METHODS

Sample

The method used was a mail survey of NVTs in the United States that received first-round funding from VCs. A list of firms was identified from the Venture Capital Journal of firms receiving first round funding from VCs during the period 1985-1990. Most of these firms were high technogy ventures. A number of popular business directories were examined in an attempt to find addresses for firms with names matching those in listed in the source. Surveys were sent out to all listed firms for which a mailing address could be identified. Surveys were addressed to the president of the new venture with instructions that the survey should be completed by a key member of the NVT that was knowledgeable about the details of the first round funding contract. The survey approach utilized Dillman’s (1978) total design methodology. The first survey was sent out, then a postcard followed two weeks later, and a second survey went out to nonrespondents one month after the first survey. 837 firms appeared to have current valid addresses based on three rounds of mailouts. Replies were received from 235 of these firms for a 28% response rate. Of 235 responding firms, 29 did not report having received first round VC funding. 144 of the remaining 206 reported receiving second round VC funding. Of this group, there were 65-90 responding venture firms with sufficient performance data (65 for income and 90 for sales) to be considered for inclusion in final modeling. Several extreme outliers were detected across some of the primary financial and funding variables used in the analysis. The following bounds were utilized to qualify the final sample to be examined. The venture had to have sales less than $12.0 million at the end of the first year of VC funding and a gain in sales between the first and second year of no more than $12.0 million. It had to have income at the end of the first or second year of VC funding between a gain of $6.0 million and a loss of $6.0 million with change in income from the first to the second year not to exceed a gain of $6.0 million or not less than a loss of $6.0 million. The change in the average amount of funding per VC firm could not exceed $8,000,000 in moving from the first to the second round. Visible breaks in the distribution of each variable were apparent beyond these values. Responses followed the date of first round funding by from 1-5 years.

Measures

Dollar Change in Investment Level per VC Firm: This represents change in risk perception of the average VC firm between the first round of VC investment and the second as measured by average change in investment between the two rounds (in dollars). It combines responses to four survey questions including: amounts of total VC funding in the first and second round and number of participating VCs in first and second rounds. The measure is average investment per VC firm in the first round subtracted from average investment per VC firm in the second round. The specific identities of participating VCs are ignored. The measure is not restricted to average change in investment levels of only those VCs that participated in first round VC funding.

Two Year Performance Means: Two variables were used to measure differing types of performance. The two year sales mean variable was calculated by identifying all responding new venture firms that received first round funding between 1985 and 1989. Sales reported for the fiscal year in which they reported first round funding was averaged with the sales reported in the following fiscal year (1990 fiscal data was latest collected). The two year income mean variable was created in the same way. Both of these variables were measured in millions of dollars. Ideally, performance levels over all years of the first round could be averaged into a measure of overall performance available to consider when making the second round funding decision. Although the date of first round funding was asked in the survey, the timing of subsequent rounds of funding was not. However, years since first round funding at the time of survey completion and number of rounds of funding completed were available for most firms. This information suggested a mean interval of just under two years between funding rounds. Financials from the year of first round funding followed the first round deal by a few weeks up to almost 12 months. Assessing performance levels over these two years allows a good picture of the early performance of the new venture from the first round deal and before second round funding is likely to have occurred. If the date of second round funding were known and sufficient performance data available, the average performance level over all relevant years could be included. In some cases this might involve one to three additional years of financial information. Inclusion of measures of rounds of funding and years since first round funding was received did not affect results of our analyses.

