility of innovation counselors who, as extension agents, are skilled in extending the range of
their services to those who are not in close proximity. A social tie to the university showed more
promise as an instrument, but given the sample size, was still only a weak instrument and was
not able to identify enough of the exogenous variance associated with treatment to provide
reliable results.

80

Even without a strong instrument, however, it would appear that bounded results can be
obtained through including an observable measure of the entrepreneur’s propensity to seek
information in order to remove the ambiguity of the selection bias’ direction. This can be done
through creating a direct scale where respondents indicate their own preference on seeking
information versus relying on intuition and trial and error. This retains the validity of positive
results, but there would need to be more work done to determine if negative results are credible
as well or if they are due to the downward bias of the selection effect.
It also appears that impact of selection bias is strongest on those who seek assistance
prior to the launch of their products than those who seek it only afterwards. When one includes
both pre-launch and post-launch only assistance seekers in the treatment group, therefore, it may
be that one is mixing two categorically different types of entrepreneurs into one, thereby
confounding both the estimated average treatment effects as well as any steps taken to control
selection bias. Instead, the two groups should be estimated separately, in addition to combined,
in regards to the impact of assistance to obtain a more complete picture.
It will be up to future researchers to determine if the aforementioned controls are
sufficient, or if other observable measures could be introduced to remove the remaining selection
biases such that the Gauss-Markov conditions for OLS, probit and Tobit can be reasonably
assumed to be valid. If so, given its efficiency and parsimony, the non-instrumented regression
techniques would be the preferred method of estimation of cross-sectional EAP treatment effects
for future impact studies. For the purposes of this paper, however, we will assume that by
including the variable “propinfoseek” in all the following regressions, only the downward bias
associated with entrepreneurial ability remains, thereby allowing all positive findings to be
considered conservative estimates that will still allow for the rejection of the null hypothesis.

81

Given that there is still some concern regarding unobservable variables biasing the
resulting estimators tested, a panel study that resamples the respondents of this survey a few
years in the future to conduct a difference in difference analysis (see Wooldridge, 2002) using
the data collected with this study as a baseline would allow researchers to control for the
unobservables and obtain more consistent results.

82

VI.

WHO FINDS ASSISTANCE USEFUL?
The next research question we must address before coming to the impact evaluation is

research question (R3):
What types of entrepreneurs find assistance useful and why?

(R3)

By answering this question prior to the impact evaluation we can get a better
understanding of how EAP clients perceive the impact the EAP is having. As was mentioned
above in the discussion on the Product Center process, entrepreneurs can choose to seek
assistance prior to the launch of their venture, while the idea is still in the development phase, as
well as after launch as needs for assistance arise. The decision on when to come in, however, is
not random, as we have seen that the more proactive information seekers and the less
experienced to seek out assistance prior to the launch of their products. This creates a selection
effect within the already discussed selection that we will need to evaluate and see if there are
significant differences between the two groups so these differences can be accounted for when
evaluating the impact assistance is bringing to each. Therefore, we need to test H4 its subhypotheses, which are summarized below with the other relevant hypotheses for R3 (see table
3c, page 35):
A. Methods
In order to answer the “what type of entrepreneurs find assistance useful” part of R3 a
probit regression analyses will be done following the same theoretical background as stated in
section VI.A. By nature of these hypotheses, only those entrepreneurs who have received
assistance will be included within the analyses for who found that assistance useful. For H4 and
its sub hypotheses (H4-a & H4-b) entrepreneurs who sought assistance prior to the launch of
83

their new venture (node a1 in the decision-tree framework, see figure page 36) will be compared
to those who sought assistance only after the launch of their venture (node c3). For H5 only
those entrepreneurs within nodes c1 & c3 will be compared to see how the respondents control
variables predict usefulness. For H6 an additional variable based on firm’s gross annual sales
will be added, and H7 will attempt to use the observed characteristics associated with venture
success to predict perceived usefulness of assistance. Insight will also be gleaned from
respondents comments to answer the “why” part of R3 using inductive reasoning.
B. Influence of timing of assistance on perceived usefulness
To test H4, the variable “prelaunchpcassist” was used a dependent variable in a probit
regression (using the methodology described in the previous chapter) to determine if any
observable characteristics could be identified that distinguished the group of respondents who
sought assistance prior to launch when compared to those who sought assistance only after
launch in order to test the sub-hypotheses H4-a and H4-b. These hypotheses can be tested from
the same regression conducted above (see table 16, page 79).
Interestingly, only industry-related experience and the propensity to seek information appears
to be driving this decision, whereas none of the entrepreneurial orientation or other demographic
variables appear to be significant. This tells us that we can weakly reject the null of H-4b in
relation to how experienced the entrepreneur is and H4-a in regards with their propensity to seek
information. This is because, other than with the amount of industry experience and information
seeking tendencies a respondent had, there does not appear to be much difference between those
who seek assistance prior to the launch of their product and those who seek it afterwards. Once
including industry-related experience and our proxy variable for the respondent’s propensity to
seek information as a control, this allows us to observe if there is any incremental difference in
84

the value of the assistance received, as well as the perceived usefulness of that assistance, based
on whether it occurs before or after the product has been launched.
Given that we just found that the only significant difference between those who seek
assistance prior to the launch of their product, and those who seek it afterwards was their level of
industry-related experience and propensity to seek information, this allows us a good opportunity
to test if those who participate in the full Product Center venture development process prior to
launching their product receive higher gains than those who come after launch (through the
weeding out and planting in process). This can be done through the testing of H4.
The first method we can test this hypothesis is to simply evaluate whether those
entrepreneurs who received assistance and have launched their product (regardless of the timing
of assistance) rated the assistance they received 4 or higher on a scale of 1 to 7 of perceived
usefulness (i.e. the variable eapuseful), where 7 indicates extremely useful and 1 indicates
extremely useless. Those respondents who met this condition where coded as a 1 for the
indicator variable pcusefullaunch, while those who did not were coded as a 0. In addition, many
respondents were present in the product center’s database, but had not indicated that they
received assistance from the Product Center. These respondents were then assumed to have
determined the assistance not to be useful, and were included in the 0 group for the
pcusefullaunch variable, if they had also gone on to launch a product. Then, by restricting the
sample to those who have had contact with Product Center’s assistance and launched a product,
we can conduct a probit regression on the variable pcusefullaunch in order to identify whether
prelaunchpcassist is a significant predictor of usefulness. This is done below in table 17.

85

Table 17: Probit regression of probability to seek assistance and find it useful on timing of
assistance and controls restricted to respondents who launched a product and had some form of
contact with the Product Center (statistically significant variables in bold N=200, Pseudo R2 =
0.1003)
pcuseful
prelaunchpcassist
propinfoseek
proprisktake
propinnovate
age
indusexp
baormore
male
white
familyentrep
inherit
pushed
disttopc
_cons

Coef.
0.65
0.03
-0.03
-0.01
-0.01
0.00
0.37
-0.16
-0.23
0.28
0.06
-0.04
0.00
0.67

Std.
Err.
0.19
0.06
0.07
0.06
0.01
0.01
0.21
0.20
0.34
0.21
0.37
0.27
0.00
0.73

z
3.33
0.63
-0.40
-0.25
-1.44
-0.15
1.79
-0.82
-0.67
1.30
0.15
-0.14
-0.64
0.92

P>|z|
0.00
0.53
0.69
0.80
0.15
0.88
0.07
0.41
0.50
0.19
0.88
0.89
0.52
0.36

[95%
Conf.
0.27
-0.07
-0.16
-0.13
-0.03
-0.02
-0.03
-0.55
-0.89
-0.14
-0.66
-0.57
-0.01
-0.76

Interval]
1.03
0.14
0.10
0.10
0.00
0.02
0.77
0.23
0.44
0.69
0.78
0.49
0.00
2.09

Here we can see very clearly that those who seek assistance prior to the launch of their
product are much more likely to find that assistance useful than those who seek it only
afterwards. So there is some evidence that those with less experience and higher information
seeking propensities are more likely to seek assistance prior to the launch of their product, and in
turn those who seek assistance prior to the launch of their product are finding that assistance
more useful. This leads us to believe that those with more experience and have been successful
in launching a product also find the assistance less useful giving some compelling support for
H4-b.
In addition, we are also seeing that those with higher education levels are also finding
assistance more useful, which may be due to the fact that the Product Center operates within a

86

university setting so those who have successfully navigated such a setting in the obtainment of
their master’s degree possess similar skills and aptitudes to navigate the process of receiving and
using the Product Center’s assistance in a way that is perceived useful. Also note, industryrelated experience is included as a control in this regression so that element of the selection of
who seeks assistance prior to the launch of a product is being controlled for.
C. Perceived usefulness amongst those who launch new ventures
Next, it is fairly straightforward to use this data to test H5 which states:
Entrepreneurs who have launched products and have less entrepreneurial
experience and probabilities of success without assistance will be more likely to
find assistance useful. (H5)

In order to test this hypothesis, the analysis was first restricted to only those respondents
who had some contact with the Product Center, since those who did not have contact would not
be able to comment on its perceived usefulness. The variable pcusefullaunch was then created as
a binary indicator variable that is coded as a “1” if the respondent indicated receiving assistance
and rated that assistance as a 4 or higher on the eapuseful scale and “0” otherwise. Because of
the nature of the survey instrument, the 60 treated entrepreneurs who did not indicate receiving
assistance but were qualitatively categorized as having received assistance would not have
responded to the eapuseful question in regards to the Product Center. In this case, however, if we
assume that the reason they did not indicate they received assistance from the Product Center
was that they did not feel the assistance was actually “assistance” because it was not useful, then
they can be categorized as a “0” for the variable pcusefullaunch.
The other controls were also included to determine if any exogenous factors could be
linked with whether the evaluation of the Product Center’s usefulness. The results of this
87

regression are presented below in table 18. In this case, it does appear that we can find some
evidence to reject the null hypothesis in regards to H5 as those respondents who have launched a
product and have more relevant industry experience (indusexp) are indeed more likely not find
the assistance useful. Not surprisingly, distance away (disttopc) is also a negative predictor of
usefulness as the treated entrepreneur will have to balance the cost of travel with the benefit of
assistance when deciding usefulness.
Table 18: Probit regression of probability to seek assistance and find it useful restricted to
respondents who launched product and had some form of contact with the Product Center
(statistically significant variables in bold N=210, Pseudo R2=0.0605)
Y=
pcusefullaunch
baormore
Disttopc
propinfoseek
Male
White
Inherit
Pushed
Indusexp
msuconnect
familyentrep
_cons

Coef.
Std. Err.
0.40
0.20
-0.004
0.00
0.06
0.05
-0.24
0.19
-0.02
0.30
-0.01
0.35
0.11
0.26
-0.01
0.01
0.13
0.19
0.37
0.20
-0.10
0.45

z
2.06
-1.79
1.26
-1.25
-0.08
-0.02
0.43
-1.69
0.69
1.88
-0.21

P>|z|
0.04
0.07
0.21
0.21
0.94
0.99
0.67
0.09
0.49
0.06
0.83

[95%
Conf.
Interval]
0.02
0.78
-0.01
0.00
-0.04
0.16
-0.62
0.14
-0.62
0.57
-0.69
0.67
-0.39
0.61
-0.03
0.00
-0.24
0.51
-0.02
0.76
-0.98
0.79

We are also seeing that having a higher education (baormore) is also a positive predictor
of perceived usefulness. This could potentially be due to the nature of this particular assistance
program operating within a university framework and hence more accessible to those familiar
with such an institution. It might also indicate that the assistance provided is more useful to those
who have higher intellectual abilities in a formal setting as could be indicated by their ability to
obtain bachelor’s degrees or higher. Whereas again we see the presence of the entrepreneur in
one’s family not only predicts the seeking of treatment, but also finding it useful. The same

88

underlying reasons as posited above are likely at play here as well. So in these areas, we can
reject the null for H5.
To deepen this analysis, another explanatory variable “prelaunchpcassist” was used
determine if the timing of assistance had any effect on perceived usefulness. The results of this
regression are present below in table 19.
Table 19: Probit regression of finding assistance useful on timing of assistance and controls
restricted to respondents who launched a product and had some form of contact with the Product
Center (statistically significant variables in bold N=210, Pseudo R2 =0.0980)
Y=
pcusefullaunch
baormore
disttopc
propinfoseek
male
white
inherit
pushed
indusexp
familyentrep
prelaunchp~t
_cons

Coef.
0.42
0.00
0.05
-0.22
-0.05
-0.05
0.06
-0.01
0.33
0.61
-0.24

Std.
Err.
0.19
0.00
0.05
0.19
0.29
0.35
0.24
0.01
0.19
0.19
0.45

z
2.23
-1.96
0.97
-1.20
-0.18
-0.14
0.25
-1.24
1.72
3.28
-0.53

P>|z|
0.03
0.05
0.33
0.23
0.86
0.89
0.80
0.22
0.09
0.00
0.60

[95%
Conf.
Interval]
0.05
0.80
-0.01
0.00
-0.05
0.15
-0.59
0.14
-0.61
0.51
-0.74
0.64
-0.41
0.53
-0.02
0.01
-0.05
0.71
0.25
0.98
-1.12
0.65

The results of this regression seem to implicate a few things. First, in regards to the
timing of the assistance, it is interesting to see that when restricting the analysis to only those
who launched new ventures, those who come in for assistance prior to the launch of their new
venture are more likely to find it useful than those who come in afterwards. This appears logical,
as much of the assistance is geared towards developing the venture concept prior to launch in
order to enhance the probability of success. Amongst the 225 treated entrepreneurs 133 (or
59.1%) have sought assistance prior to the launch of their new venture, and so this is not a small
percentage of the group. It may be other unobserved factors are at play as well, such as those
89

who are experiencing more trouble early on in their new venture creation process may have
lower entrepreneurial abilities and therefore seek assistance, and find it more useful. In fact,
when the prelaunch variable is added, we can see that the industry experience variable is no
longer significant. This is consistent with our earlier finding that those with less experience are
more likely to seek assistance prior to the launch of their product. It also shows us that, in turn,
those who seek assistance prior to the launch of their product are finding that assistance more
useful. Therefore, those with more experience and have been successful in launching a product
also find the assistance less useful giving some compelling support for H5, though this is not a
definitive result as we are only indirectly showing this.
D. Influence of firm size on perceived usefulness
We are also interested to see if respondents from smaller firms (based on gross annual
sales) are finding assistance more useful than from larger firms. To determine if this is so we will
test the following hypothesis:
Entrepreneurs from firms whose gross annual sales are less than $200,000 per year will be
more likely to find assistance useful. (H6)
We can test this hypothesis through using a probit regression on pcusefulaunch while also
including the smallfirm indicator, which indicates a “1” for firms whose gross annual sales is
under $200,000 per year and a “0” if otherwise, as well as the prelaunchpcassist variable since
we have determined that to be significant. The results are presented below in table 20.

