Chapter Listing | Return to 1997 Topical Index


NEW VENTURE RESEARCH: METHODOLOGICAL APPROACHES AND RESULTS

New venture research has used a variety of statistical data analysis techniques for examining the relationship between new venture strategy and new venture performance. Biggadike (1979), Miller and Camp (1985), MacMillan and Day (1987), and Roure and Keeley (1990) found that the assumptions underlying parametric statistical data analysis techniques were violated by the financial performance data in their new venture strategy research. Nonetheless, the vast majority of these and other studies have utilized parametric techniques for exploring relationships between strategy and new venture performance. Furthermore, with the exception of those listed above, the majority of such studies have often failed to explicitly check or fully discuss whether the assumptions underlying the appropriate usage of parametric techniques were satisfied by the data utilized in such studies. By contract, Sandberg (1986) and Kunkel (1991) are among the few studies which utilized nonparametric techniques to examine the impact of new venture strategy on venture performance.

Despite these differing approaches, the majority of these and other studies have found that new venture strategy is an important determinant of new venture performance.

Prior Findings vis--vis Scope

The vast majority of prior research has found that the scope of a new venture's strategic activities has a significant impact on new venture performance (e.g., Biggadike, 1979; Kunkel, 1991; Miller & Camp, 1985; Roure & Madique, 1986; Sandberg, 1986). Thus, we hypothesize:

H4: There will be differences in new venture performance based on the breadth of scope a new venture pursues.

Most, though not all, prior studies have also found that broad scope strategies are more successful than intermediate or narrow scope strategies. This study expects to find that broad
scope strategies will be the most successful, followed by intermediate scope strategies:

H4a: New ventures pursuing board scope strategies will have higher levels of economic performance than those ventures pursuing intermediate or narrow scope strategies.

H4b: New ventures pursuing intermediate scope strategies will have higher levels of economic performance than those ventures pursuing narrow scope strategies.

Prior Findings via-a-vis Strategic Weapons

There is substantial evidence that new venture performance is influenced by the strategic weapons a new venture emphasizes (e.g., Biggadike, 1979; Kunkel, 1991; Miller & Camp, 1985; Roure & Madique, 1986). Thus, we hypothesize:

H5: There will be differences in new venture performance based on the strategic weapons which a new venture emphasizes.

The majority of prior studies found that emphasizing both higher relative quality and lower relative prices resulted in superior new venture economic performance. This study expects to find that new ventures which emphasize higher relative quality and lower relative prices will achieve superior levels of performance:

H5a: New ventures pursuing strategies based on higher relative quality and lower relative price will achieve superior levels of economic performance.

Most, though not all, prior studies also indicate that ventures emphasizing higher relative quality are more successful than those emphasizing lower relative price. This study expects to find new ventures which pursue a strategy based on emphasizing higher relative quality will outperform those ventures which pursue a strategy emphasizing either lower or comparable relative prices:

H5b: New ventures pursuing strategies based on higher relative quality will have higher levels of new venture economic performance than those ventures which pursue a strategy based on either lower or comparable relative prices.

RESEARCH METHODS

Research Design

This research utilizes a longitudinal research design for a cross-section of new ventures in various industries. The primary reason for the use of this type of design was to determine the appropriateness and validity of using analogous parametric and nonparametric statistical data analysis techniques for examining relationships in new venture research.

Sample/Site Selection

A total of 455 initial public offering prospectuses were examined for firms which had
undertaken an IPO between 1980 and 1987. The new ventures were selected for inclusion into this study's sample based on the following criteria: (1) less than six years old at the time of the IPO; (2) independence from corporate sponsorship or control; and (3) still operating under the founding management team. The final sample size for this study consisted of 199 new ventures.

Variable Specification & Classification

The measures used to examine the economic performance of new ventures have varied widely in prior entrepreneurship research (Murphy, Trailer, & Hill, 1993). Murphy et. Al found that "change in sales" was the most commonly used measure of performance in 52 prior entrepreneurship research studies between 1987 and 1993.

At the dame time, there have been increasing calls for using the "creation of shareholder value" as a more global measure of economic performance for publicly held companies (Helfert, 1994; Rappaport, 1986). In addition, Robinson (1995) found that the shareholder value created measure of new venture economic performance was the most effective measure for accurately assessing the long-term economic value creation of such ventures. Thus, this research utilized (1) change in sales, and (2) shareholder value created as the two dependent measures of new venture performance.

Both of these ratio measures are based on three year periods. More specifically, the change is sales variable is based on data on data from each venture's first three full fiscal years following the venture's IPO. The shareholder value created variable is based on the change in stock price and cumulative dividends over each venture's first three full fiscal years (adjusted for stock splits). IPO price is not utilized due to the high variability of returns in initial trading prices.

The two strategy elements examined in this research are scope and strategic weapons, as they are the two frequently examined elements of new venture strategy in prior research.

Since these strategy assessments involved gathering and interpreting data contained in IPOs, inter-rater reliability checks were made of each strategy classification. Chi-square tests revealed a probability of less than 0.001 that the initial agreement level could have resulted from random chance. Moreover, all classifications were agreed upon after further review by the raters.

Data Gathering Methods

Financial information for assessing new venture performance was gathered for each venture's first four fiscal years, following the venture's IPO, from Standard and Poor's Compustat Services. The information for classifying new venture strategy was gathered from the IPO prospectuses of new ventures submitted to the SEC between 1980 and 1987.

Data Analysis Techniques for Assessing Distributional Characteristics

This study tested the assumption of normality through both the analysis of distributional statistics and plots. More specifically, this study calculated the: (1) Shapiro-Wilk statistic; (2) the skewness statistic,; and (3) the kurtosis statistic. In addition, this study examined: (1) normal probability plots; (2) histograms; and (3) box plots to assess the normality of the data utilized.

The two most frequently utilized procedures for testing equality of variances are the Bartlett and Hartley test (Neter, Wasserman, & Jutner, 1990). However, both of these tests procedures are based on strict assumptions of normality (Glaser, 1982; Layard, 1973). This research utilized the Cochran and Cox procedure, which does not depend on normal distributions, to test the null hypothesis of equal variances. It should be noted, however, this test allows only pairwise testing of the null hypothesis of equal variances. Thus, the results are presented for such pairwise comparison of variances for the sampled populations of the dependent variables.

The assumption of continuous distributions underlying the appropriate usage of the parametric and nonparametric statistical data analysis techniques utilized in this study was tested by rank ordering the variables of interest. The ranks of each of these variables were examined to assess the presence and number of ties present as recommended by Gibbons (1985).

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


1997 Babson College All Rights Reserved
Last Updated 03/03/98