Building upon the cluster analysis results, we now attempt to validate our typology of crowdinvestors. We develop a set of hypotheses proposing how these distinct crowdinvestor types may react to project quality signals of entrepreneurs, project-related information reducing the degree of uncertainty and social influence by fellow investors when making their investment decision. We consider four relevant mechanisms identified by the entrepreneurial finance literature: human capital of project creators, third-party certifications, financial projections and observed behavior of other investors.
5.1 Hypotheses
Project quality may be signaled to potential investors by the
human capital of project creators. Generally, human capital encompasses an individual’s skills and knowledge acquired through education, on-the-job training and other types of experience which may increase one’s productivity at work (Becker
1964). Entrepreneurship researchers have investigated the influence of human capital on entrepreneurial success for over three decades (Unger et al.
2011). The importance of entrepreneurs’ human capital is further highlighted by prior studies that suggest managerial skills and experience are among the selection criteria that are most frequently used by venture capitalists (Baum and Silverman
2004; Zacharakis and Meyer
2000).
Similarly, crowdinvestors perceive the qualifications and education of project creators as a reliable signal of project quality, see Ahlers et al. (
2015), Li et al. (
2016), Nitani and Riding (
2017), and Piva and Rossi-Lamastra (
2018) for related empirical evidence. Therefore, we expect a general tendency of crowdinvestors to fund projects of creators with higher levels of human capital. Due to their distinct profiles, however, such quality signals may not be perceived as equally important across the different types of crowdinvestors. In particular, Crowd Enthusiasts and Sophisticated Investors tend to invest more in innovative projects than Casual Investors. Since innovative investments reflect a higher degree of uncertainty regarding both the probability and magnitude of economic returns, Crowd Enthusiasts and Sophisticated Investors might more strongly rely on signals of project quality related to human capital of the entrepreneurs than Casual Investors. Therefore, we hypothesize:
Besides signals provided by the project creators, potential investors may also look for
external certifications of project quality. In the context of CI, Ralcheva and Roosenboom (
2016) analyze three signaling mechanisms, in which a third party is involved. Ventures with business angel backing, grants, and intellectual property rights have significantly increased chances of success. Furthermore, Bapna (
2019) studies the effect of certification by expert intermediaries on the tendency to invest.
Again, taking investor heterogeneity into account, we expect such external certifications not to be of equal importance to all crowdinvestor types. We argue that entrepreneurs’ external certifications about project quality are relatively more important to Casual Investors. Sophisticated Investors tend to be experienced and more knowledgeable regarding reliable indicators of future project success (e.g., business plan, patents held, industry outlook). They know how to gather this information elsewhere in case it is not provided by the project creators, which makes them less likely to rely on third-party certifications of project quality. Crowd Enthusiasts’ are less concerned with making a profit; instead, they have the desire to support project creator and cause. They tend to invest in a project if the business idea convinces them. Whether projects provide external certifications then seems to be of secondary relevance for them. Thus, we propose the following hypothesis:
Informational asymmetries between investor and entrepreneur are a key challenge in CI (Agrawal et al.
2014). If the degree of information asymmetry is too high and the supply of projects of inferior quality is relatively higher than the supply of good quality projects, markets of venture financing, such as CI, might even collapse. In order to attract funding, entrepreneurs need to provide reliable information that helps reduce the degree of uncertainty regarding the quality of the start-up project (Ireland et al.
2003). For instance, disclosing
financial projections facilitates potential investors forming of expectations of the start-up’s future returns, detailing the risks and opportunities of the investment. See Michael (
2009) for similar evidence from the franchise sector and Ahlers et al. (
2015) who find a positive link between the disclosure of financial information (like roadmaps) and the funding success of CI campaigns.
Following previous research, there should also be a positive relationship between the provision of financial projections and the likelihood that a project receives investments in our data. We argue however that the assessment of financial projections disclosed by entrepreneurs differs across crowdinvestor types. Crowd Enthusiasts’ portfolios contain the highest share of innovative projects. In order to make their investment choices, they may focus on a project’s idea, its vision, while they may put less emphasis on expected economic outcomes. By contrast, Casual Investors are mainly motivated by financial interests. They tend to keep the probability of financial losses small by diversifying their investments among less innovative (and therefore less risky) projects. Thus, Casual Investors are likely to strongly rely on information directly related to possible future outcomes most strongly. This assumption is in line with other research demonstrating that disclosed financial information is particularly appealing to risk-averse investors (Epstein and Schneider
2008). Sophisticated Investors tend to have a professional approach to investing. While they are open to risky investments in innovative projects, they also consider provided financial projections when they decide on an investment. Accordingly, we hypothesize:
Recent research on CF and CI stresses the role of
social influence for the funder’s decision to invest. The availability of information concerning the timing and the number of investments are found to be critical factors affecting the investment behavior of subsequent funders. In this context, empirical findings have been related to economic models of herding behavior (Herzenstein et al.