90

Table 20: Probit regression of Usefulness of Assistance on Controls and Firm Size
(N= 247, statistically significant variables in bold)
Y=
pcusefulaunch
age
indusexp
baormore
familyentrep
smallfirm
white
male
propinfoseek
inherit
pushed
disttopc
prelaunchpcassist
_cons

Coef.
Std. Err.
-0.01
0.01
-0.01
0.01
0.29
0.24
0.24
0.25
0.87
0.43
0.34
0.15
0.10
0.24
0.06
0.06
-0.10
0.40
-0.01
0.29
0.00
0.00
0.42
0.23
-0.44
0.93

z
-1.19
-0.65
1.22
0.95
2.05
0.44
0.40
0.92
-0.25
-0.03
-0.35
1.79
-0.48

P>|z|
0.23
0.52
0.22
0.34
0.04
0.66
0.69
0.36
0.80
0.98
0.73
0.07
0.63

[95%
Conf.
Interval]
-0.03
0.01
-0.03
0.01
-0.18
0.77
-0.25
0.72
0.04
1.71
-0.52
0.82
-0.37
0.56
-0.07
0.18
-0.89
0.69
-0.57
0.56
-0.01
0.01
-0.04
0.87
-2.27
1.38

Here we can see that both the small firm and the pre-launch assist indicator variables are
significant in determining whether or not an entrepreneur finds assistance useful. This is
interesting because, as was hypothesized, we are seeing that assistance is perceived as more
useful from the firms with less resources. This is consistent with our hypothesis that assistance
acts as a substitute for the strategic resources that firms may otherwise be lacking as opposed to a
complement. That is to say, EAP assistance appears to be filling gaps more than enhancing
underlying assets. In addition, even when the size of the firm is controlled for, those seeking
assistance prior to launch are still finding it more useful than those who seek it afterwards. This
indicates that the value of assistance is perceived as more helpful when entrepreneurs engage in
the assistance process from start to finish, instead of coming afterwards to put out a fire.

91

E. Why certain entrepreneurs find assistance useful when others do not.
Finally, we wish to delve further into the second half of R3, which asks why do
entrepreneurs find assistance useful? One method we can use to answer this question is to
examine the respondents’ reported feedback on the services they received. This can help give us
a direct indication of how beneficial those services were to them. Before doing so, however, it is
important to point out that the types of entrepreneurs who come to the Product Center are very
diverse. For example, one client described her venture in the following way, “my business is a
small cottage foods business that will never be big due to the nature of the products (wild
harvested). My goal is to educate people about wild foods, good health, and get people out-ofdoors.” For this type of entrepreneur the goals and needs will be quite different from the one who
said the following, “Because of (the Product Center’s trade show) Making it in Michigan 2011
(we were) able to meet with Kroger Michigan and Meijer Corporate. We are working with
Meijer to provide a (product) for the 2013 season in their 199 bakeries.” For this reason, while
we can gain some insight through the performance measures used in this analysis, this might not
always indicate the full measure of the impact provided. Furthermore, the existing skill set of the
entrepreneur who said, “I never worked in this industry prior to starting this business - previously
a scientist employed in the automotive industry,” will be quite different from the one who tells
the following tale, “Farming has been our family business since 1854. I am the 5th generation on
our farm. My sons are the 6th generation. Upon the retirement of my father & uncle I took over
operation of our farm in 1991. In 1994 I initiated a shift to organic production--which placed me
closer to the end-user, and opened up inquiries from consumers for direct purchase of products.
My initial product was a wheat free pancake/waffle mix. I made it for personal consumption after
I discovered that I am allergic to wheat. Friends commented on how good it tasted--asked for

92

mix to take home--and a literal cottage/kitchen business was born!!” This is why it is important
to include the appropriate controls such as industry experience, presence of entrepreneurs’ in
one’s family, education and goals for the business in the quantitative analysis that will follow.
In the survey, respondents were asked if they had received assistance from the MSU
Product Center. If the respondent indicated that they had, then a follow up question was asked on
how they rated their satisfaction, the usefulness, and the influence on their decision to launch the
assistance provided. For this question, respondents were given a 7-point Likert scale where 1
indicated very dissatisfied (useless or un-influential) and 7 indicated very satisfied (useful or
influential).
i.

Reported Satisfaction with Assistance

Beginning with satisfaction, the majority of respondents (>70%) indicated they were not
dissatisfied (i.e. 4 or higher) with the most common response being a 5 out of 7, which we could
interpret as somewhat satisfied (see figure 2, below). The mean out of the 247 respondents was
3.97 or approximately neutral indicates that there appears to be a significant amount of
respondents who fall both on the satisfied and dissatisfied end of the spectrum.

93

Figure 2: Histogram of Product Center Clients’ Reported Satisfaction Levels.

45

Product Center Clients' Reported Satisifcation Levels
(N=247)

Percent of Respondents

40
35
30
25
20
15
10
5
0
1

2

3

4
5
6
Satisfaction Level
1 = Very Dissastified , 7= Very Satisfied

7

These findings are consistent with the comments of respondents where responses range
from the very positive such as one respondent said, “I really liked meeting with the MSU
Counselor and she gave very good information and some great ideas and thought provoking info
too. I must say that the initial visit was very beneficial,” to some mixed comments such as
another respondent said, “Initially the product development center was most helpful. The last
time I asked a question they were not answering the question. I tried twice and felt I was
spinning my wheels and did not want to waste their time,” to also some negative experience such
as another respondent put it, “Whenever asked, I tell people that the MSU Product Center was
not only a waste of time, it lead to unnecessary governmental interference and start-up
expenses... If the purpose for funding this MSUPC is to increase employment opportunities in

94

Michigan it has shown me no proof of offsetting the cost of faculty expense with an equal
amount of tax-paying employment.” Therefore, while it is clear that some clients walk away very
satisfied with their experience, there are also those where more could be done to address their
needs in a fulfilling manner.
ii.

Reported Usefulness of Assistance

More directly related to research question (3) is the overall perceived usefulness of the
assistance received. Once again, the majority of respondents (>75%) reported that the assistance
rated a 4 out of 7 or higher in terms of its usefulness, with the most common response being a 5
out of 7 (see figure 3, below). The mean out of the 250 respondents who answered this question
was slightly higher than the previous one at 4.16, but still squarely within the neutral zone.
Figure 3: Histogram of Product Center Clients’ Perceived Usefulness of Assistance Levels.

45

Product Center Clients' Reported Satisifcation Levels
(N=250)

Percent of Respondents

40
35
30
25
20
15
10
5
0
1

2

3
4
5
Perceived Usefulness of Assistance
1 = Very Useless , 7= Very Useful

95

6

7

Once again, comments on the usefulness of services range from the very positive to very
negative with a few in between. On the positive side of the spectrum there are many comments
that praise the assistance and access to resources provided by the Product Center, such as one
respondent put it, “I have found AgBioResearch and the Product Center to be very helpful in
many ways including research, marketing, and technical aspects for food production.” Another
respondent had a similar experience and said the following, “The MSU Product Center is a
valuable resource for small businesses. The amount of knowledge and experience available is
immeasurable.” There were still some mixed comments such as the following, “We received
some starting help but the follow through and following the business progress was in my mind
set was lacking.” There were also a number of comments on the negative side of the spectrum,
such as one respondent put it, “Ultimately, I never heard back from (the Product Center) again. I
emailed and called a few times with no response… Ultimately, I came to the decision that I
could not sustain my product launch without the outside help I was looking for. By November of
2011 I stopped all expenditures and work and by year end, I decided to fold my venture as it was
evident it would not feasible for me to move forward on my own given my limitations.”
When investigating further into the comments of respondents with low satisfaction and
perceived usefulness levels, two common themes emerge: (1) there is often a significant lag-time
and follow-through between the initial counseling session and hearing feedback from the
counselor on the next steps. This can be best summarized in the words of another respondent
who said, “I found that the biggest hurdle to overcome was having to wait for answers from the
MSU Product Center. My days were structured with lots to get done in a short amount of time in
order to launch. Unfortunately the MSU Product Center Staff were not as available as I needed. I
found myself spending most of my time getting answers to my questions by others means;” and

96

(2) counselor will give clients a list of things to do, but not provide enough assistance in getting
those things done. Again, in the words of another respondent, “they give you some ideas and
"tell" you what to do, but don't have any real connections in making things happen. In my own
experience, I need HELP...connections, meetings, networking for knowledge in my industry
(food production) NOT necessarily just financial help...I need marketing help. They tell you "oh
you need new packaging" but then there is no follow up with helping to get it done.” There also
appears to be a significant disparity in between the skill levels of different counselors. The same
respondent just quoted also went on to say, “I had an awesome mentor named (omitted) helping
to "inspire" me, but he left and the person who took his position has literally done nothing in the
past year to even keep up communication with me.” Though these comments reflect poorly on
some of the services provided by the Product Center, it is important to emphasize for every
negative comment there are just as many that say positive things such as how, as one respondent
commented, “The MSU product development center has been extremely helpful to me in starting
this business venture.” This might therefore suggest that those entrepreneurs with the least
ability to do things on their own are the ones that are most dissatisfied with the assistance. This
certainly would give some support to the hypothesis H7 which states:
Entrepreneurs who seek assistance and then decide to not launch a new venture
will have been less likely to be successful than those who have persisted with the
assistance process. (H7)
Therefore, the next area of impact we wished to assess, was how influential the Product
Center’s assistance was on respondent’s decision to launch a product or not. Unfortunately, there
are not enough observed “no-launches” in this group to get any regression model to fit properly.
Therefore, no strong conclusions can be made using this data about H7. Despite this limitation, it
may be possible to infer from the analysis of the impact of treatment on survival that will be

97

done below whether or not there is still more support for this hypothesis. Namely, if treated
entrepreneurs are enjoying higher survival rates than the non-treated ones, this could be due, in
part, to the “weeding out” effect, though this would not be the only influence of the EAP
In order to test H7 a probit regression analysis was done of the launch status of the
entrepreneurs’ venture idea (launched) on the control variables mentioned above for both the
treated and non-treated group to see if any noticeable differences could be observed. Initially
when the age variables were included the treatment group’s regression there was a high
“badness” of fit value with the reported probability > chi2 at 0.2296. However, by excluding
these variables on the basis that within this particular regression they may be highly collinear
with the industry experience variable, a reasonable fit (prob>chi2 at 0.0904) was obtained with
some interesting results as is presented below in table 21.
Table 21: Regression of launch status on controls restricted to treated group (N=248, pseudo
R2=0.0746, statistically significant variables in bold)
Variables
propinfoseek
familyentrep
baormore
white
male
indusexp
pushed
inherit
_cons

Coef.
Std. Err.
-0.08
0.06
0.36
0.19
0.22
0.19
0.42
0.24
-0.29
0.18
0.01
0.01
0.63
0.32
-0.12
0.34
0.24
0.41

z
-1.43
1.90
1.18
1.76
-1.57
1.69
1.97
-0.35
0.58

P>|z|
0.15
0.06
0.24
0.08
0.12
0.09
0.05
0.72
0.56

[95%
Conf.
Interval]
-0.19
0.03
-0.01
0.74
-0.15
0.59
-0.05
0.89
-0.65
0.07
0.00
0.03
0.00
1.25
-0.80
0.55
-0.57
1.05

For the treated group, it appears those with a history of family entrepreneurship, are
Caucasian, have higher levels of industry experience and are “pushed” into self-employment
because of no better options for work are observed to be more likely to have launched their new
venture versus the converse of those variables. As will be seen later on, both the Caucasian
98

variable and industry experience variable have been linked with higher performance. This does
appear to allow us to reject the null with respect to H7. However, in order to observe whether the
Product Center is the cause of this impact, by “weeding out” bad ideas, there must be some
evidence in the non-treatment group to compare against.
F. Implications of these Results
Broadly speaking, we found that those respondents with higher education levels and
those who sought assistance prior to the launch of their product were more likely to find
assistance useful than those with lower education levels or those who sought assistance only after
their product had already launched. In addition, we found that those who were seeking assistance
prior to the launch of their product tended to have less experience than those who sought it only
afterwards, but no other identifiable differences. This finding that the lack of industry-related
experience is an important predictor of both the decision to seek assistance prior to launch and to
find that assistance useful, appears to imply that the assistance provided acts as a substitute for
resources that entrepreneurs with less experience might otherwise not have. This was also found
to be true for those smaller firms within the sample, whereas the finding that larger firms and
more experienced entrepreneurs find assistance less useful seems to imply that the assistance
provided does not necessarily complement existing firm level resources for the more experienced
entrepreneurs or larger ventures.
We found that the majority of respondents felt they were at least neutral or satisfied with
assistance, with the most common response being somewhat satisfied, but very few were highly
satisfied and approximately a third were dissatisfied. A common reason given was that the initial
help was very beneficial but there was a lack of strong follow through in helping to drive the
clients’ venture creation or improvement process to completion. These findings were consistent
99

with clients’ responses on the perceived usefulness of assistance where some felt they could not
have launched their venture without the assistance of the Product Center and others found that
the assistance was not responsive enough for the needs of their venture and therefore relying on
it became a stumbling point. This disparity in responses also appears to be at least somewhat
connected to the particular counselor providing assistance as some respondent who had contact
with multiple counselors noted the differences in services provided. However, it may also
indicate that the assistance provided is most useful for those who have enough entrepreneurial
ability to launch their new venture on their own, but are in need of filling certain information or
resource gaps, as opposed to those entrepreneurs who rely on assistance to accomplish things
that they themselves cannot do. In this case, an EAP should be careful not to give too much
assistance as this would unnaturally support the venture when perhaps the most beneficial
decision in the long term, though painful in the short term, would be for the client to decide to
drop the business idea.
Unfortunately, we did not find strong evidence to support or reject the notion that those
who receive assistance and decide not to launch would not have been as successful had they
launched than the ones who received assistance and did. However, we did find evidence that
those with less experience in a related industry are more likely to find assistance useful, but this
relationship is moderated by whether or not the client sought assistance before the launch of their
new venture. That is to say those who seek assistance prior to the launch of their new venture are
more likely to find it useful, and in turn those with less experience are more likely to seek
assistance prior to the launch of their new venture. This gives strong support for the notion that
an EAP acts as a special type of knowledge resource provider to inexperienced entrepreneurs,
that is not as beneficial to those entrepreneurs with more experience.

100

We also found that those entrepreneurs with higher levels of education and a history of
entrepreneurship in their family were more likely to find assistance useful as well. The former
could be indicative of the institutional setting (i.e. a University) where the Product Center
provides assistance as well as the need for a certain level of cognitive capabilities to translate the
assistance provided into a tangible benefit for the venture. The latter is a rather surprising result,
though consistent with earlier findings that having an entrepreneur in the family makes one more
likely to seek assistance as well. The same conjecture provided above, that knowing first-hand
the difficulties of the entrepreneurial process is the likely cause of this result. Not only does firsthand knowledge of the difficulty in launching a new venture make one more likely to seek
assistance in doing so, it also provides for a greater appreciation of the assistance provided. It
also might help support this notion that assistance is perceived as most useful by those who can
best capitalize on it as having an entrepreneur in the family could also indicate that the
entrepreneur has access to a greater set of resources to help him or her succeed. Finally, the
distance traveled to reach the Product Center was found to be a significantly negative predictor
of finding assistance useful. This is relevant because it provides an interesting way that
researchers or EAP managers could use to value the “worth” of assistance. Namely, this finding
shows clients are clearly trading off time travelled for assistance with its perceived usefulness.
This makes sense as there is a cost to the time and money spent to travel to receive assistance,
even if the assistance itself is provided without a cost. Therefore, further researchers could look
to quantify this relationship using standard mileage expense rates in combination with the
amount of assistance received and perceived usefulness to create a valuation of that assistance.

101

VII.

IMPACT OF PC ASSISTANCE ON PERFORMANCE
Now that a clear picture of who seeks assistance and why they find it useful has been

developed, we can proceed to measure the impact that assistance has created on those who
receive it. This will be done through answering research question 1 (R1) through the testing of
hypothesis (H1) and its associated sub-hypotheses (see table 3a, page 34)
In order to test the impact of EAP assistance three different estimation methods will be used:
ordinary least squares (OLS), probit regression and Tobit (Wooldridge, 2002). A brief
explanation will be given on each of these.
A. Methods
i.

