Chapter Listing | Return to 1997 Topical Index | Order hard copy editions of Frontiers of Entrepreneurship Research by mail |

The sample for this research consisted of high-technology new ventures issuing an IPO from 1989-1995. The firm must have been an independent startup less than 10 years old and listed as being in one of the high technology industries included in this study. NSF defines "high" technology industries based on direct and indirect R&D product expenditures as a percentage of net sales of that particular product (National Science Board, 1987). Specifically, fabricated metals (34), machinery including computer and computer related equipment (35), electronic equipment(36) professional and scientific instruments (38) communications(48) and business services (73) are the high-technology industries chosen for this dissertation. These industries were selected due to their similarity to each other and , the fact that they were cited as high technology industries in the National Science Board study. Most of theses companies also were manufacturers with the exception of 16 firms represented by SIC code 73. Also, these industries led in terms of R&D expenditures (65% of all company expenditures) for 1986 and later-stage investment dollars from venture capitalists (65% of venture capital dollars invested in high-tech industries). Later-stage investment provides capital for the expansion of new ventures that are already producing and marketing products. It includes bridge financing for new ventures, which are going public, as well as research and development partnerships which fund new product development by established firms. This suggested that these industries would represent a large pool from which to draw.

Data was collected from the IPO statements including SEC forms S-1. The use of IPO registration statements as a source of data is considered relatively reliable due to reporting requirements, SEC scrutiny, and sanctions for falsifications (Marino, Castaldi, & Dollinger.1989; Mosakowski, 1991; McGee et.al., 1995). Accounting data was recorded from financial statements and accompanying footnotes. Other data such as strategic alliance motivations, governance structures and business strategy required a content analysis of the SEC documents.

The variables collected from the IPO statements included continuous measures of firm performance and control measures for size, age and industry. Categorical measures of business strategy, alliance motivation and governance structure were coded from the descriptive portions of the IPO documents. Each of the variable measurements are discussed in Table 1 and provide a summary of the independent variables.

Average growth in sales was adopted as the measure of performance since it has been suggested that sustained growth in revenues is often indicative of overall new venture success(Feeser & Willard, 1990; McGee et.al., 1995). The following formula was used:

Average Sales Growth = ((Sales3 / Sales1 )1/3- 1)100

where Sales3 was the annual sales of the firm at the time of the IPO and the annual sales 3 years prior to the IPO respectively.

YR3Assets and Employee totals were used to control for size. YRINC and IPO date were

used to control for timing in entering the industry and age effects. The SIC code
variables controlled for industry effects.

TABLE 1

Description of Independent Variables

# of Alliances | The total number of alliances the firm had at time of IPO |

IPO | Date of IPO |

Empl | Number of employees at time of IPO |

YRINC | Year of Incorporation |

YR3ASSETs | Total assets at time of IPO |

SIC 36,38,50,73 | Dummy variable representing 5 SIC codes |

Scost, Smkt, Stech | Dummy variables representing a modified Porter typology |

Struc 1 | Equity alliances such as minority investment or joint venture |

Struc 2 | Nonequity alliances such as contracts and agreements |

MOTRD12 | Horizontal and vertical alliances based on resource dependency motives |

MOTRD3 | Reciprocal alliances based on resource dependency motives |

MOTTC | Transaction cost motives |

MOTSC | Strategic choice motives |

Moderated regression analysis will be used to test the various relationships of
governance structure and motivation on performance. This procedure will be used to test
the individual hypotheses since they suggest that the relationship between the independent
variables of motivations and structure and the independent variable new venture
performance was moderated by an additional independent variable; the interaction of
structure and alliance motive. The moderated regression model will include the interaction
term of structure and motivation. The basic form of the regression equation was:

Y = B_{0} + B_{1}X_{1} + B_{2}X_{2} + B_{3}X_{3}
+ ... + B_{p}X_{p} where

Y = Sales Growth

B_{0} = Intercept

B_{1} = Number of alliances

B_{2} = Age

B_{3} = Employee Total

B_{4} = Assets in most recent year

B_{5} = Year Incorporated

B_{6} = Date of IPO

B_{7} - B_{11} = Qualitative variables representing SIC code categories

Motive variables, structure variables and interaction terms were added depending on the
hypothesis being tested. The actual test of the hypotheses relationships used either a
t-test of an individual variables Beta or a partial F-test if more than one variable was
involved in the hypotheses.

In the analysis of the data, t-tests were used to determine if the interaction terms
were significant. According to Neter, Wasserman, & Kutner (1990), when interaction
effects are

present which include a qualitative variable, these effects can best be seen by graphing
the two lines. When the interaction term was significant, a graph has also been included.
To create these graphs, the mean values of the other variables in the equation not of
primary interest (control variables) were multiplied by their beta coefficients and summed
to create the Y-intercept. This procedure shows the difference in the slope and intercept
of the regression equation when interaction effects are present. To determine the
endpoints of the line, the variable's minimum amount was used and the high endpoint was
the mean plus two standard deviations. This ensured over 95% of the data was included in
the graph (McGee et. al., 1995).

After the models were developed a battery of diagnostic techniques were administered to ensure that the data was appropriate for multiple regression techniques. Specifically, tests were performed to ensure that the model was linear, error terms had constant variance, error terms were independent and normally distributed, and the model fits all but a few outliers.

Top of page | Chapter Listing | Return to 1997 Topical Index |

© 1997 Babson College All Rights Reserved

Last Updated 06/02/98