2011; Lee and Lee
2012; Zhang and Liu
2012; Colombo et al.
2015; Crosetto and Regner
2018; Vismara
2018) or substitution (i.e., crowding-out; Burtch et al.
2013). Lin et al. (
2014) find that whether crowding-out or herding takes place depends on the backer’s motivation to engage in CF. They show that with an increasing number of backers the probability decreases that a project is chosen by users with altruistic motives to participate in CF. Likewise, altruists do not seem to imitate the investment decision of experienced investors. On the contrary, risk-averse and reward-driven crowdfunders are found to imitate the decision of others, resulting in general herding behavior.
Based on these findings, we expect different responses to observed peer behavior along the crowdinvestor typology. In particular, among the identified crowdinvestor types, we expect Crowd Enthusiasts to most likely engage in herding behavior. Their investment portfolio accounts for the highest share of innovative projects which are characterized by a particularly high degree of uncertainty. Moreover, Crowd Enthusiasts should be positively influenced by the project creators’ expressed belief in the quality of the business idea. In the light of Crowd Enthusiasts’ strong sense of community, this positive feeling may be reinforced when peers invest in the same project, eventually inducing the Crowd Enthusiasts to invest as well. In this respect, Casual Investors differ from Crowd Enthusiasts. Casual Investors pursue an investment strategy of risk-diversification among less innovative projects. With their investment decisions, they seem to rely on less ambiguous information which may not include the decision of other investors. Finally, Sophisticated Investors can be characterized as being more experienced and knowledgeable compared to Casual Investors and Crowd Enthusiasts. They tend to undertake thorough due diligence and primarily base their investment decisions on own knowledge and expertise. Therefore, we expect that investments of Sophisticated Investors are rather not influenced by others’ investment decisions but instead are used for guidance, predominantly by Crowd Enthusiasts. The following hypothesis applies:
5.2 Estimation strategy
To test our hypotheses, we adopt a choice model (using probit regression) and investigate how crowdinvestors select a start-up project for funding from a set of alternatives. We implement a binary choice task that indicates whether or not the crowdinvestor decides to fund a particular project. Given that at any point in time there are multiple projects available for funding, we have to construct a temporal choice set for each crowdinvestor. Hence, for each instance when an investment is being made we identify the possible set of alternative projects, including the eventual choice.
2 We let
Yit represent the binary choice task of funding a particular project out of the temporal set of alternative projects
i, which is available at time
t representing the day of investment.
3Yit returns 1 if crowdinvestor
j decides to fund the particular project and 0 otherwise. It is important to note though that the binary choice task
Yi may repeat over time. For example, the same investment alternatives are presented to different crowdinvestors if they access Companisto at the same time. Consequently, their investment decisions are likely to be correlated. This could potentially lead to biased estimates. To correct for the temporal correlation of outcomes in our data, we used the generalized estimation equation (GEE; Zeger et al.
1988) approach. Essentially, GEE is an extension of the generalized linear model (GLM) but allows for the use of a correlation matrix structure which takes into account the lack of independence of observations. The main advantage of GEE resides in the unbiased estimation of population-averaged (marginal) regression coefficients despite possible misspecification of the correlation structure (Cui
2007). For our analysis, we used the commonly specified exchangeable correlation as the working correlation matrix.
As we are interested in the determinants of crowdinvestors’ investment decisions, we let μit = E(Yit) denote the marginal expectation of whether a start-up project of the temporal choice set i is funded in time t. The marginal expectation of investment depends on the vector of explanatory variables (project, creator, and crowdinvestor characteristics), Xij, through the probit link function Probit (μit) = Xijβ + vi, where vi represents the stochastic term related to the choice set i which is assumed to be normally distributed.
Finally, common model selection criteria like Akaike’s (
1974) information criterion (AIC) cannot be applied as GEE is a quasi-likelihood-based method (Cui
2007). Pan (
2001) proposed the independence model criterion (QIC) which is equivalent to the AIC in evaluating the goodness-of-fit of competing models.