Ordinary Least Squares (OLS)

For continuous performance variables that are not truncated, Ordinary Least Squares
(OLS) will be used to measure EAP impact. To use OLS, we must make the first standard GaussMarkov assumption (see Wooldridge, 2002) and assume that our model is linear in parameters.
We can then estimate the impact of treatment on continuous performance variables such as sales
and employment through estimating the coefficients in equation (1):
(1)

ܻ௜ = ܺ௜ ߚ + ܼ௜ ߙ + ߝ௜

Where for a set of N individual entrepreneurs, the ith entrepreneur’s performance variable
{Y} is determined by a vector of explanatory variables {W} including the control variables {X}
and a treatment variable of assistance {Z}, which can be either a general or service specific
participation dummy or a continuous indicator of hours of assistance received, and an error term
{ε} that represents all other unobserved influences on {Y}. With the key estimator of interest

102

{α} represents the impact of assistance on performance to be solved by minimizing the sum of
the squared error terms in respect to the β and α coefficients.
ii.

Probit Regressions

For the discrete binary performance variables, such as survival (which is coded 1 for still
alive and 0 for no longer in operations) a probit regression will be used. In the case of survival,
for example, we can postulate there exists a latent factor Y*, which represents the viability of the
new venture to the entrepreneur, that will determine whether or not that entrepreneur decides to
keep the venture “alive” or in operations, such that if Y* > 0 the venture will still stay in
existence and if Yi* ≤ 0, the entrepreneur will decide to drop the business idea. Furthermore, let
us assume that Yi* is influenced both by observable characteristics represented by a vector {Xi}
such as the venture’s age, the entrepreneur’s experience, education and so on as well as the
impact that treatment {Zi} has had. It is likely that the viability of the new venture is also
influenced by unobservable characteristics (to us) as well such as the entrepreneur’s dedication,
interest and satisfaction with the venture that will be represented by the error term {εi}. Then we
can characterize the viability of the venture by equation (4):
Y௜∗ = ܺ௜ ߚ + ܼ௜ ߙ + ߝ௜ 	(݅ = 1,2 … , ܰ)

(4)

However, as Yi* is a latent variable we only observe its sign indicated by the survival
dummy variable {Di} where Di = 1 (for example indicating still alive) if Yi*>0 and Di =0 if Yi* ≤
0. Then we can rewrite equation (4) as follows:
ଵ:		௜௙	௑ ఉା௓ ఈାఌ 	வ	଴

	‫ܦ‬௜ = {଴:	௜௙	௑೔೔ఉା௓೔೔ఈାఌ೔೔		ஸ	଴ 	

(4a)

103

Then, using equation (4a), then we can see that the probability that an entrepreneur (i) is
still alive (i.e. Di =1) can be written as follows in equation (4b):
(4b)

ܲ‫ߝ(ܾ݋ݎ‬௜ ) 	 > −ܺ௜ ߚ−ܼ௜ ߙ	

Next, if we assume that the error term {εi} follows an independently and identically
distributed normal distribution with mean (0) and variance σε2, then the probability that εi is
larger −ܺ௜ ߚ−ܼ௜ ߙ is simply 1 – Φ(−ܺ௜ ߚ−ܼ௜ ߙ) where Φ represents the normal cumulative
density function (CDF). Furthermore, since the normal distribution is symmetrical in nature, 1 –
Φ(−ܺ௜ ߚ−ܼ௜ ߙ) is equivalent to Φ(ܺ௜ ߚ+ܼ௜ ߙ). Therefore, the probability that an entrepreneur’s
new venture is still alive, given the above assumptions, can be characterized by equation (4c):
(4c)

ܲ‫	ܦ(ܾ݋ݎ‬௜ = 1|ܺ) = ߔ(ܺ௜ ߚ + ܼ௜ ߙ)
This equation can be solved using MLE to obtain an estimate of the impact of assistance
represented by ߙො.
iii.

Tobit Regressions

For truncated variables, such as gross annual sales (which is upper-bound at $200,000) and
our proxies for perceived legitimacy (which are lower-bound at zero) a Tobit (see Wooldridge,
2002) regression will be used. This model can be characterized as in equation (2) for the upper
bound or right-hand censored regressions and equation (3) for the lower bound or left-hand
censored regressions:
		ை௕௦௘௥௩௘ௗ	ெ௔௫௜௠௨௠																																				௜௙	௑ ఉାఢ வ௠௔௫	

(2)

	௑ ఉା௓ ఈାఢ 																																																							௜௙	௑ ఉାఢ வ଴	

(3)

ܻ௜ = {	௑೔ఉା௓೔ ఈାఢ೔ 																																																							௜௙	௑೔೔ఉାఢ೔೔ஸ௠௔௫	 		߳௜ ~݊(0, ߪ ଶ )
೔
೔
೔
೔
೔
ܻ௜ = {	଴																																																																														௜௙	௑
								߳௜ ~݊(0, ߪ ଶ )
೔ ఉାఢ೔ ஸ଴	

104

Where again for a set of N individual entrepreneurs, the ith entrepreneur’s performance
variable {Y} is determined by a vector of explanatory variables {W} including the control
variables {X} and a treatment variable of assistance {Z}, which can be either a general or service
specific participation dummy or a continuous indicator of hours of assistance received, and an
error term {ε} that represents all other unobserved influences on {Y}. However, in this case
instead of solving the equation by minimizing the sum of squared errors, a maximum-likelihood
estimation (MLE) will be used that models the uncensored variables as a probability density
function and the censored variables as a cumulative density function within the sample at that
point. This method is less efficient that OLS, but necessary for the cases, such as with gross
annual sales, the observed value of $200,000 indicates the entrepreneur’s new venture is making
more than $200,000 (not $200,000 per se), so a censored regression technique should be used.
B. Impact of EAP assistance on new ventures’ success and survival
Now that the methodology to be used has been thoroughly developed as to how we will
assess the impact of EAP assistance, the next step is begin quantifying the impact. This will be
done through testing the main hypothesis of this paper:
Entrepreneurs who receive assistance from an EAP will be more likely to have higher
performance, survival rates and legitimacy than had they not sought assistance. (H1)
As it turns out, this hypothesis is rather broad and vague and for full elucidation of the
impact that the case study’s EAP has had on its entrepreneurial clientele it becomes prudent to
break this hypothesis down into the many sub-hypotheses listed above in table 3a. Therefore, we
will begin by first testing the impact that assistance has on the decision to launch, followed by

105

the impact on survival after launch, then the ability of the EAP to build legitimacy with resource
providers and marketing outlets, followed by impact on sales and employment and then finally
differentiating these impacts for the smaller firms grossing under $200,000 per year versus the
larger sample.

C. Impact of assistance on the decision to launch
We saw in the previous chapter that those entrepreneurs who were seeking assistance
prior to the launch of their new venture were more likely to have less experience and in turn
more likely to find assistance useful. The natural question that arises from this result is why this
would be so. To this end we have hypothesized the following:
Entrepreneurs who receive assistance from the example EAP prior to the launch of a new
venture will be less likely to launch that new venture than those who do not (weeding out
hypothesis). (H1-a)
When one compares respondents who have received assistance with those who have not,
the percentage of those respondents who went on to launch a product is smaller in the treatment
group than in the non-treated group (see figure 4, below). Whether this is a positive or negative
effect, however, has yet to be seen. If treatment is successful in preventing the launch of
untenable business ideas than this would be a positive effect that would be borne out in higher
survival rates amongst those that do launch. If, however, the lower launch rates are due to poor
quality of assistance or selection bias issue, then we would not expect to find any positive
impacts of treatment on survival.

106

Figure 4: Comparison of launch status between Product Center clients and non-clients
Comparison of Launch Status by Treatment
120%
100%
80%
60%

Did not launch

40%

Launched

20%
0%
Product Center Clients
(N=310)

Non-Product Center Clients
(N=154)

Further investigation into the reasons that respondents did not launch a product reveals a
few noticeable differences between the treated group and the non-treated group (see figure 5,
below). For both sets of entrepreneurs, the most common reason they decided not to launch the
product was not enough capital (46% and 45% for the treated and non-treated respectively).
When compared to the treated entrepreneurs, non-treated entrepreneurs were more likely to
claim not enough demand (7% to 4%) and no market (7% to 2%) or to have pursued other
interests (21% to 17%) than the treated entrepreneurs, whereas treated entrepreneurs were more
likely to claim they did not enough time (25% to 17%), couldn’t find a co-packer (8% to 3%)
and to still be in progress (17% to 13%). Some of the comments from respondents such as the
following, “the Product Center did help us determine that we could not take a product idea to
market, so it was not pursued further,” however, do clearly indicate that the Product Center
helped to “weed out” an untenable idea.

107

Figure 5: Comparison between treated and non-treated entrepreneurs on the reasons why those
that did not launch a product gave for not launching.

Reasons Why Respondents did not Launch Product
50%
45%
40%
35%
30%
25%
20%

Product Center Clients
(N=84)

15%

Non-Product Center
Clients (N=29)

10%
5%
0%

i.

Testing the “weeding out” component of assistance

To test H1-a a new binary indicator variable “prepcassist” was created that identified
whether the respondent sought assistance at point a1 in our conceptual framework as a “1” and
those who did not (including those entrepreneurs who may have sought assistance later on) were
coded as a “0”. This was done because if the entrepreneur sought assistance after the launch of
their product, then it would not make sense to test whether assistance had an impact on the
decision to launch. The variable, prepcassist, was included in a probit regression of launch status
on controls to test this hypothesis with the results presented below in table 22.

108

Table 22: Probit of launch status on treatment (at decision node a1) and controls (statistically
significant variables in bold N=346, pseudo R2 = 0.1197.)
Variables
Coef.
Std. Err. z
prepcassist
-0.77
0.16
propinfoseek
0.05
-0.07
proprisktake
0.08
0.05
propinnovate
-0.13
0.05
propgrowbiz
0.03
0.04
age
0.00
0.01
indusexp
0.00
0.01
baormore
0.19
0.17
male
-0.14
0.16
white
0.29
0.23
familyentrep
0.11
0.17
inherit
0.02
0.32
pushed
0.27
0.24
_cons
1.13
0.60

-4.72
-1.50
1.46
-2.68
0.63
-0.06
0.58
1.18
-0.85
1.26
0.66
0.06
1.12
1.91

P>|z|
[95% Conf. Interval]
0.00
-1.09
-0.45
0.13
-0.17
0.02
0.15
-0.03
0.19
0.01
-0.22
-0.03
0.53
-0.06
0.11
0.95
-0.02
0.01
0.56
-0.01
0.02
0.24
-0.13
0.52
0.40
-0.46
0.18
0.21
-0.16
0.75
0.51
-0.22
0.44
0.95
-0.61
0.65
0.26
-0.20
0.74
0.06
-0.03
2.30

As can be seen from this regression, receiving assistance prior to launch was very
strongly correlated with the decision to not launch, thus giving support for the first element of
the “weeding out” hypothesis and allowing us to reject the null in regards to H1-a.
ii.

Testing the “planting in” component of assistance

It still remains to be seen whether those who do launch and receive assistance are more
likely to survive than those who do not. That is to say, if the weeding out we observed in the
previous section truly was a beneficial weeding mechanism, then we would expect to find
evidence to reject the null of the following hypothesis:
Entrepreneurs who receive assistance from the example EAP prior to the launch of a new
venture and decide to launch that new venture will be more likely to stay in business than those
who launched a new venture but did not receive assistance (planting in hypothesis).(H1-b)
109

To test this hypothesis, the survival rates of entrepreneurs who received assistance were
compared to those who did not. From a purely comparison of means test, it does appear that the
treated group is enjoying higher survival rates than the non-treated group (see figure 6 below).
This indicates support for our previous hypothesis that treatment is helping to prevent untenable
business ideas from going forward. The survival rates, however, for both the treated and nontreated entrepreneur respondents appear to be higher than would be expected (i.e. from Shane’s
2008 research on the low expected survival rates of new ventures). This probably indicates a bit
of response bias in this survey. This is likely due to the lack of interest in responding to surveys
about a business idea that has already been dropped. Nonetheless, assuming this response bias is
not skewed to either the treatment or non-treatment group then it should not impact the
comparison across these two groups. Given this assumption, then the responses indicate that
treated entrepreneurs (with approximately 95% of new ventures still in operation) are enjoying
higher survival rates than their non-treated counterparts (with 89% of new ventures still in
operation) (see figure 6, below).
Figure 6: Survival Rates between the Treated and Non-Treated Groups

Comparison of Survival Rates between Treated and
Non-Treated Groups
105%
100%
95%
Dropped Business Idea

90%

Still Alive

85%
80%
Product Center Clients
(N=219)

Non-Product Center
Clients (N=122)

110

Therefore, it is not surprising that when one conducts a probit regression on the
probability of a new venture being still in operation on the key treatment indicator variable,
while controlling for the venture’s age and other relevant factors, having product center’s
assistance appears to be a significantly positive predictor of survival (see table 23, below).
Table 23: Probit regression of the probability of the new venture being still in operation on
treatment and control variables (statistically significant variables in bold, N=256 R2 =0.23)
Y=stillalive
ventureage
pcassist
propinfoseek
proprisktake
propinnovate
propgrowbiz
indusexp
age
baormore
male
white
familyentrep
inherit
pushed
_cons

Coef.
Std. Err.
-0.13
0.06
0.57
0.31
0.03
0.07
-0.15
0.10
0.00
0.07
0.17
0.07
-0.02
0.01
0.02
0.01
-0.06
0.24
0.74
0.32
0.71
0.41
0.48
0.28
-1.00
0.44
-0.04
0.32
-0.13
0.82

z
-2.26
1.83
0.50
-1.45
-0.01
2.32
-1.99
1.74
-0.25
2.35
1.73
1.72
-2.28
-0.11
-0.16

P>|z|
0.02
0.07
0.61
0.15
0.99
0.02
0.05
0.08
0.80
0.02
0.08
0.09
0.02
0.91
0.88

[95%
Conf.
Interval]
-0.25
-0.02
-0.04
1.18
-0.10
0.17
-0.36
0.05
-0.14
0.14
0.03
0.32
-0.05
0.00
0.00
0.05
-0.53
0.41
0.12
1.37
-0.09
1.52
-0.07
1.03
-1.86
-0.14
-0.67
0.60
-1.74
1.48

The results of this regression provide a lot of very interesting insights. To begin with it
appears we can reject the null that assistance has no impact on survival, with a 90% confidence
level. As can be seen from the assistance indicator variable used, we did not distinguish at this
point between those who received assistance prior to launch or not. The reasoning for that
distinction before was to directly test the influence of assistance on the decision to launch. In this
case, we wish to see if assistance, regardless of when received, influences the new venture’s
ability to survive, which it appears it does. Therefore, in combination with the rejection of the

111

null that assistance has no impact on the decision to launch a product, we can see that assistance
is effective at “weeding out” the untenable venture ideas as well at nurturing the ones “planted
in”. Therefore, it appears we have provided sufficient evidence to reject then null in regards to
H1-b.
In addition to finding assistance a significant positive indicator of survival, we also find
that the venture’s age is a significant negative predictor of survival, as one might expect, that
being Caucasian and a male are significant positive predictors, which is consistent with the other
findings in this analysis with positive relationships being associated with Caucasians. In addition,
the entrepreneur’s underlying propensity to grow the business is also shown to be significantly
positively correlated with survival, giving further support to the notion mentioned in the choice
of performance variables that growth is correlated with survival.
These past results are all what one would expect, but there are some unexpected results
uncovered in this analysis as well. The first is that while age is shown to be positively correlated
with survival, perhaps indicating that greater life experience is correlated the likelihood of
survival, there is also a significantly negative relationship found with the entrepreneur’s industry
experience. Since age is being held constant, this means there is an interesting relationship
occurring with the tradeoff between these two variables likely at play. As was seen in the cluster
analysis, there are distinct clusters based on age and experience, so this may be picking up some
of this variation. In addition, it is also surprising that having received an inheritance is negatively
correlated with survival. But as was seen in the summary statistics, the percentage of respondents
who received inheritances was quite low, so it could be just random variation that is causing this
relationship to occur.

