Conjoint analysis is a strong tool for decision modeling research providing significant, structured insight into venture capitalists' decision criteria (Zacharakis & Meyer, 1996; Shepherd & Zacharakis, forthcoming). Theoretical basis for this study's use of conjoint analysis is information integration theory (Anderson, 1981; Louviere, 1988).
In developing conjoint profiles, levels were chosen to represent variation that typically occurs in the decision environment of venture capitalists, thereby maintaining believability and response validity. Venture capitalists evaluated a series of conjoint profiles which describe new ventures in terms of eight attributes. These eight attributes and their levels are detailed Table 1. Discussions with venture capitalists, accountants and academics provided face validity for the attributes and their levels. "Profitability: was defined as " net profit on sales, using a ten year time horizon". An eleven point scale was used and anchored by very high profitability and very low profitability. Ten years exceeds Biggadike (1976) and others choice of eight years to define a new venture. It appears reasonable to assume that this captures sustainable performance after entry.
Each of the eight attributes is varied at two levels in an orthogonal fractional factorial design consisting of 16 profiles (Hahn & Shapiro, 1966). The fractional factorial design allowed each main effect and selected two way interactions to be investigated. Two way interactions investigated are timing's interaction with each other factor, with the exception of industry related competence. Each of the 16 profiles were fully replicated. These 32 profiles were randomly assigned to avoid order effects, with a further practice case and 6 hold out cases used. The practice case familiarized respondents with the task and the 6 hold out cases were used to test the models' predictive ability. Therefore the experiment presented 39 profiles. Sixty six senior individual venture capitalists representing 47 Australian venture capital firms completed the survey.
To identify new venture profitability determinants that are statistically significant, an individual-subject analysis of variance was performed on the decision making of each venture capitalist. Although two or more attributes may significantly affect the decision process, it is unlikely that those attributes will be of equal importance. Therefore significance at the individual level of analysis is supplemented with a measure of relative importance. Hay's (1973) omega squared (w2), which is a measure of explained variance, was used to assess each respondent's relative importance of the eight attributes and selected 2-way interactions. Averages corresponding to all main effects and selected 2-way interactions were calculated.
Regression analysis was performed at the individual level to produce Beta coefficients for each factor and two-way interactions. The betas were aggregated across all venture capitalists and for inferential statistics the corresponding individual t-statistics were aggregated to a single Z score (Dechow, Huson & Sloan, 1994). To interpret significant contingent relationships at the aggregate level, interaction means were averaged across venture capitalists. Predictive ability of individual and aggregate decision making models were tested using a Pearson R correlation between the observed score on six hold out cases and a predicted score calculated by the individual decision making model(s). Sixteen replicated profiles were used in a test retest measure with the original 16 cases using Pearson R correlations to test the consistency of responses.
© 1997 Babson College All Rights Reserved
Last Updated 06/01/98