5.4 Results
Drawing upon the three crowdinvestor profiles that we identified in the cluster analysis (see Sect.
4.2), we now study how the evaluation of investment alternatives differs across investor types. In particular, we are interested in the individual responses of Casual Investors, Crowd Enthusiasts, and Sophisticated Investors to various signals of project quality and social influence when making investment choices. This analysis contributes to a more detailed understanding of the heterogeneity of crowdinvestors and their individual behavior on CI platforms.
For each investment made, we identify the possible set of alternative projects to invest in at that point of time. This results in a dataset with 67,833 investment alternatives. Table
3 provides descriptive statistics, and Table
4 displays the correlation among the variables used for the regression analysis.
Table 3
Descriptive statistics for the variables used in the regression analysis
PhD | 0.12 | 0.32 | 0 | 0.00 | 1 |
Ties | 0.44 | 0.50 | 0 | 0.00 | 1 |
Financial forecast | 0.18 | 0.38 | 0 | 0.00 | 1 |
Sophisticated investors_before | 19.19 | 15.07 | 0 | 20.00 | 93 |
Days after project start | 45.71 | 30.63 | 0 | 45.00 | 100 |
Team members | 4.24 | 1.49 | 2 | 4.00 | 8 |
Awards | 0.15 | 0.36 | 0 | 0.00 | 1 |
Updates | 0.29 | 0.45 | 0 | 0.00 | 1 |
Table 4
Correlation matrix for the variables used in the regression analysis
[1] | PhD | 1 | | | | | | |
[2] | Ties | − 0.05 | 1 | | | | | |
[3] | Financial forecast | − 0.17 | 0.20 | 1 | | | | |
[4] | Sophisticated investors_before | 0.17 | − 0.25 | 0.15 | 1 | | | |
[5] | Days after project start | − 0.09 | − 0.02 | − 0.11 | 0.12 | 1 | | |
[6] | Team members | 0.09 | 0.17 | 0.20 | 0.43 | − 0.05 | 1 | |
[7] | Awards | 0.21 | 0.49 | 0.08 | − 0.09 | − 0.02 | 0.34 | 1 |
[8] | Updates | 0.03 | 0.22 | 0.13 | − 0.22 | − 0.34 | 0.03 | 0.07 |
Table
5 reports the results of the GEE estimation for the prediction of investment decisions of crowdinvestors. Model 1 is a baseline specification without considering our cluster variables. It can be regarded as a test whether we find support for the impact of human capital, third-party certifications, financial projections, and observed behavior of other investors in our sample without distinguishing investor types. The coefficients for PhD (
β = 0.491,
p < 0.001) and Financial forecast (
β = 0.408,
p < 0.001) show up positive and significant. Crowdinvestors, in general, seem to take into consideration the educational attainment of project creators and the availability of financial forecasts as project quality signal when selecting among investment alternatives. In contrast, the likelihood to attract an investment seems to diminish if a start-up project has pre-existing relationships with traditional investors (Ties;
β = − 0.0358,
p < 0.05). The negative effect of Sophisticated Investors_before (
β = − 0.0164,
p < 0.001) further suggests that crowdinvestors, in general, do not necessarily follow the investment decision of the group of Sophisticated Investors.