112

D. EAP Impact on Legitimacy with Resource Holders
The next area of potential impact this paper will look at is the impact that EAP assistance
has on the perceived legitimacy with resource holders such as banks and private equity sources.
The reason behind conducting this test is the notion that through the counseling process on the
business idea, the assistance in writing a formal business plan, and through the leverage of the
EAP’s direct ties with resource holders, EAP clients will be able to build a higher level of
perceived legitimacy with the holders of financial resources and thereby have greater success in
obtaining external financing for their new venture. Therefore, we will test the following
hypothesis against its null:
Entrepreneurs who receive assistance from the example EAP will be more likely to obtain
external financing than had they not received assistance. (H1-c)
In order to first test this a left- and right-hand censored Tobit model will be run on the
level of investment received from banks and then from private equity sources. This model is
censored on both sides as zero investment indicates less than the required threshold level of
legitimacy required to obtain such financing, and on the right hand side as the survey instrument
allowed for a maximum of $200,000 for a response, so those values need to be treated as an
upper limit, not as $200,000. The results of conducting this regression on the entire sample is
presented below in table 24.

113

Table 24: Left- and Right-hand censored Tobit of external financing level on controls and
treatment. (statistically significant variables in bold, N=251, Pseudo R2=0.0253)
Y =invbank
Coef.
Std. Err. t
P>|t|
[95% Conf. Interval]
ventureage
15,689
6,589
2.38
0.02
2,709
28,670
propinfoseek
-4,087
8,658
-0.47
0.64
-21,143
12,969
propgrowbiz
5,852
7,908
0.74
0.46
-9,726
21,430
age
-472
1,435
-0.33
0.74
-3,298
2,355
indusexp
3,143
1,376
2.28
0.02
431
5,855
baormore
-6,302
31,510
-0.20
0.84
-68,377
55,773
inherit
-19,790
57,764
-0.34
0.73
-133,586
94,007
white
72,172
1.33
0.19
-46,193 238,168
95,987
male
50,821
30,762
1.65
0.10
-9,780 111,423
pushed
-44,328
42,181
-1.05
0.29
-127,425
38,769
familyentrep
10,102
32,271
0.31
0.76
-53,473
73,677
propinnovate
-18,906
9,126
-2.07
0.04
-36,883
-928
proprisktake
13,431
10,824
1.24
0.22
-7,894
34,755
pcassist
3,159
31,852
0.10
0.92
-59,590
65,908
_cons
-336,361
133,853
-2.51
0.01
-600,054
-72,668
Obs. summary:
212 left-censored observations at invbank<=0
32 uncensored bservations
7 right-censored observations at invbank>=200000
As can be seen, the majority of respondents (84.3%) are below the threshold and
receiving no bank financing. For those that do, industry experience, age of the venture and
gender and the propensity to innovate appear to be the only significant predictors of financing,
whereas treatment is not significant. Interestingly, where we saw the propensity towards
innovation was positively correlated with employment but not with sales or survival, it is
negatively correlated with this measure of legitimacy. This could indicate that banks, especially
in the recent strict credit market following the recession caused by the 2008 collapse of the
housing market, are not willing to take risks on the more non-traditional products out there. This
seems logical as the risk is much higher for the more innovative but unproven products, despite
having more potential for creative destruction.

114

In this case, the results are no different when restricting to the small and medium sized
enterprises in the sample. In the case of those receiving private equity financing, none of the
variables were found to be significant at the 10% level.
Given the apparent difficulty in achieving external financing, perhaps it is better just to
simply analyze whether the Product Center is helping clients to cross that the legitimacy
threshold at all. This can be done by conducting a probit regression on the binary “externfinance”
variable which simply indicates if the respondent received any financing from bank or private
equity sources. When this is done on the entire sample, we see that assistance does not appear to
be a significant predictor of receiving financing, whereas the industry experience, age of the
venture, the propensity for risk-taking and age variables all do as is presented below in table 25.
Table 25: Probit regression of receiving External Financing on Treatment and Controls
(statistically significant variables in bold, N=233, Pseudo R2= 0.0765)
Y=externfinance
ventureage
propinfoseek
propgrowbiz
age
indusexp
baormore
inherit
white
male
pushed
familyentrep
propinnovate
proprisktake
pcassist

Coef.
0.08
-0.06
0.01
-0.02
0.02
0.01
0.38
0.53
0.11
-0.30
-0.09
-0.08
0.13
0.13

Std. Err.
0.04
0.05
0.05
0.01
0.01
0.20
0.33
0.36
0.19
0.26
0.20
0.06
0.07
0.20

z

P>|z|
1.88
-1.09
0.26
-2.19
2.38
0.08
1.14
1.49
0.60
-1.14
-0.48
-1.35
1.80
0.66

0.06
0.28
0.80
0.03
0.02
0.94
0.25
0.14
0.55
0.25
0.63
0.18
0.07
0.51

[95%
Conf.

Interval]
0.00
-0.17
-0.09
-0.04
0.00
-0.37
-0.27
-0.17
-0.26
-0.80
-0.48
-0.19
-0.01
-0.26

0.15
0.05
0.11
0.00
0.04
0.40
1.03
1.23
0.48
0.21
0.29
0.04
0.27
0.53

It is not surprising that more experienced entrepreneurs and older ventures are more
likely to receive financing. The negative impact of age and positive impact of propensity towards
115

risk-taking are a bit more surprising however. Given the previous discussion about innovations,
one might have thought the same could be said for risk-takingness. However, as this variable is
not directly observed by the lender, it could be that those willing to take the right risks are being
rewarded by lenders who agree that those risks are worth taking. The impact of age is less clear,
but it could indicate that lenders are hesitant to finance ventures for the older class of
entrepreneurs given that they may decide to retire sooner than younger ones.
These results, therefore, do not allow us to reject the null in regards to H1-c when we
consider the entire sample of entrepreneurs who responded to our survey. However, as will be
seen later, different results will be obtained when we restrict the sample to those ventures
grossing under $200,000 per year.
E. EAP Impact on Legitimacy with Trading Partners
Next, we wish to assess the impact that treatment has had on the perceived legitimacy of
the new ventures within the marketing channels, as measured by the volume sold through retail
and wholesale outlets through the testing of the following hypothesis:
Clients from the example EAP will have access to a greater marketing opportunity set
than they otherwise would have without assistance (H1-d)
For this measure, a left-hand censored Tobit is proposed as roughly half of the observed
respondents have a volume of sales sold through retail that is zero. The reason this is taken to be
a censored observation is that if this measure is a proxy for legitimacy is based on the notion that
in order to gain access to retail outlets the new venture must have met a certain threshold level of
legitimacy with the retail provider. After gaining access, to maintain or increase the volume sold
through this outlet, the new venture must then maintain or increase their level of perceived
legitimacy with the end consumers who frequent those retail stores. Those who have zero sales
116

through retail can be considered to be below this threshold, though not necessarily the same level
of legitimacy as the other respondents who also have zero sales through retail. Therefore, the
zero observations should be categorized as a censored observation of legitimacy and modeled as
cumulative density function, whereas for those above the threshold it is logical to associate
increased sales level with incrementally increasing levels of legitimacy and should be modeled
as such, hence the appropriateness of a left-hand censored Tobit model as was described in the
methods section.
In addition, because this measure is a compilation of two other measures (gross annual
sales x % sold through retail), one of which has an upper bound of $200,000, we must restrict the
sample to only those ventures grossing under $200,000 per year. This is because for those
grossing over $200,000 their response enters in as $200,000 which is then multiplied by the %
sold through retail giving an unrealistic expectation of their total sales volume sold through this
channel. For example, consider two firms: one grossing $1 million per year and selling 10%
through retail outlets and another grossing $200,000 per year and selling 50% through retail.
These are two very different firms in size and scope, but would have the same measure of
legitimacy as we have structured it here. This is an obvious flaw in experimental design that can
be improved upon for future research, but for this case can be eliminated through the restriction
of the sample size to the proposed amount and keeping this restriction in mind when interpreting
results.
The result of this model for the respondents grossing under $200,000 per annum is
shown below in table 26.

117

Table 26: Tobit on volume of sales sold through a retail outlet on treatment and controls for
respondents grossing under $200,000 per annum (statistically significant variables in bold,
N=192, Pseudo R2 = 0.0119)
Y=volretail
ventureage
propinfoseek
propgrowbiz
age
indusexp
baormore
inherit
white
male
pushed
familyentrep
propinnovate
proprisktake
pcassist
_cons
Obs. summary:

Coef.
Std. Err.
t
P>|t|
1,390
735
1.89
0.06
956
0.28
0.78
271
114
914
0.13
0.90
-214
156
-1.37
0.17
-12
156
-0.08
0.94
3,520
-1.89
0.06
-6,664
-1,102
6,308
-0.17
0.86
8,806
5,816
1.51
0.13
4,519
3,303
1.37
0.17
-5,468
4,443
-1.23
0.22
-1,045
3,596
-0.29
0.77
1,240
1,048
1.18
0.24
11
1,185
0.01
0.99
12,939
3,817
3.39
0.00
-18,205
12,781
-1.42
0.16
95 left-censored observations at volretail<=0
97 uncensored observations
0 right-censored observations

[95%
Conf.
Interval]
-60
2,840
-1,616
2,158
-1,690
1,918
-521
94
-320
296
-13,610
281
-13,550
11,346
-2,671
20,284
-1,998
11,037
-14,235
3,299
-8,140
6,051
-827
3,307
-2,328
2,350
5,406
20,471
-43,426
7,016

In this case, it appears that there is a strong indication that treatment is helping to increase
respondent’s ability to sell their product through retail outlets, and by proxy improve their
overall legitimacy within that channel within this restricted sample of ventures. Therefore,
though it appears we cannot reject the null on that assistance has no impact on improving the
growth of total gross annual sales, it does appear that assistance is helping to improve small and
medium sized enterprises (defined in this case as having under $200,000 gross annual sales) to
increase their perceived legitimacy with retailers and end consumers. Therefore we can reject the
null hypothesis in regards to H1-d for this group.

118

This finding also supports the results of the author’s master’s thesis that found:
“The Product Center helps entrepreneurs to legitimize their product through the
adherence to social norms, rules and regulations… It can help assist entrepreneurs
in both figuring out what the steps are and by connecting them to people who can
help them complete those steps… It has developed direct ties to on and off
campus service providers for each step… It can provide advice how to market
their (clients’) products at a level that is scale appropriate for them… and it can
create opportunities by using the University’s name, reputation and relationships
with different partners such as major grocery retail chains or local hospitals to try
to get the appropriate clients into their operations.” (Lovgren, 2012).
F. EAP Impact on Gross Annual Sales
The next area of interest to investigate is the impact that the Product Center has had on the
size of the venture evidenced by the venture’s gross annual sales level. This is one of the areas
that we would hope to find a significant impact of the business planning and legitimacy building
activities that have been discussed above. Namely, we want to know has assistance impacted
overall sales growth of the new ventures, which will be answered through the testing of the
following hypothesis against its null:
Clients from the example EAP will have higher gross annual sales rates than they otherwise
would have without assistance. (H1-e)
As a reminder, gross annual sales was collected as a continuous variable from $0-$200,000
and then at the upper limit of $200,000 indicates that sales were above this amount. The impact
of assistance in this section will therefore consider both the full sample for all those who
provided sales information (as a right-hand censored Tobit regression) as well as the impact on
those new ventures grossing under $200,000 annually (as an OLS regression). The majority of
respondents for both the treated and non-treated entrepreneurs were grossing less than $12,500

119

annually with percentages that taper off as sales go up, but 12.3% of respondents have ventures
grossing over $200,000 per year.
We can see that if there is selection bias occurring, it is likely manifesting itself in the
larger than $200,000 per year group which is nearly double, in percentage terms, in the nontreated group(16.7%) than in the treated group (8.7%). However, in order to determine if this
difference is significant when other controls, including the age of the venture (which allows this
measure to represent sales growth), a right-hand censored Tobit must be conducted. The results
of this regression are presented below in table 27.
Table 27: Right-hand censored Tobit of Gross Annual Sales on treatment and controls
(statistically significant variables in bold, N=219, Pseudo R2=0.0106)
Std.
[95%
Coef.
Err.
t
P>|t|
Conf.
Interval]
ventureage
4,258
2,172
1.96
0.05
-25
8,541
pcassist
-16,157
10,354
-1.56
0.12
-36,571
4,257
propinfoseek
-1,856
2,553
-0.73
0.47
-6,889
3,176
proprisktake
2,655
3,370
0.79
0.43
-3,989
9,300
propinnovate
610
3,161
0.19
0.85
-5,622
6,842
propgrowbiz
9,864
2,541
3.88
0.00
4,853
14,875
indusexp
1,222
491
2.49
0.01
254
2,190
age
411
-1.89
0.06
-1,585
35
-775
baormore
8,143
10,140
0.80
0.42
-11,849
28,135
male
32,930
9,238
3.56
0.00
14,716
51,143
white
31,641
10,402
3.04
0.00
11,132
52,150
familyentrep
-16,242
10,182
-1.60
0.11
-36,317
3,834
inherit
-10,405
17,109
-0.61
0.54
-44,138
23,328
pushed
-8,421
11,463
-0.73
0.46
-31,021
14,180
_cons
-9,272
31,713
-0.29
0.77
-71,797
53,253
Obs. summary:
0 left-censored obs
192 uncensored obs
27 right-censored obs grosssales>=200000
Y=grosssales

120

In this case, what we can see is that, though the coefficient is negative, the treatment
variable does not appear to be a significant influence on the level of sales ceteris paribus.
Instead, the food processor indicator variable, the age of the venture, the propensity to grow the
business, the industry experience, age, gender and ethnicity of the entrepreneur appear to be the
variables that can be said to significantly impact the level of sales. None of these findings, except
perhaps the food processor indicator, are particularly surprising, as all are consistent with the
expectations laid out in variables section as reasons for their inclusions as controls. In the case of
the food processors, there was no a priori expectation either positively or negatively about how
this would influence sales, but was included to control for the different industry effects.
However, it is a bit surprising that despite the larger percentage of ventures grossing over
$200,000 in the non-treatment group, the treatment variable is not significant. It is at this point,
that the correction for selection bias would be most useful. This would require a valid IV. As was
seen above (in section VII.A), however, no potential candidates were found. This leads us to be
unable to reject the null for H1-e, though results remain largely inconclusive and require further
investigation on how to remove the downward selection bias impact.
G. EAP Impact on Employment
The next area of impact on performance to assess is employment levels and growth. For
similar reasons as with sales growth, we might hope to find that assistance has been effective in
helping new ventures to create more jobs than they otherwise would have. To determine if this is
the case, we will test the following hypothesis against its null:
Clients from the example EAP will higher employment rates than they otherwise would
have without assistance (H1-f)

121

To begin, a comparison between the treated and non-treated entrepreneurs reveals that on
average, the non-treated group has higher levels of employment (8.69 to 7.26) but the relative
distribution between these two groups is fairly similar with the majority of respondents having 5
or fewer employees
In this case there were relatively fewer outliers, however, with only six respondents who
had employment levels above 30 (3 for each group), of those six the range was from 63-250(the
maximum response allowed) with an average of 186.16. With only 2 points hitting the maximum
of 250, this regression was done as an OLS to improve its efficiency. The results are presented
below in table 28.
Table 28: OLS regression Total Employment on Treatment and Controls (statistically significant
variables in bold, N=227, R2=0.0923)
Y=totemp
ventureage
propinfoseek
propgrowbiz
age
indusexp
baormore
inherit
white
male
pushed
familyentrep
propinnovate
proprisktake
pcassist
_cons