Procedural Justice: Two variables were used to examine differing views of the important procedural elements in the VC-NVT relationship. A three-item measure labeled global justice was used to examine the overall perception of the NVT of the fairness of the VCs in their relationship with the NVT. A five part Likert-type scale ranging from strongly disagree (1) to disagree (2) to neutral (3) to agree (4) to strongly agree (5) was used for each response. The statements used included: our venture capital firms (VCFIRMS) force us to accept their business views (reverse coded), our VCFIRMS are willing to compromise with us, and our VCFIRMS have hampered development of new ideas (reverse coded). This scale had a coefficient alpha of .7148. These three items are intended to mirror some of the complex breadth of the procedural justice construct discussed in the organizational justice literature as reviewed by Gilliland (1993). Forcing acceptance of views reflects the procedural justice concept of two way communication (an interpersonal treatment characteristic). Willingness to compromise focuses on the procedural justice concept of representativeness (a formal characteristic). Hindering the development of new ideas refers to the procedural justice dimension of justification (an explanation characteristic). These aspects are combined into a global measure of procedural justice. Gilliland (1993) suggests that all three types of justice rules are important components of the overall fairness perception. The second variable was a single-item measure of mutuality-based justice, a much more specific approach to understanding how justice is being evaluated by the NVT in conflict situations. The same five part Likert scale was used for the following statement: our VCs emphasize mutual benefits in resolving conflicts with us. We distinguish mutuality-based justice perceptions from global justice perceptions in order to focus more specifically on how fairly the VCs are perceived to balance interests in high conflict situations. Ideally both measures would have been taken immediately prior to the onset of negotiations between the VCs and the NVT concerning the terms of the second round deal to make sure that there was no corruption of the justice perceptions of the second round deal or actions or rounds following thereafter. If it is assumed the second round occurred about 18 months after the first round, the average response occurred about within 6-18 months after completion of the second round. Despite the imperfect timing of the measures, we employ these here based on our assumption that the initial framing effects of the first round deal and the first extensive interactions between the VCs and the NVT which occur during the first round largely determine final NVT procedural justice perceptions.

Reactions of the NVT to Assistance from VCs: Reactions to three different types of assistance from VCs were gathered through responses to five part Likert type scales ranging from strongly disagree (1) to disagree (2) to neutral (3) to agree (4) to strongly agree (5). A three-item scale labeled strategic assistance was developed from reactions to the following statements: our venture capital firms (VCFIRMS) provide excellent financial advice; our VCFIRMS give us sound business advice; and our VCFIRMS have provided sound management advice. The coefficient alpha for this scale was .8817. A four-item scale labeled operational assistance was developed from reactions to the following statements: our VCFIRMS introduce us to customers; our VCFIRMS have assisted us with suppliers; our VCFIRMS help us recruit potential employees; and our VCFIRMS tell us about competition in our industry. The coefficient alpha for this scale was .7235. A single-item variable labeled conflict recognition assistance was developed from the reaction to the following statement: our VCFIRMS bring conflict out into the open.

Control Variables: Three variables were used as controls in the regression analyses including: the natural log transformation of the amount of total VC funding in the first round; the two year employee count mean, and the change in the proportion of VC funding provided by the largest investor. These controlled for prefunding performance, size, and change in syndicate structure.

Data Analysis

Aiken and West (1991) describe in detail a recommended procedure for testing three-way interactions using moderated regression analysis. The procedure involves first testing for the significance of the change in the R2 statistic from adding the interactions effects to the first-order effects of theoretical variables of interest and the effects of control variables. This test is designed to preclude use of models in which higher-order terms do not contribute significantly to understanding of the dependent variable. Testing using individual coefficients on interaction terms therefore proceeds only on higher-order models that meet this initial contribution threshold. Where the change in R2 test is nonsignificant, the moderation hypothesis can be rejected without looking at the partial t test on the interaction parameter in the high-order model. In such cases all individual coefficient estimates for the higher-order model must be ignored, even if individually significant. Testing a higher-order interaction requires entry of all lower-order component effects. Therefore, for testing a 3-way interaction, all three 2-way interactions must also be added to the original model involving the three component variables of theoretical interest and any control variables. The component variables are centered by subtracting mean values for each sample from the actual values for that variable for each observation in the sample. To interpret a significant test of a hypothesized 3-way interaction term coefficient, the relationship must be graphed to understand the form of the relationship between the component variables. Aiken and West (1991) recommend using two graphs with two lines per graph to show continuous relationships between one predictor variable (the relevant two year performance mean measure) and the dependent variable at levels one standard deviation above and one standard deviation below the mean on the other two predictor variables. These levels are labeled high and low in the graphs in Figure1. (Graphic unavailable)

Top of Page
Previous Page Main Page Next Page
Return to 1996 Topical Index


1997 Babson College All Rights Reserved
Last Updated 4/5/97 by Cheryl Ann Lopez

To sign-up for the Center for Entrepreneurial Studies' publication lists,
please register with the
Entrepreneurship WebTeam.