Table 5
Generalized estimation equation regressions for the prediction of investment decisions
Project quality signals |
PhD | 0.491*** (24.22) | 0.329*** (13.68) | 0.453*** (22.04) | 0.489*** (24.16) | 0.365*** (18.15) |
Ties | − 0.036* (− 2.49) | − 0.046** (− 3.24) | 0.071*** (4.36) | − 0.029* (− 2.03) | − 0.033* (− 2.32) |
Financial forecast | 0.408*** (26.58) | 0.447*** (28.95) | 0.423*** (27.69) | 0.474*** (26.02) | 0.495*** (31.32) |
Social influence of investor peers |
Sophisticated investors_before | − 0.016*** (− 31.55) | − 0.018*** (− 34.08) | − 0.017*** (− 33.09) | − 0.017*** (− 31.99) | − 0.033*** (− 42.47) |
Control variables |
Days after project start | − 0.013*** (− 52.00) | − 0.012*** (− 49.34) | − 0.013*** (− 52.00) | − 0.013*** (− 52.06) | − 0.014*** (− 54.49) |
Team members | 0.176*** (37.44) | 0.168*** (35.22) | 0.190*** (39.44) | 0.175*** (37.13) | 0.158*** (34.42) |
Awards | − 0.372*** (− 16.30) | − 0.331*** (− 14.61) | − 0.350*** (− 15.25) | − 0.370*** (− 16.19) | − 0.335*** (− 15.00) |
Updates | 0.612*** (45.37) | 0.609*** (44.92) | 0.605*** (44.66) | 0.611*** (45.46) | 0.572*** (41.99) |
Investor types |
Crowd enthusiasts | 0.068*** (6.71) | 0.003 (0.23) | 0.271*** (17.94) | 0.116*** (7.90) | − 0.687*** (− 35.95) |
Sophisticated investors | 0.010 (0.48) | − 0.026 (− 1.03) | 0.150*** (4.15) | 0.000 (0.01) | − 0.397*** (− 9.87) |
Hypothesis 1: |
Crowd enthusiasts × PhD | | 0.496*** (11.44) | |
Sophisticated investors × PhD | | 0.272** (2.77) | |
Hypothesis 2: |
Crowd enthusiasts × Ties | | | − 0.437*** (− 14.67) | |
Sophisticated investors × Ties | | | − 0.242*** (− 4.01) | |
Hypothesis 3: |
Crowd enthusiasts × Financial forecast | | | | − 0.185*** (− 6.24) | |
Sophisticated investors × Financial forecast | | | | 0.020 (0.31) | |
Hypothesis 4: |
Crowd enthusiasts × Sophisticated investors_before | | | | | 0.034*** (38.10) |
Sophisticated investors × Sophisticated investors_before | | | | | 0.023*** (13.07) |
Constant | − 1.086*** (− 53.96) | − 1.041*** (− 50.87) | − 1.184*** (− 55.82) | − 1.088*** (− 54.17) | − 0.697*** (− 30.75) |
QIC | 57,461.488 | 57,189.150 | 57,330.856 | 57,277.032 | 55,069.290 |
As regards the control variables, the team size (Team members; β = 0.176, p < 0.001) as well as recently posted updates (Updates; β = 0.612, p < 0.001) significantly and positively predict funding of a project. The number of days since the funding campaign has started (Days after project start; β = − 0.0129, p < 0.001) and awards won (Awards; β = − 0.372, p < 0.001) seem to have a negative effect on the investment choice of crowdinvestors. Finally, Crowd Enthusiasts are relatively more likely to invest than Casual Investors, the reference group (Crowd Enthusiasts; β = 0.0678, p < 0.001), while there is no such difference between Sophisticated Investors and Casual Investors (Sophisticated Investors; β = 0.0102, n. s.).
Next, models 2 to 5 in Table
5 focus on differences in the prediction of investment decisions across the distinct types of crowdinvestors in order to test hypotheses 1 to 4. Therefore, we include interaction effects between the cluster variables Casual Investors, Crowd Enthusiasts, and Sophisticated Investors on the one hand and the main explanatory variables PhD (model 2), Ties (model 3), Financial forecast (model 4), and Sophisticated Investors_before (model 5) on the other hand. The coefficients for the interaction effects are to be interpreted relative to the baseline group Casual Investors. Hence, they indicate how much larger or smaller the parameter estimates for the respective explanatory variables are for Crowd Enthusiasts and Sophisticated Investors relative to Casual Investors. Table
6 further allows for a more detailed analysis of the coefficients for the interaction effects. Here, corresponding to models 2 to 5 in Table
5, the effect sizes for the parameter estimates of the main explanatory variables are reported. The effect size (see last column in Table
6) represents the percentage change in the predicted probability of a start-up project being chosen by an investor type before and after the value for the respective explanatory variable increased by one standard deviation (for continuous scales) or changed from 0 (base level) to 1 (for dummy variables), while keeping all other variables at their mean (for continuous scales) or median value (for dummy variables).