Coef.
-0.12
0.93
2.50
0.03
0.34
5.63
-7.84
5.68
7.28
3.25
-2.22
2.16
-1.84
-4.94
-19.48

Std. Err.
0.85
1.15
1.07
0.18
0.18
4.14
7.21
6.52
3.92
5.36
4.13
1.17
1.42
4.22
14.24

t
-0.14
0.81
2.33
0.18
1.87
1.36
-1.09
0.87
1.86
0.61
-0.54
1.85
-1.3
-1.17
-1.37

P>|t|
0.888
0.42
0.021
0.86
0.062
0.175
0.278
0.385
0.065
0.545
0.592
0.065
0.196
0.244
0.173

[95%
Conf.
-1.80
-1.34
0.39
-0.33
-0.02
-2.53
-22.06
-7.18
-0.44
-7.32
-10.36
-0.14
-4.64
-13.26
-47.54

Interval]
1.56
3.20
4.62
0.39
0.70
13.78
6.37
18.53
15.00
13.82
5.93
4.47
0.96
3.38
8.58

In this case, we do not find any significant correlation (positive or negative) of assistance
on the employment levels of the business. Instead, experience in the industry, gender, the
122

propensity to grow the business and the propensity towards innovation appear to be the primary
drivers of employment. The last three of these is most interesting as it suggests that those who
are more inclined to created new and different products are experiencing growth in terms of total
employment. This variable, however, was not significant in terms of sales, so it may be that
investing in innovation requires more manpower in order to do so.
In addition, whereas the age of the venture was significantly correlated with sales level,
there does not appear to be such a relationship between age of the venture and employment. This
is not surprising given the sample’s population, as we have seen that many are small in terms of
total sales and are primarily owner operated, and the inclusion of the large firms with high
employment numbers on both sides of the treatment variable is likely obscuring the marginal
gains brought about by the smaller firms.
In terms of the unrestricted population, therefore, we cannot reject the null hypothesis
that assistance has no measurable impact on sales or employment. As will be seen later on,
however, when one drills down into different subcategories of entrepreneurs the impact of
assistance becomes more evident.
H. Impact of the timing of assistance on performance
The next logical question then becomes, if those receiving assistance prior to launch are finding
it more useful, then are they also receiving greater gains from that assistance? To answer this
question the following hypothesis will be tested against its null:
Clients who receive assistance prior to the launch of their new venture will receive
greater impact from assistance than those who receive it only afterwards. (H1-g)

123

As it turns out, there is some weak evidence that this may be occurring. By restricting
the sample to only those respondents who have received assistance (i.e. pcassist=1), and
conducting a right-hand censored tobit on the level of gross annual sales, as was done earlier, and
including the prelaunchpcassist variable as a regressor, we see that there is some weakly
significant (i.e. just below the 90% level) positive indication that those receiving assistance prior
to the launch of their product and having greater gains in sales performance. This is shown below
in table 29.
Table 29: Right-hand censored Tobit on Gross Annual Sales restricted to only those respondents
who have received assistance. (statistically significant variables in bold, N=130, Pseudo
R2=0.0091)
Std.
[95%
Coef.
t
P>|t|
Interval]
Y=grosssales
Err.
Conf.
prelaunchassist
18,457
11,254
1.64
0.104
-3,835
40,748
propgrowbiz
4,637
3,022
1.53
0.128
-1,349
10,622
propinnovate
-4,610
3,405
-1.35
0.178
-11,355
2,135
propcompaggresive
-411
3,461
-0.12
0.906
-7,266
6,444
proprisktake
5,994
4,087
1.47
0.145
-2,102
14,090
propinfoseek
-2,376
3,052
-0.78
0.438
-8,422
3,670
age
-823
534
-1.54
0.126
-1,880
233
indusexp
1,039
488
2.13
0.035
73
2,005
baormore
6,410
11,892
0.54
0.591
-17,145
29,966
male
21,771
10,780
2.02
0.046
418
43,124
white
24,745
16,827
1.47
0.144
-8,586
58,076
familyentrep
-6,306
12,098
-0.52
0.603
-30,270
17,659
inherit
-10,528
18,534
-0.57
0.571
-47,241
26,185
pushed
-15,087
14,384
-1.05
0.296
-43,579
13,406
ventureage
7,764
2,574
3.02
0.003
2,666
12,862
_cons
42,093
-0.21
0.838
-92,018
74,738
-8,640
Obs. summary:
0 left-censored observations
119 uncensored observations
11 right-censored observations at grosssales>=200000

124

In addition, we are also seeing that industry-related experience, being a male and the age
of the venture are also significant factors (as is consistent with our earlier findings).
While no significant differences were found using the same technique in regards to total
employment, survival rates and volume in retail, there was strong evidence that receiving
assistance prior to launch was significant in helping clients to obtain external financing as is
shown below in table 30. This particular finding is extremely interesting as it provides perhaps
the strongest support that the venture development counseling and services are indeed achieving
their goal helping to weed out bad ideas and plant in good ones.

Table 30: Probit regression of the ability to obtain external financing on controls and prelaunch
assistance restricted to Product Center clients who have launched a product. (statistically
significant variables in bold, N=142, Pseudot R2=0.1559)

Y=externfinance
prelaunchassist
propgrowbiz
propinnovate
propcompaggresive
proprisktake
propinfoseek
age
indusexp
baormore
male
white
familyentrep
inherit
pushed
ventureage
_cons

Coef.
Std. Err.
0.60
0.28
-0.08
0.07
-0.17
0.08
0.23
0.09
0.12
0.10
-0.05
0.07
-0.01
0.01
0.02
0.01
-0.19
0.27
-0.06
0.26
0.73
0.49
-0.27
0.27
0.61
0.42
-0.49
0.34
0.07
0.06
-1.20
1.06

z
2.14
-1.14
-2.02
2.63
1.19
-0.67
-1.18
1.99
-0.71
-0.22
1.50
-1.00
1.46
-1.45
1.24
-1.13

P>|z|
0.03
0.26
0.04
0.01
0.23
0.50
0.24
0.05
0.48
0.83
0.13
0.32
0.14
0.15
0.21
0.26

[95%
Conf.
Interval]
0.05
1.16
-0.22
0.06
-0.33
0.00
0.06
0.39
-0.08
0.32
-0.20
0.10
-0.04
0.01
0.00
0.05
-0.72
0.34
-0.56
0.44
-0.22
1.69
-0.81
0.26
-0.21
1.43
-1.16
0.17
-0.04
0.19
-3.28
0.89

Another interesting finding that comes from this regression is the significance of two of
the entrepreneurial orientation variables. First, it appears investors are not showing interest in
125

respondents who consider their products more innovative and different from their competitors,
but are favoring those who seek out more competitive “undo-the completion” posture. So this
seems to indicate that while investors might not be interested in financing unproven products,
even in less competitive markets like food processing they still wish to see some competitive
business instincts.
I. EAP impact on small ventures (grossing under $200,000 per year)
Lastly, we are also interested to see, given the focus the Product Center has traditionally had
on small firms, if the impact of assistance is different when restricting our sample to only those
firms who are grossing under $200,000 per year. This distinction is of interest because we have
hypothesized that the value that an EAP can create for its clients will provide the most
incremental benefit to those less experience and/or fewer resources available to them. Though
not categorically so, smaller sized ventures will often fall into this category as: (1) they have less
revenue, by definition, to hire and employ such resources; and (2) firms grossing over $200,000
will have already accomplished much of what EAP has to offer. Therefore, we will test
hypothesis H1-h, which states:
When restricting the sample to entrepreneurs whose new venture grosses under $200,000 per
year, the impact of the example EAP assistance will be more likely to be significantly positive
than when compared to the full sample. (H1-h)

However, before this is done, it will be useful to give a characterization of how these firms
differ compared to the larger firms in our sample. To do this, an indicator variable, smallfirm,
was created that denoted a “1” if a respondent’s venture grossed under $200,000 last year, and
“0” if they grossed $200,000 or more. Then a probit regression was conducted on smallfirm to
indicate differences in the control variables.
126

As the small firm variable is created from a performance variable, this regression was
restricted to only those respondents who had launched their product and thus reported sales
numbers. This regression indicates that those grossing under $200,000 per year are less likely to
be interested in growing their business as much as possible, not surprisingly, nor are they white
males (note that the white indicator was omitted due to collinearity, indicating that after
controlling for all other variables there is no variation in those grossing over $200,000, i.e. they
are all white males). They also tend to be older and have less industry-related experience (which
will turn out to be key drivers of the different clusters identified later on). They also are more
likely to have a history of family entrepreneurship, which we have shown is highly correlated
with seeking assistance. Otherwise there is no significant difference in the entrepreneurial
orientation variables, whether a food processor or not, their education status, whether they’ve
received an inheritance, the age of their business or their reasons for getting into selfemployment. This can be seen below in table 31:

127

Table 31: Probit regression of firm size on control variables (statistically significant
variables in bold, N=187, Pseudo R2=0.2862)
Y=smallfirm
propgrowbiz
propinnovate
propcompaggresive
proprisktake
propinfoseek
age
indusexp
baormore
male
white
familyentrep
foodproc
inherit
pushed
ventureage
_cons

Coef.
-0.26
-0.04
0.04
-0.09
0.04
0.03
-0.03
-0.05
-1.37
0.00
0.74
0.11
0.34
0.57
-0.06
1.84

Std. Err.
0.09
0.09
0.10
0.11
0.10
0.02
0.01
0.32
0.38
(omitted)
0.33
0.39
0.57
0.46
0.07
1.28

z
-2.76
-0.50
0.36
-0.81
0.43
2.02
-2.26
-0.16
-3.59

P>|z|
0.01
0.62
0.72
0.42
0.67
0.04
0.02
0.88
0.00

[95%
Conf.
-0.44
-0.21
-0.16
-0.30
-0.15
0.00
-0.06
-0.68
-2.11

2.23
0.29
0.60
1.23
-0.89
1.44

0.03
0.78
0.55
0.22
0.38
0.15

0.09
-0.65
-0.78
-0.33
-0.20
-0.66

Interval]
-0.07
0.13
0.23
0.13
0.23
0.06
0.00
0.58
-0.62
1.38
0.87
1.47
1.47
0.07
4.35

When focusing our attention on those entrepreneurs who are grossing less than $200,000 per
annum, we can see that treatment does appear to be having some positive impact not on overall
sales but on the amount of sales sold through retail outlets. To see this, first consider a standard
OLS of sales (restricted to those grossing under $200,000) as is shown below in table 32. In this
case, the treatment variable though now with a positive coefficient, is not found to be significant.
In fact, only the age of the venture and the industry indicator appear to be significant predictors
of sales in this case.

128

Table 32: OLS of gross annual sales on treatment and controls, restricted to respondents
grossing less than $200,000 per annum (statistically significant variables in bold, N=203,
R2=0.1266)

grosssales
propinfoseek
age
indusexp
baormore
male
white
familyentrep
foodproc
inherit
pushed
ventureage
pcassist
_cons

Coef.
Std. Err.
-992
1,423
-49
238
29
213
5,550
5,370
2,058
5,019
7,970
6,339
3,075
5,440
16,854
5,568
-7,634
9,277
-4,490
6,044
4,223
1,275
6,073
5,479
-13,790
17,068

t
-0.70
-0.20
0.13
1.03
0.41
1.26
0.57
3.03
-0.82
-0.74
3.31
1.11
-0.81

P>|t|
0.49
0.84
0.89
0.30
0.68
0.21
0.57
0.00
0.41
0.46
0.00
0.27
0.42

[95%
Conf.
Interval]
-3,798
1,815
-517
420
-392
449
-5,042
16,142
-7,842
11,957
-4,534
20,474
-7,655
13,806
5,872
27,836
-25,933
10,664
-16,411
7,431
1,709
6,737
-4,735
16,880
-47,458
19,877

This result is particularly interesting because when we were evaluating the impact
assistance had on the perceived legitimacy of clients, and by construction of the volume in retail
variable we had to restrict the sample to those grossing under $200,000 per year, we found a
significant positive impact. In this case, it appears that assistance was not significant in boosting
overall sales however. This may indicate that the results of the increase in legitimacy don’t
always directly translate into increased sales (at least not immediately) but may be behind the
increased survival rates as well as future potential for sales increase once the higher legitimacy
obtained percolates into increased customer loyalty and brand recognition.
Continuing on to the other performance variables, we do find a positive correlation with
assistance and employment as is evidenced below in table 33.

129

Table 33: OLS regression of Total Employment on treatment and controls restricted to those
grossing under $200,000 per annum in sales. (statistically significant variables in bold, N=186,
R2=0.1085)
totemp
Coef.
Std. Err. T
propinfoseek
-0.03
0.14
age
0.03
0.03
indusexp
0.03
0.04
baormore
1.14
0.60
male
0.44
0.57
white
-0.36
0.81
familyentrep
-0.68
0.70
foodproc
0.75
1.52
inherit
-1.43
0.71
pushed
-0.93
0.48
ventureage
0.18
0.14
pcassist
1.02
0.61
_cons
-1.23
2.10

P>|t|
-0.19
1.25
0.70
1.92
0.76
-0.44
-0.96
2.03
-2.00
-1.94
1.28
1.67
-0.59

0.85
0.21
0.48
0.06
0.45
0.66
0.34
0.04
0.05
0.05
0.20
0.10
0.56

[95%
Conf.
Interval]
-0.31
0.26
-0.02
0.09
-0.05
0.10
-0.03
2.32
-0.70
1.57
-1.96
1.24
-2.06
0.71
0.05
3.00
-2.84
-0.02
-1.87
0.02
-0.10
0.45
-0.18
2.22
-5.38
2.92

It should be noted, however, that this variable is only weakly significant, and once
additional entrepreneurial orientation variables are added (such as proprisktake and
propinnovate) we see this variable drop out of significance slightly. Nevertheless, there does
appear to be at least some impact being observed here. We can also see that the inheritance
variable remains a significant negative predictor for this group, which gives some support to the
previous claim as one would not expect the lifestyle entrepreneur to be among the group
excluded from this regression (i.e. those with sales >$200,000 per annum). Also, those pushed
into entrepreneurship are, not surprisingly, less likely to be experiencing employment growth.
Whereas, those in the food processing industry and higher education levels are experiencing
higher employment growth than those who are not. Furthermore, it also appears when one
restricts the sample on the basis of level of sales as we have, the strong correlations between
males as well as those with Caucasian ethnicities seems to disappear. A similar effect also
130

occurred with the sales analysis above, especially with the gender variable, where the same effect
occurred, and the impact of the ethnicity variable was lessened. This perhaps indicates that being
Caucasian and male is highly correlated with operating those large firms, but is less apparent in
the smaller ones.
The results of this regression appear to give us another area where we can reject H1-f for
those grossing under $200,000 per annum in sales. It is also interesting to note that in terms of
total sales for this group, there was no significant difference found between the treated and nontreated group (though there was a difference found in volume in retail). It is difficult to speculate
why this might be, but it is an area that would warrant further research.
Next, we find that EAP assistance is significant in improving the probability a small
firm’s entrepreneur will receive external financing as is indicated below in table 34.
Table 34: Probit of receiving external financing on treatment and controls restricted to those
grossing under $200,000 per annum. (statistically significant variables in bold, N=188, Pseudo
R2=0.0711)

Y=externfinance
indusexp
baormore
age
age2
propinfoseek
ventureage
pushed
inherit
familyentrep
foodproc
pcassist
white
male
_cons

Coef.
0.01
-0.19
-0.19
0.00
-0.02
0.08
-0.10
0.09
-0.07
-0.08
0.43
-0.05
-0.02
3.74

