An empirical investigation of determinants of effectual and causal decision logics in online and high-tech start-up firms

According to Milliken (1987), event uncertainty relates to an “individual’s inability to predict what the impact of environmental events or changes will be on his/her organization.” Response uncertainty “is defined as a lack of knowledge of response options and/or an inability to predict the likely consequences of a response choice” (p. 137).

Although extant entrepreneurship literature recognizes new venture stages, studies are ambiguous regarding what constitutes a certain stage and how many exist. Some use a dichotomy (e.g., new vs. established, Brinckmann et al. 2010); others differentiate more stages (e.g., discovery, emergence, young, established; Klyver and Terjesen 2007). Herein, we distinguish early stage, expansion, and later stages as experienced by our informants.

For more interviewee quotes in support of our findings, see Web Appendix 3.

Little et al. (1999) suggest keeping such indicators in a model to prioritize conceptual concerns over maximizing indicator reliabilities.

In order to check whether a more parsimonious model could also explain the phenomena of interest in a similar way (see, for example, Edwards 2001), we also computed a structural model in which we substituted the four dimensions of effectuation by effectuation as a second-order construct. According to Jarvis et al.’s (2003) decision rules for reflective versus formative construct measurement specification, effectuation must be specified as a formative higher-order construct (see also Chandler et al. 2011). We, therefore, computed factor scores for each effectuation dimension, using our CFA results and Thurstone’s regression technique (Estabrook and Neale 2013). We, then, computed an effectuation index by multiplying those factor scores, in line with prior research (e.g., Homburg et al. 2002). The resulting structural model, with one effectuation index instead of four effectuation dimensions, displayed the following model fit: χ²(d.f.) = 397(199), p < 0.001, χ²/d.f. = 1.99, RMSEA = 0.05, SRMR = 0.04, NFI = 0.93, NNFI = 0.96, and CFI = 0.97. These values are better than those of our full model but can mostly be explained by model parsimony. While relationships between antecedents and causation differed only very slightly from the original model (average difference in path coefficients = 0.005), there was only one significant relationship between uncertainty and effectuation (γ = .184, p < 0.001), but no significant relationships between entrepreneurial experience (γ = 0.083, ns), management experience (γ = − 0.056, ns), or investor influence (γ = − 0.037, ns) and effectuation, respectively. These results indicate that effectuation is a more valuable explanatory concept if it is understood and measured as a multidimensional composite of its dimensions rather than as a unidimensional higher-order construct.

Because of sub-sample size requirements, we combined three different answers to our venture stage question into the second group, called “later stages.” It comprised n = 147 respondents who saw their venture in “expansion,” n = 23 in “later stage,” and n = 18 who had answered “I do not know.” We included the latter cases for theoretical and empirical reasons: on a data-collection level, we assumed that the most probable reason for respondents to tick “I do not know” regarding development phase was that they were somewhat beyond the “early phase,” but they could not tell how far beyond. To further base this decision on the data properties, we also compared the age-wise distribution of those 18 cases with the 170 cases from the “later stages,” because development stage and age of the venture should be related. The age distributions of “later stages” and “I do not know” were very similar and also quite different from the distribution of the “early stages.”