Std.
Err.
z
0.01
0.23
0.06
0.00
0.06
0.04
0.28
0.38
0.23
0.24
0.23
0.34
0.21
1.67

0.90
-0.84
-3.18
3.04
-0.36
1.82
-0.36
0.23
-0.32
-0.32
1.86
-0.15
-0.09
2.24

131

P>|z|
0.37
0.40
0.00
0.00
0.72
0.07
0.72
0.82
0.75
0.75
0.06
0.88
0.93
0.03

[95%
Conf.
-0.01
-0.64
-0.32
0.00
-0.14
-0.01
-0.66
-0.65
-0.52
-0.55
-0.02
-0.71
-0.43
0.47

Interval]
0.03
0.25
-0.07
0.00
0.10
0.16
0.46
0.82
0.38
0.39
0.89
0.61
0.39
7.02

Interestingly, this finding only holds true, however, when a squared age term is included.
In fact, without the squared age the model does not even fit with any level of significance for any
variable. With it in, however, the upside down U-shaped relationship with age gives some
indication that investors have a preference for entrepreneurs in the middle of the age spectrum
(which for this sample would mean in their late 40s and 50s). Only after this preference is
controlled for do we see that assistance has an influential impact. The same is true for the age of
the venture, though industry experience and “pushed” status are no longer significant as they
were in the unrestricted model. It appears, once again, the impact of assistance is most profound
on those small and medium sized ventures. So while again we cannot reject H4o for the entire
sample, we can do so for those who gross under $200,000 in sales per annum.
J. Implications of these findings
In regards to what the impact is that assistance this particular EAP provides, it appears
that it is most effective at creating value for those smaller business with gross annual sales under
$200,000 per annum. This is not entirely surprising as those businesses that are already doing
well enough to be obtaining significant revenues are probably the least likely to need assistance
and are in fact driving the selection bias that is occurring. In fact, this research has shown that
these small businesses are more likely to be run by older females with a history of family
entrepreneurship, less experience in their venture’s related industry and have a smaller
propensity to grow their business that those with larger firms. It is interesting to note that this
demographic group has already been identified (except the female part) as the cluster who has
received the most incremental benefits from assistance, on average, namely the older
inexperienced cluster from the cluster analysis,

132

Furthermore, until recently, the Product Center’s focus has been on aiding the start-up of
small businesses to go from producing a cottage industry good out of their kitchen to taking the
next step towards becoming a growing commercial product. It helps these businesses at each step
of the way from developing the venture concept, acting as a sounding board to help “weed out”
untenable ideas and “plant in” good ones, to developing a business plan to attract investors, to
obtaining the proper packaging and labeling, to adhering to industry regulations and norms, to
finally gaining access to marketing outlets and developing proper growth strategies. These are all
things that the large business grossing over $200,000 per year has already done and hence do not
find that beneficial. For those small business and nascent entrepreneurs without a lot of
experience, however, this assistance can be invaluable and the data clearly shows that the EAP
has helped to increase survival rates, perceived legitimacy with retailers and external investors,
and produced employment growth. The increased legitimacy brought about by assistance with
external resource holders was also shown to be more prominent with those who received
assistance prior to the launch of their new venture. This indicates that for EAP assistance to be
most effective, it should be engaged in at the beginning of the entrepreneurial process. Trying to
change courses midstream seems to have been less effective.
Those who are finding it useful, are also benefitting from a larger start-up network than
those who do not, which can also help to further improve their probabilities of success. The
assistance also appears to encourage them to be more proactive in the entrepreneurial orientation
(EO) of their firm.
That is not to say that this assistance has always been effective or that there is no room
for improvement. As some of the comments and evaluations from the less satisfied users bears
out, clearly there are ways in which the Product Center can improve its ability to create value for

133

entrepreneurs. The significant lag times between counselor contact with entrepreneurs, the
apparent heterogeneity between counselor effectiveness and the passing on of information on
what to do without enough follow up and support to get it done all indicate areas that the EAP
could improve its effectiveness. Whether this is due to a scarcity of counselors to respond to the
abundance of clients seeking assistance or a scarcity of skill level amongst those counselors is an
area that is beyond the scope of this analysis, but should be investigated further.
Furthermore, while helping the small and medium sized businesses improve their
survival, legitimacy and growth rates is important, so too is helping the most successful ventures
become more successful. These large, fast growing ones are often the ones most associated with
the ability of entrepreneurial activity to create economic growth (e.g. Reynolds et al, 2000). In
fact, with its High Impact Value Action Team (HI-VAT) the Product Center has started to cater
to this group a bit more, but it is still too early to gauge its impact in the data. Therefore, future
research will need to take this into account as well.
It is also important to note that the entrepreneurs within this sample are quite a diverse
and heterogeneous mixture. Therefore, it is not surprising that, though many significant
relationships were found, the overall R2 value of the regressions was often under 0.1.
Furthermore, there was clearly evidence of response bias in terms of survival rates when
compared to figures provided in the literature (e.g. Shane, 2008). While these should raise some
concern on the part of the validity of the results contained within this paper, it is hoped that so
long as the response bias is not skewed towards treatment or non-treatment, and the known
differences in heterogeneity across entrepreneurs (i.e. the selection bias effect) has been
sufficiently accounted for in terms of bounding the resulting estimates as conservative, and no

134

other unobserved bias is at work, then some confidence can be had in the validity of the average
treatment effects uncovered.
Therefore, in conclusion, it appears that EAPs can significantly create value for those
entrepreneurs most at risk from the “liability of newness.” The pre-launch counseling that helps
to “weed out” untenable ideas and “plant in” good ones before resources are committed to the
new venture creation process appears to be effective at increasing overall survival rates and
potentially even the employment growth seen. The strategic counseling on marketing strategies,
the specialized services provided on product testing, labeling and packaging, as well as the
direct ties developed by the Product Center with retailers appears to be helping to legitimize the
smaller new ventures with key trading partners and end consumers. While lastly, the business
planning services and ties to resource holders such as GreenStone Farm Credit Services and
Comerica bank also appear to be helping the smaller ventures gain more legitimacy with
potential investors. Though not perfect, it does appear that EAPs can help facilitate the new
venture creation process to the point that for those who have received assistance, and found it
useful, in the words of one respondent, “It would have been much more difficult to do this
without the help MSU has provided.”

135

VIII.

SUMMARY OF FINDINGS
The motivation behind this research was to determine if one particular Entrepreneurial

Assistance Program (EAP), the Michigan State University Product Center, a publically financed
organization, has created value for its food and agribusiness clientele. This lead to the positing of
the primary research question:
How much value has the Product Center, as an example EAP, created for its clients? R1
To answer this question, we first conducted an extensive literature review of past EAP
impact studies. This review gave some indication that these programs were creating value in
some particular situations, such as with less experienced entrepreneurs and those with limited
resources at hand. But there was a strong concern of selection bias on the basis of who seeks
assistance. This bias results from the fact that known factors that cause entrepreneurs to seek
assistance such as: 1) being a proactive information seeker and/or 2) having lower
entrepreneurial ability will also influence the overall probability of success of that entrepreneur
(positively in the former and negatively in the latter). Therefore we also determined that we must
first answer the following two secondary research questions before a conclusive answer to the
primary research question could be given:
What types of entrepreneurs seek assistance and why?

R2

What types of entrepreneurs find assistance useful and why?

R3

To conduct this analysis, a set of hypotheses was developed for each question and a
survey created to obtain the necessary data to test these hypotheses. This survey was sent out to
all individuals who had contact with the Product Center from 2004-2012, as well as to a control
group of entrepreneurs who had applied for a new food license in Michigan during that same
time period. Non-entrepreneurial respondents were screened out, as well as incomplete

136

responses, and the remaining respondents were used as the data set to test the proposed
hypotheses.
Using this data we first conducted a cluster analysis on the respondents to determine if
there were distinct groupings of entrepreneurs within the sample. Three different types of
entrepreneurs emerged from this analysis: the Seasoned Entrepreneurs – who consisted of older
veterans of the industry with decades experience who have spent a substantial amount of time as
self-employed; the Late Bloomers – older newcomers to the industry with a modest amount of
experience owning their own business but less so in the industry of their new ventures; and the
Young Bucks – who are well educated but less experienced younger entrepreneurs often starting
up a business for the first time. The Young Bucks were the most likely cluster to seek out
assistance, but the least likely to be satisfied with it, whereas the Seasoned Entrepreneur was the
least likely to seek out assistance but most likely to be satisfied with it. Interestingly enough,
however, when looking at the incremental gains provided by assistance by comparing within the
clusters we found that the Late Bloomers who received assistance had the highest difference
from the non-assisted group in median sales and volume in retail, but the assisted Seasoned
Entrepreneurs had the highest overall improved survival rates and ability to obtain external
financing when compared to the non-assisted Seasoned Entrepreneurs. The Young Bucks, in
comparison, had the lowest overall improvement from assistance. This appears to indicate that
the Product Center is providing the most benefit to its older entrepreneurial clientele than its
younger ones and satisfaction appears most correlated with ability of the Product Center to help a
venture improve its survival probabilities and increase its perceived legitimacy with resource
holders.

137

Next, for the sample as a whole, we analyzed what types of entrepreneurs seek assistance
and why. This was done in order to ensure we were taking proper steps to deal with selection
bias so we could quantitatively determine if the Product Center was indeed creating value for its
clientele in the ways postulated above. In this regard we found support for the notion that seekers
of assistance are proactive information seekers, though could not find any support for the notion
that seekers of assistance were more likely to have lower entrepreneurial abilities (see table 35
(below) for a summary of R2, the hypotheses we put forth to answer this question and the
findings).
Table 35: Research question 2, hypotheses and findings
R2

What types of entrepreneurs seek assistance and why?

Reject Null
(y/n)

H2

Entrepreneurs with less entrepreneurial experience and/or ability will
be more likely to seek assistance than those with more experience
and/or ability.

N

H3

Entrepreneurs with a higher propensity to seek information will be more
likely to seek assistance than those with a lower propensity to seek
information.

Y

Unfortunately, we were not able to identify a valid instrument for the aforementioned
selection bias concern, but by using the omitted variable proxy (propinfoseek) we can mitigate
the positive selection bias concern associated with proactive information seekers being more
likely to succeed through removing the correlation of success and propensity to seek information
from the error term of the treatment coefficient’s estimator. In regards to the downward selection
bias effect of treatment due to seekers of assistance being more likely to have lower
entrepreneurial abilities, though we have not been able to instrument this away or find an
appropriate proxy for this omitted variable, we have shown that this bias is downward bounded

138

and would therefore not cause any significant positive impacts identified to be invalidated. This
is because the effect of this bias would only cause estimators to be more negative than the actual
effect, and therefore if a regressant coefficient is estimated to be significantly positive, the actual
effect of that variable would only be more positive assuming the upward bias is removed through
the use of the propinfoseek variable.
In addition, we also found evidence to suggest that seekers of assistance were more likely
to have a strong propensity to grow their business (propgrowbiz), have achieved higher
educational status (baormore) and have a history of entrepreneurship in the family
(familyentrep). Given the influence that these characteristics could also have on overall
performance, they were also included in the impact analyses to remove any other potential
selection bias concerns. We also found that although the majority of the entrepreneurs in the
sample were Caucasian (approx. 90%), non-Caucasians we slightly more likely than their
Caucasian counterparts to seek assistance.
We next examined what the differences were between those who found assistance useful
or those that did not. This was done to give us an indication of what type of entrepreneur would
be most likely to receive value from EAP assistance. In this case we found that those who sought
assistance prior to the launch of their new venture were more likely to find that assistance useful
than those who sought out assistance only afterwards. Though there was some weak indication
that the timing of when one sought assistance was positively correlated with the respondents’
propensity to seek information and negatively correlated with their industry related experience.
In addition, we also found that those respondents who had launched their ventures and had sales
under $200,000 per year as well as had other characteristics associated with lower probabilities

139

of entrepreneurial success, such as a lack of experience in the industry, were more likely to find
assistance useful (see table 36, below).
Table 36: Summary of research question 3, relevant hypotheses and findings
R3

What types of entrepreneurs find assistance useful and why?

H4

Entrepreneurs who seek assistance prior to launch will be more likely
to find that assistance useful than those who seek assistance only after
launch

Reject Null
(y/n)
Y

H4-a

Entrepreneurs who seek assistance prior to launch will be more likely
to have a higher propensity to seek information than those who seek
assistance only after launch.

Y

H4-b

Entrepreneurs who seek assistance only after launch will have
significantly less entrepreneurial ability than those who seek assistance
prior to launch.

Y

H5

Entrepreneurs who have launched products and have less
entrepreneurial experience and probabilities of success without
assistance will be more likely to find assistance useful.

Y

H6

Entrepreneurs from firms whose gross annual sales is less than
$200,000 per year will be more likely to find assistance useful.

Y

H7

Entrepreneurs who seek assistance and then decide to not launch a
new venture will have been less likely to be successful than those who
have persisted with the assistance process.

Not enough
information

Respondents who were well-educated entrepreneurs and those with a family history of
entrepreneurship were, in addition to more likely to seek assistance, also more likely to find that
assistance useful. This poses further questions to be investigated as to why the Product Center’s
services attract and satisfy those entrepreneurs with higher educational degrees and/or those with
other entrepreneurs in the family more than those without higher degrees or family
entrepreneurs. One potential explanation is that the language and paradigms used at the Product
Center are presented in a way that is easier to understand for entrepreneurs who are already
140

familiar with the business or the university environment, but is less accessible for those who are
relatively inexperienced in either arena.
In contrast to the findings above, those who did not find assistance satisfactory or useful
often complained that the information provided by the Product Center was too vague, and
counselors did not follow up quick enough or provide enough specific direction on what to do
next. This gives further support to the idea that assistance is most valuable to those with enough
business experience or education that allows them to fully appreciate and utilize the information
provided as opposed to the inexperience nascent entrepreneur who is new to the industry.
Lastly, after fully examining and accounting for the different characteristics of the
clusters of entrepreneurs, those who seek assistance and those who find it useful, we sought to
quantitatively identify the value created by the Product Center. Here we saw first and foremost
that the Product Center was very effective at “weeding out” less tenable business ideas and
“planting in” sound ones evidenced by the lower launch rates but higher survival rates amongst
those entrepreneurs who sought assistance prior to the launch of their new ventures. We also
found that those who received assistance prior to the launch of their ventures with those who
received assistance only afterwards were more likely to have improved performance statistics
such as having higher overall sales and the increased ability to obtain external financing, when
controlling for other relevant variables, indicating the primary benefit from EAP assistance
comes from the pre-launch business idea counseling and development.
When considering the entire sample group as a whole we did not see any noticeable
differences in total employment growth or ability to obtain financing, and actually lower overall
sales growth with the assisted group versus the non-assisted. However, when one eliminated the
large firms with gross sales over $200,000 per year as characteristically different businesses from

141

the smaller remaining 85% of respondents (i.e. gross sales under $200,000 per year), we did
observe that assistance appeared to be significant in increasing the legitimacy of new ventures
and led to an improved ability to obtain financing and larger marketing opportunity sets. We also
saw that amongst smaller firms, assistance was more likely to increase total employment growth
(see table 37 below). In short, assistance is most beneficial to those with smaller firms and who
are seeking out assistance prior to the launch of their product. This assistance helps to improve
survival rates and perceived legitimacy amongst resource holders, but does not necessarily lead
to dramatic performance improvements.

142

Table 37: Summary of research question 1, relevant hypotheses and findings
R1

How much value has the Product Center, as an example EAP,
created for its clients?
Entrepreneurs who receive assistance from an EAP will be more likely
H1
to have higher performance, survival rates and legitimacy than had
they not sought assistance.
Entrepreneurs who receive assistance from the example EAP prior to
H1-a the launch of a new venture will be less likely to launch that new
venture than those who do not (weeding out hypothesis).
Entrepreneurs who receive assistance from the example EAP prior to
the launch of a new venture and decide to launch that new venture will
H1-b be more likely to stay in business than those who launched a new
venture but did not receive assistance (planting in hypothesis).

H1-c

Entrepreneurs who receive assistance from the example EAP will be
more likely to obtain external financing than had they not received
assistance.

Reject Null
(y/n)
Mixed

Y

Y

N

H1-d Clients from the example EAP will have access to a greater marketing
opportunity set than they otherwise would have without assistance.

Y

H1-e

Clients from the example EAP will have higher gross annual sales
rates than they otherwise would have without assistance.

N

H1-f

Clients from the example EAP will higher employment rates than they
otherwise would have without assistance.

N

Clients who receive assistance prior to the launch of their new venture
H1-g will receive greater impact from assistance than those who receive it
only afterwards.
H1-h When restricting the sample to entrepreneurs whose new venture
grosses under $200,000 per year, the impact of the example EAP
assistance will be more likely to be significantly positive than when
compared to the full sample.

Y

Y

In addition to the support provided by the Product Center, we also found that younger
entrepreneurs with older ventures, high propensity to grow the business and levels of industry
experience were correlated with increased sales growth. These characteristics were also strong
predictors of the ability to obtain financing, so it appears that resource holders show some
143

efficiency in recognizing the market potential of entrepreneurs who seek finances. In terms of the
cluster analysis, this would indicate that late bloomers would have the hardest time obtaining
these resources as well as succeeding in the market. This likely explains why the highest
incremental gains within a particular cluster when comparing assisted respondents to nonassisted respondents’ sales, average volume in retail and ability to obtain financing from
assistance comes from the late bloomers.

144

IX.

EAP Managerial Implications

The results of this research provide several implications to those who are in the business of
managing entrepreneurial assistance programs. The first implication for managers to consider is
based on who they desire as the target clientele of an EAP. Our research has shown that
proactive information seekers, well-educated entrepreneurs, and those with entrepreneurs in the
family are more likely to seek assistance than those without these characteristics. If the goal of
the EAP is to attract as many entrepreneurs as possible, EAP managers should look how to better
target and provide useful assistance to those with lower education levels and without a history of
entrepreneurship in their families as well. This can be done through outreach programs aimed at
nascent entrepreneurs without business experience or educational backgrounds. Care should be
taken not to involve too much business jargon or know-how as an implied pre-requisite and
instead promote workshops and services that provide the most basic level of services. In
addition, managers should look for rural or urban venues that would attract food entrepreneurs of
this type as well as hire counselors who come from less traditional backgrounds. Our research
has also shown that targeting small ventures run by entrepreneurs with less experience in the
industry and are still in the pre-launch phase of the venture creation process will also be more
likely to find that assistance useful.
Given the limited resources EAPs have allocated to them through public funding, EAP
managers must balance catering to the most nascent of entrepreneurs with the concern of
providing too much assistance to those ventures with less tenable ideas or less able
entrepreneurs. This may seem counterintuitive, given EAPs are meant to help entrepreneurs fill
resource gaps such as lack of business knowledge or supply chain networks. While this is true, it
would appear that EAP managers should target the “Goldilocks”-type entrepreneurs.
145

For example, we saw in the cluster analysis that small ventures grossing under $200,000
per year with a late bloomer founder were achieving the most incremental gains from assistance.
By contrast, the gains in the other two clusters may have been less because the seasoned
entrepreneur may already have the knowledge and resources that the EAP was providing and the
young buck may not have enough experience to know how to use those resources effectively.
This is not to say that an EAP should refuse assistance towards any group, but instead, the EAP
should structure its process to facilitate entrepreneurs who will use the information and
assistance provided to maximum effect and discourage those who will not.
In this paper, we have found evidence based on the lower launch but higher survival rates
of its clients that this particular EAP is achieving this goal for those that seek assistance prior to
the launch of their products by “weeding out” untenable venture ideas whilst “planting in” good
ones. According to the Product Center’s core staff, this is done through the “concept definition”
and “business planning” process where an entrepreneur has to define its business idea, target
market, customers, suppliers, processes and so on and gives the entrepreneur a better
appreciation of the work involved and a “reality check” on the viability of their idea (Lovgren,
2012). This sudden realization of the work involved and external validity (or lack thereof) of the
business idea creates a screening process that can help entrepreneurs decide whether to launch
the venture prior to committing significant time and resources to the idea. This is perhaps why
we see that those small ventures with entrepreneurs who have received assistance and persisted
through the Product Center process from start to finish are achieving measurable gains in
survival and employment rates, plus increasing their perceived legitimacy amongst resource
holders and trading partners. Managers in turn can highlight this screening process to those

146

resource holders to further enhance this effect and thus create a strong signal or “badge” ”(Bell et
al., 2002) to the market.
However, this EAP has been less successful at creating value for those who seek
assistance only after they have launched their ventures. It may be that for this group the scope of
services required (i.e. obtaining nutritional fact labeling) is beneficial but will have less
noticeable impact on performance than the pre-launch assistance provided. This is because prelaunch assistance such as business strategy planning can have a greater overall impact on future
performance and may not be explicitly carried out by the control group, whereas post-launch
assistance such as nutritional facts labelling is something the control group who must also obtain
for market entry and therefore the incremental impact between treatment and control would be
less noticeable. Managers should consider whether public dollars should be spent on such
services or whether these services should be fee-based. The rationale would be that public dollars
should go to services that promote the public good. One-off services such as nutritional fact
labeling, packaging or other market entry requirements that all entrepreneurs must obtain
promote more the private good of the entrepreneur by saving them the cost of paying for this
service than the deeper level strategic and business development services which have a greater
impact on improving the overall probability of success and survival of the venture and thus are a
greater benefit to society.
We have also seen that those who did not find assistance useful often complained of
information as being too general and/or vague with regards to the specific steps to take next.
These comments are more likely to come from those entrepreneurs who already have some
experience and business know-how, such as the seasoned entrepreneur and the larger businesses
(i.e. >$200,000 in sales) who maybe seeking deeper strategic consulting services. This is not to

147

be confused with the earlier finding that those with some level of entrepreneurial or business
background were more likely to seek assistance and find it useful, but rather to indicate that after
a certain threshold of experience and background, that assistance becomes less useful. So while
some business knowledge will drive enterpreneurs to seek assistance and find it useful by
allowing the entrepreneur to more readily absorb the information provided, those with extensive
business knowledge will not find additional benefit from hearing what they already know. This is
likely why these groups have been shown to have less incremental benefit from assistance and is
a gap EAP managers must decide whether or not to address. One the one hand, these
entrepreneurs already have a significant amount of knowledge and resources at hand and it may
be more equitable for an EAP to focus instead on the “have-nots” of the entrepreneurial world.
On the other hand, the larger businesses have the potential to create more jobs and customer
value so could be providing more societal benefits. This decision will depend on the mission and
vision of each EAP on a case by case basis. This research has shown, however, that other than
survival, no other significant differences in venture performance from assistance were evident
when including the larger firms in the sample. So if improving the larger businesses’
performance is a goal, the EAP must develop more specific services and acquire or develop more
experienced personnel towards larger businesses who have moved past the start-up phase. In the
case of the Product Center, the HI-VAT (HIgh Velocity Action Team program was developed to
respond to this need.
Lastly, there also were consistent complaints of long lag times between Product Center
contact and follow up from EAP counselors. While we had mentioned earlier that building in
screening mechanisms that “weed out” untenable business ideas can be a good thing, the tradeoff is screening out potentially good ideas based on frustrations with the services provided. In

148

this manner, EAP managers should consider whether they feel they have an adequate staff of
counselors and specialized service providers to handle the workload of cases brought to the EAP.
Given the public financing, which by nature will be limited, and the lack of charge for most
services, it is likely that most EAPs will therefore consistently be understaffed in regards to
demand from entrepreneurs. Managers should consider this in their structuring of the assistance
process by making the service provided streamlined and tiered. That is to say, managers should
develop a clear screening mechanism for each stage of the assistance process and tailor the
amount of service provided to be inversely proportional to the amount of cases likely to be
present at each stage. For example, initial counseling on the business idea will have the most
cases, and therefore should have the least amount of time associated with it and have a built in
screening mechanism, such as homework on the concept definition process. This is, in fact what
is currently being practiced at the Product center (Lovgren, 2012) for this stage. However, it was
unclear if the additional stages or next steps were incorporating additional screens, such as
market validation or food license obtainment, as a requirement for future service. Nor was it
clear that the amount of time allocated for more advanced stages was inversely proportional to
the amount of clients to reach that stage. EAP managers should work to make this process
explicit and incorporate the core staff team in its structuring so as to ensure buy-in from the key
personnel in charge of implementation. By implementing this change, an EAP can go a long way
towards reducing the long lag-times due to understaffing without the ability to use a market
mechanism (i.e. raising the price of service) to limit demand.
These implications are summarized below in table 38:

149

Table 38: Implications for EAP managers
Implications for EAP Managers
EAP managers should improve targeting of nascent entrepreneurs without business
1 experience or educational backgrounds by promoting the most basic level of services,
especially in rural or urban areas which do not have strong support infrastructures built for
entrepreneurs.
2 EAP managers should structure the assistance process to facilitate entrepreneurs who will
use the information and assistance provided to maximum effect and discourage those who
will not by creating screening mechanisms and explicit market validation checks.
EAP managers should charge fees for services that promote the private good of the
3 entrepreneur and have not been shown to increase the overall public good through evidence
of increased venture performance.
4 If improving the larger businesses’ performance is a goal, EAPs must develop more specific
services and acquire or develop more experienced personnel directly targeted towards larger
businesses who have moved past the start-up phase.
In order to improve the efficiency and effectiveness of assistance provided, EAP managers
5 should develop a clear screening mechanism for each stage of the assistance process and
tailor the amount of service provided to be inversely proportional to the amount of cases
likely to be present at each stage.

150

X.

FUTURE RESEARCH IMPLICATIONS
In terms of the contribution this research provides to the entrepreneurial and EAP

literature streams, there are many potential research implications. We find support for Chrisman
and McMullan’s (2005) assertion that assistance acts as a special type of knowledge resource for
nascent entrepreneurs. The fact that more experienced entrepreneurs find assistance less useful
and that the impact of assistance is most readily observable for the small and medium sized firms
both indicate that this assistance acts to help fill resource gaps for entrepreneurs that lack the
particular set of resources the EAP is skilled at providing. The larger, more experienced firms
likely have such resources, knowledge and social ties already established and therefore do not
benefit observably from such assistance. However, Chrisman and McMullan’s (2005) claim that
they have overcome the selection bias problem by showing that the benefits accrued from
assistance correlate with the amount of contact hours in a curvilinear relationship does not appear
well supported. Indeed, we have found the same relationship to hold true for those who have
found assistance useful, but there does not appear to be any grounds for the exogeneity of this
measure (thus it was omitted from the results). Those entrepreneurs who find assistance useful
will likely return for more assistance, thus increasing the relationship between contact hours and
performance benefits. Whereas those less motivated individuals whose business is faltering
because of their lack of commitment, or lack of time to devote to the venture idea and
development, will likely receive less contact hours and also exhibit lower performance. In
addition, those individuals who substitute assistance for actual entrepreneurial activity will likely
receive the most contact hours, but have low performance characteristics, and will hence drive
the negative quadratic term that dictates the diminishing returns to scale. As all of these factors
influence both number of contact hours received and performance, it is ]difficult to conclude

151

there is a definitive causal link between the two that can be estimated exogenously. Without such
an estimation, this method does nothing to overcome selection bias, but simply adds another
form of bias into the estimation of the average treatment effect on the treated. For such a
procedure to be valid, the number of contact hours given to the entrepreneur would have to be
randomly assigned ex-ante to treatment, which is obviously not feasible for most EAPs since the
goal of the EAP is not to measure its own effect, but rather to help each entrepreneur that walks
through its doors to the best of its ability.
We have also shown that researchers are well-founded in their concern over selection
bias as we have provided evidence that both the propensity to seek information and the lack of
experience are driving factors in the decision to seek assistance and find that assistance useful
(and hence persist with it). Even though we were not able to find an instrumental variable to
exogenously identify our regressions, we have shown that selection bias can be removed or at
least bounded in linear regression models if one can sufficiently account for the various
influences of the selection through observable measures that proxy for the underlying effect.
That is to say, as selection bias can be viewed as a missing variable bias that, when uncontrolled
for, causes correlations between the error term and the regressor, if one can control for it with an
observable measure such that it is removed from said error term, then the resulting estimation of
the treatment variable would be contemporaneously exogenous. This is in fact what the Heckman
procedure does with the Inverse-Mills ratio (though this procedure requires an instrument to
identify the selection model to create the ratio in the first place). An attempt was made in this
study follow this procedure. The propensity to seek information bias was controlled for through
creating a direct scale that respondents could indicate their own preference on seeking
information versus relying on intuition and trial and error. In addition, the entrepreneurial ability

152

bias was in part controlled for by including education and industry experience. Given that there
is some concern that true ability might still transcend all of those measures as an innate
characteristic of the respondent, however, there is still some concern over this bias so all results
are presented as lower bound.
We also found that the entrepreneurs who seek assistance prior to the launch of their
venture were categorically different than those who sought it only afterwards. This is an
important finding, because the main distinguishing characteristic between those who sought
assistance pre- and post-launch of their venture were their propensities to seek information and
overall industry-related experience. Since these two factors are the same factors that are causing
selection bias within EAP studies, if researchers do not control for the timing of assistance in
their impact evaluations they will be confounding their results by mixing two heterogenous types
of entrepreneurs within the treatment group. Since those who seek assistance pre-launch are the
most likely to be influenced by selection bias in estimation of the treatment effect, researchers
should look to separate out this group via the statistical methods mentioned in this paper in order
to achieve unbiased results.
Next, are the three clusters of entrepreneurs identified in this study: the seasoned
entrepreneur, the young buck, and the late bloomer in fact distinct types of entrepreneurial
groups that are common amongst entrepreneurs in all industries or was this a simple by-product
of random chance in the observed sample? Are there other types of entrepreneurs not identified
here that we should also be concerned with?
There were also many interesting relationships found within this study that warrant
further investigation. Researchers should consider why those with higher education levels and a
history of entrepreneurship in the family are more likely to seek out assistance and find it useful.

153

Is the assistance provided not accessible enough for those with lower education levels or direct
exposure to the business world? Does the university setting of this particular EAP naturally
encourage others with university educations to seek out assistance from it? Or are those
entrepreneurs with higher levels of education more able to utilize the assistance provided in an
effective manner?
Are those who grew up around entrepreneurs more likely to seek assistance because they
know firsthand the hardships of entrepreneurship or are they simply more aware of the support
networks out there? Is there a genetic component to this behavior? Why do those with
entrepreneurs in the family find it more useful as well, when they should also have a “built-in”
counselor already in their families to talk to?
Furthermore, another area of research to look into is why were non-Caucasians more
likely to seek treatment as a group than their Caucasian counterparts? The overwhelming
majority of the sample was Caucasian, indicating non-Caucasians were less likely to go into food
entrepreneurship but more likely to seek assistance. Why is this happening? Is it a result of the
Product Center’s presence in metropolitan areas or does this indicate there is less support for
minority entrepreneurs in Michigan and thus a higher need to seek assistance?
Another interesting question to pursue further is the relationship between those receiving
treatment and the size of their start-up network. It would appear obvious that those clients who
seek assistance prior to the launch of their product and find that assistance useful, would increase
the size of their start-up network simply by the inclusion of the EAP counselors themselves.
Furthermore, Lovgren (2012) suggests one of the primary mechanisms used by EAPs to create
value is to link entrepreneurs indirectly to resource holders, trading partners and other potential
collaborators that could work with the entrepreneur for mutual benefit. However, it may be that

154

those who have strong social skills and the ability to collaborate with others on projects are also
more likely to seek assistance from others in order to solve a problem and this is what is driving
the apparent correlation. Given that the size of one’s start-up network has been linked with
performance, it will be important to determine if assistance is exogenously impacting clients’
start-up networks in order to count this as benefit provided by the EAP. This could best be done
by measuring the size of the client’s start-up network when they first receive assistance to the
size when the product is launched, and then compare this difference to the difference between the
size of the entrepreneurs in the control group’s start-up network of at the beginning of their work
on the venture idea to when they launch product.
Another puzzling result was the apparent negative relationship between industry
experience and survival rates uncovered in this study. Was this a by-product of the apparent
survival response bias, where those entrepreneurs who have dropped their business ideas appear
less likely to respond to surveys about those ideas? Or is there a relationship between the
strategic actions taken by entrepreneurs with more experience, such as engaging in riskier
behavior or faster expansions when the economy was doing well, that resulted in higher failure
rates once the economy went into a recession such as the one brought on by the recent housing
market crash?
Similarly, the negative relationship between receiving an inheritance (which was meant
to proxy for not having capital constraints) and employment levels uncovered is also puzzling.
The lack of capital constraints, one would presume, would allow entrepreneurs more flexibility
in hiring employees to help grow the business. It may be, however, that given the relatively small
size of most of the ventures within this sample, those with inheritances have decided to become
entrepreneurs to pursue a “hobby” or “lifestyle” business that is not intended to grow into a large

155

commercial operation, but rather provide a reasonable return to the entrepreneur’s passion for
producing a particular product. Many entrepreneurs in this survey have indicated such motives
for this business, as was the case with the entrepreneur involved in selling the “wild-harvested”
goods quoted above, and for those types of entrepreneurs it is unlikely that they would be
interested in hiring additional employees.
Finally, why is it that these results have not shown an increase in total sales growth for
the assisted entrepreneurs grossing under $200,000 per annum versus the control group, but have
shown an increase in employment levels for the same category? One would expect higher sales
and higher employment levels to be strongly correlated (as one needs more sales to pay more
employees). This survey did not take account of the salary being afforded these employees, so it
is difficult to ascertain whether assisted clients are running their operations more efficiently and
can therefore hire more employees, or if the unassisted ones are more efficient by not having
more employees. Therefore, further research should look deeper into what is occurring here.
These questions for future research are summarized below in Table 39.

156

Table 39: Implications for future research.
Implications for future research
Chrisman and McMullan’s (2005) assertion that assistance acts as special type of knowledge
1 resource for nascent entrepreneurs appears to have found some support within this
dissertation, but not so the claim that selection bias is overcome through the curvilinear
relationship with contact hours.
There is evidence that the concern about selection bias for entrepreneurial ability (based on
2 experience) and the propensity to seek information is well-founded. While an instrumental
variable was no identified, proxies for these unobservable variables were and can be used to
mitigate this concern.
The three clusters of entrepreneurs were identified in this study: the seasoned entrepreneur,
the young buck, and the late bloomer. Are these in fact distinct types of entrepreneurial
3
groups that are common amongst entrepreneurs in all industries or was this a simple byproduct of random chance in the observed sample?
Entrepreneurs who seek assistance prior to the launch of their venture are more likely to
show the characteristics associated with selection bias (e.g. high propensities to seek
4
information and a lack of industry-related experience), and therefore impact evaluations
should estimate this group separately from those who seek assistance only after launch
Entrepreneurs with higher education level and a history of entrepreneurship in the family are
5 more likely to seek out assistance and find it useful indicating that assistance may be
unintentionally targeting these groups over those with less business knowledge/education.
6 The sample was overwhelming Caucasian (90%), but non-Caucasians as a group were more
likely to seek assistance. Does this indicate there is less support for minority entrepreneurs
and thus a higher need to seek assistance?
7 Negative correlations with industry experience and survival, as well as receiving an
inheritance and total employment seem counterintuitive. Are other deeper factors at play?
Why is it that these results have not shown an increase in total sales growth for the assisted
8 entrepreneurs grossing under $200,000 per annum versus the control group, but have shown
an increase in employment levels for the same category

157

REFERENCES

158

REFERENCES

Akerlof, G. (1970), ‘The market for ‘lemons’: quality uncertainty and the market mechanism’
The Quarterly Journal of Economics, 84(3), 488-500.
Aldrich, Howard E. and Marlene C. Fiol. 1994. “Fools Rush In? The Institutional Context of
Industry Creation.” Academy of Management Review, 19, 4 (October): 645-670.
Amit, R., L. Glosten and E. Muller (1990), ‘Entrepreneurial ability, venture investments, and risk
sharing’, Management Science, 38(10), 1232-45.
Anderson, James C. and David W. Gerbing (1988), “Structural Equation Modeling in Practice: A
Review and Recommended Two-Step Approach,” Psychological Bulletin, 103 (3): 411-423.
Arrow, K. (1974), ‘Limited knowledge and economic analysis’, American Economic Review,
64(1), 1-10.
Bateson, P. (1988), “The biological evolution of cooperation and trust.” In D. Gambetta (Ed.),
Trust: Making and breaking cooperative relations: 14-30. New York: Blackwell.
Baron, R.A. (2004) ‘The Cognitive Perspective: A Valuable Tool for Answering
Entrepreneurship’s Basic ‘‘Why’’ Questions.’ Journal of Business Venturing 19(2):221–239.
Barney, J. B., Busenitz, L.W., Fiet, J. O. & Moesel, D. D. (1996). ‘New Venture Teams’
Assessment of Learning Assistance from Venture Capital Firms’. Journal of Business Venturing,
11, 257-272.
Barzel, Y. (1987), ‘The entrepreneurs reward for self-policing’, Economic Inquiry, 25, 103-16.
Behncke, S. Frohlich, M. & Lechner, M. (2010). "Unemployed and their caseworkers: should
they be friends or foes?" Journal of the Royal Statistical Society Series. A, Royal Statistical
Society, 73(1): 67-92.
Bhide, A. (2000), The Origin and Evolution of New Businesses, New York, US: Oxford
University Press.
Blanchflower, D., A. Oswald and A. Stutzer (2001), ‘Latent entrepreneurship across nations’,
European Economic Review, 45, 680-91.
Burt, R. (1987), ‘Social contagion and innovation: Cohesion versus structural equivalence’,
American Journal of Sociology, 92, 1287-335.
Casson, M. (1982), The entrepreneur. Totowa, NJ: Barnes & Noble Books.

159

Calantone, R. J., Cavusgil, S.T. & Zhao, Y,(2002) “Learning Orientation, Firm Innovation
Capability, and Firm Performance.” Industrial Marketing Management. 31:515-524.

Chrisman, Hoy and Robinson (1987) ‘New Venture Development: The Costs and Benefits of
Public Sector Assistance,’ Journal of Business Venturing 2 (fall), 315-328.
Chrisman, J. J. & Leslie, J. (1989). Strategic, Administrative, and Operating Problems: The
Impact of Outsiders on Small Firm Performance. Entrepreneurship Theory & Practice, 13(3),
37-51.
Chrisman, J.J, McMullan W.E., and Hall, J. (2005) ‘The influence of guided preparation on the
long-term performance of new ventures.’ Journal of Business Venturing, 20, 769–791.
Chrisman, J.J. McMullan, W.E. (2000) ‘A Preliminary Assessment of Outsider Assistance as a
Knowledge Resource: The Longer-term Impact of New Venture Counseling’ Entrepreneurship
Theory and Practice. Spring, 2000, 35-53.
Chrisman, Nelson, Hoy, and Robinson (1985) ‘The Impact of SBDC Consulting Activities,’
Journal of Small Business Management 23 (July), 1-11.
Clark, B.W., Davis, C.H. and Harnish, V.C. (1984), “Do courses in entrepreneurship aid in new
venture creation?”, Journal of Small Business Management, Vol. 22, pp. 26-31
Cooper, A., Woo, C., & Dunkelberg, W. (1988) ‘Entrepreneurs' perceived chances for success.’
Journal of Business Venturing, 3, 97-108.
Covin, J.G. & Slevin, D. P. (1989), “Strategic management of small firms in hostile and benign
environments.” Strategic Management Journal. 10: 75-87.
Deeds, D. L., Mang, P. Y., & Frandsen, M. (1997) The quest for legitimacy: A study of
biotechnology IPO's. Paper presented at the annual meeting of the Academy of Management,
Boston.
Dupong, W.D, & Plummer W.D., (1998) ‘Power and Sample Size Calculations for Studies
Involving Linear Regression.” Controlled Clinical Trials, Elsevier.
Granovetter, M. (1985), ‘Economic action and social structure: The problem of embeddedness’,
American Journal of Sociology, 91, 481-510.
Gulati, R. (1995), ‘Does family breed trust? The implication of repeated ties for contractual
choice in alliances’, Academy of Management Journal, 38(1), 85-112.
Hannan, Michael T. and Glenn R. Carroll. (1992) ‘Dynamics of Organizational Populations:
Den- sity, Legitimation, and Competition’. New York: Oxford University Press.

160

Heckman, J., Ichimura, H. and Todd, P. (1998) “Matching as Econometric Evaluation
Estimator,” Review of Economic Studies 65:2, 261-294.
Heckman, J. & Navarro-Lozano, S. (2004) ‘Using Matching, Instrumental Variables, and
Control Functions to Estimate Economic Choice Models’ The Review of Economics and
Statistics, 86(1), 30-57.
Heckman, J. (1979) ‘Sample Selection bias as a specification error.’ Econometrica, 47(1), 153161.
Kaplan, D. (2009) Structural Equation Modeling: Foundations and Extensions. Los Angeles:
Sage Publications.
Knight, F. (1921) Risk, uncertainty and profit. New York: Augustus Kelley.
Kosters, S and Obschonka, M (2010), “Building Winners? An Empirical Evaluation of Public
Business Assistance in the Founding Process,” Jena Economic Research Papers, 25.
Kulicke, M. (2004). Bewertung der EXIST-Initiativen durch die von ihnen betreuten
Gründungen. Umfrage im Rahmen von EXIST „Existenzgründungen aus Hochschulen”.
[Beneficiaries’ assessment of EXIST-initiatives – Survey in the context of EXIST „Start-ups
from universities“]. Arbeitspapier der wissenschaftlichen Begleitung von EXIST, FraunhoferInstitut für Systemtechnik und Innovationsforschung, Karlsruhe.
Larson, A. (1992), ‘Network dyads in entrepreneurial settings: A study of the governance
exchange processes’, Administrative Science Quarterly, 37, 76-104.
Lumpkin, G.T. & Dess, G.G. (1996), “Clarifying the Entrepreneurial Orientation Construct and
Linking it to Performance.” The Academy of Management Review, 21 (1): 135-172.
Low, M. and V. Srivatsan (1994), ‘What does it mean to trust an entrepreneur?’ in S. Birley and
I. MacMillan (eds), International Entrepreneurship, London: Routledge, 59-78.
Marsden, P. (1982), ‘Introducing influence processes into a system of collective decisions’,
American Journal of Sociology, 86, 1203-35.
McMullan, W.E., Chrisman, J.J., Vesper, K. (2001) ‘Some Problems in Using Subjective
Measures of Effectiveness to Evaluate Entrepreneurial Assistance Programs.’ Entrepreneurship,
Theory and Practice. Fall, 2001, 37-54.
Mole, K., Hart, M., Roper, S. & Saal, D. (2008). Differential gains from Business Link support
and advice: a treatment effects approach. Environment and Planning C: Government and Policy,
26(2), 315-334.
Parker, S. C. & Belghitar, Y. (2006). What Happens to Nascent Entrepreneurs? An Econometric
Analysis of the PSED. Small Business Economics, 27(1), 81-101.

161

Reynolds, P.D., Hay, M. and Camp, S.M. (2000), Global Entrepreneurship Monitor: 1999,
Kauffman Center for Entrepreneurial Leadership, Kansas City, MO.
Rosenbaum, P. and Rubin, D. (1983), ‘The Central Role of the Proepensity Score in
Observational Studies for Causal Effects,” Biometrika 70:1, 41-55
Rotger, P.G., Gortz, M., Storey, D.J. (2012), ‘Assessing the effectiveness of guided preparation
for new venture creation and performance: Theory and practice.’ Journal of Business Venturing,
27, 506-521.
Shane, S. (2003) A General Theory of Entrepreneurship: The Individual-Opportunity Nexus.
Cheltenham: Edgar Elgar Publishing limited
Shane , S.(2008) The Illusions of Entrepreneurship, Yale University Press.
Shane, S. and D. Cable (2002), ‘Network ties, reputation, and the financing of new ventures’,
Management Science, 48(1), 122-37.
Solomon, Weaver (1983), ‘Small Business Institute Economic Impact Evaluation’ American
Journal of Small Business, Vol. VIII, No.1, 1983 41.
Stevenson, H.H. and Sahlman, W.A. (1988), “How small companies should handle advisors”,
Harvard Business Review, Vol. 66 No. 2, pp. 28-34.
Stinchcombe, A. L. (1965). ‘Social Structure and Organizations’. In: J. G. March (Ed.),
Handbook of Organizations (pp. 142–193). Chicago, IL: Rand McNally & Company.
Storey, D. (2000) ‘Six steps to heaven: Evaluating the impact of public policies to support small
business in developed economies.’ In D. Sexton & H Landstrom (eds.) Handbook of
entrepreneurship. pp. 176-193. Oxford: Blackwell.
Stubner, S., Wulf, T. & Hungenberg, H. (2007). Management Support and the Performance of
Entrepreneurial Start-ups – An Empirical Analysis of Newly Founded Companies in Germany.
Schmalenbach Business Review, 59(2), 138-159.
Suchman, M. “Managing Legitimacy: Strategic and Institutional Approaches.” Academy of
Management Review, 1995, 20 (3), 571–610.
Stinchcombe, A. (1965) "Social structure and organizations." In J. G. March (ed.), Handbook of
Organizations: 142-193. Chicago: Rand McNally.
Uzzi, B. (1996), ‘The sources and consequences of embeddedness for the economic performance
of organizations: The network effect’, American Sociological Review, 61, 674-98.
Welbourne, T.M. & Andrews, A.O. (1996). Predicting the performance of initial public
offerings: Should human resource management be in the equation? Academy of Management
Journal, 39(4), 891–919.

162

Williamson, O. (1975), Markets and Hierarchies: Analysis and Anti-Trust Implications, New
York, US: Free Press.
Williamson, O. (1985), The Economic Institutions of Capitalism: Firms, Markets, Relational
Contracting, New York, US: Free Press.
Williamson, O. (1986), Economic Organization: Firms, Markets and Policy Control, New York
University Press.
Wren C. and Storey, D. (2002) ‘Evaluating the effect of soft business support upon small firm
performance’ Oxford Economic Papers 54(2).
Wooldridge, J. (2002) Econometric Analysis of Cross Section and Panel Data, Cambridge,MA:
The MIT Press. p.50-51.
Wu, S. (1989), Production, Entrepreneurship and Profit, Cambridge, MA, US: Basil Blackwell.
Yusuf, (2010) ‘Meeting Entrepreneurs’ support needs: are assistance programs effective?’
Journal of Small Business and Enterprise Development, 17(2), 294-307.
Zimmerman MA, Zeitz GJ. (2002)” Beyond survival: achieving new venture growth by building
legitimacy.” Academy of Management

163