Table 6
Predicted probabilities and effect size of coefficients
2 | PhD | Casual investors | 0.329** | 0.11 | 0.18 | 68.19% |
Crowd enthusiasts | 0.825** | 0.11 | 0.34 | 213.76% |
Sophisticated investors | 0.601* | 0.10 | 0.25 | 145.44% |
3 | Ties | Casual investors | 0.071** | 0.10 | 0.11 | 13.31% |
Crowd enthusiasts | − 0.366** | 0.15 | 0.08 | − 46.40% |
Sophisticated investors | − 0.171** | 0.12 | 0.09 | − 25.48% |
4 | Financial forecast | Casual investors | 0.474** | 0.10 | 0.22 | 107.99% |
Crowd enthusiasts | 0.289** | 0.13 | 0.20 | 55.48% |
Sophisticated investors | 0.474** | 0.10 | 0.22 | 113.75% |
5 | Sophisticated investors_before | Casual investors | − 0.033** | 0.10 | 0.04 | − 61.84% |
Crowd enthusiasts | 0.002** | 0.10 | 0.10 | 4.75% |
Sophisticated investors | − 0.009** | 0.11 | 0.09 | − 21.87% |
Turning to Hypothesis 1 and the proposed signaling effect of project creators’ human capital on the likelihood to receive funding, model 2 in Table
5 provides positive and significant coefficients for Casual Investors (
β = 0.329,
p < 0.001), Crowd Enthusiasts (
β = 0.825 (0.329 + 0.496),
p < 0.001), and Sophisticated Investors (
β = 0.601 (0.329 + 0.272),
p < 0.01). Accordingly, it seems that higher levels of creator human capital are perceived as reliable project quality signal by all three crowdinvestor types. Interestingly, model 2 in Table
6 reveals significant differences among the three groups. Accordingly, a change in the variable PhD from base level to 1 increases the likelihood that a project receives funding from the group of Crowd Enthusiasts by 213.76%. For Sophisticated Investors, a project creator holding a PhD increases the likelihood of investment by 145.44%, while the likelihood that Casual Investors choose a project is still 68.19% higher. In relative terms, however, Casual Investors are least likely to respond to higher levels of human capital in their investment decision. Therefore, hypothesis 1 is supported by the data. Moreover, our findings complement related results from other CI platforms stressing the importance of education (Ahlers et al.
2015; Nitani and Riding
2017; Piva and Rossi-Lamastra
2018).
Testing hypothesis 2, model 3 takes account of the signaling effect of third-party endorsement by venture capitalists or business angels (Ties). We find significant interaction effects for Crowd Enthusiasts (
β = − 0.437,
p < 0.001) and Sophisticated Investors (
β = − 0.242,
p < 0.001). Consequently, the coefficients for Crowd Enthusiasts and Sophisticated Investors are decreasing to
β = − 0.366 (
p < 0.001) and
β = − 0.171 (
p < 0.001), respectively. Crowd Enthusiasts and Sophisticated Investors seem to be less likely to invest in start-up projects when project creators possess preexisting ties to traditional investors. This is in strong contrast to Casual Investors (
β = 0.071,
p < 0.001). Table
6 reveals that this group is 13.31% more likely to invest in a project if there are existing relationships to external investors before the start of the funding campaign, while for Crowd Enthusiasts and Sophisticated Investors the likelihood to invest decreases by 46.4% and 25.48%, respectively. Overall, we find support for our hypothesis 2.
In model 4, we investigate differences in investment activities of the investor types in response to the provision of financial forecasts (hypothesis 3). Table
5 indicates that Casual Investors and Sophisticated Investors share the same parameter estimate for Financial forecast (
β = 0.474,
p < 0.001) since the interaction effect for Sophisticated Investors turns out not significant (
β = 0.0204,
n.
s.). Thus, Casual Investors and Sophisticated Investors do not significantly differ in their response to provided financial forecasts. Table
6 further reveals that the likelihood to invest in projects increases for Casual Investors by 107.99% and for Sophisticated Investors by 113.75%, respectively, if financial forecasts are provided. Moreover, examining the effect size of the coefficient for Crowd Enthusiasts (
β = 0.289,
p < 0.001), we find that a change of the dummy variable Financial forecast from base level to 1 increases the probability to invest in the focal project by 55.48%. Taking these results together, we find strong support for the importance of financial information as project quality signal to reduce the degree of information asymmetry between project creators and investors. In more detail, though, Crowd Enthusiasts seem to be least likely to respond to this quality signal relative to the other two investor types. Hence, we find support for hypothesis 3.
Finally, model 5 tests for different responses to observed peer behavior along the crowdinvestor typology (as proposed in hypothesis 4). Table
5 considers the interaction effects of the cluster membership variables and the explanatory variable Sophisticated Investors_before. The coefficient for Casual Investors is negative and statistically significant (
β = − 0.0326,
p < 0.001), revealing that Casual Investors are less likely to follow more sophisticated peers in their investment decision. Assessing the economic magnitude of the coefficient in Table
6, a one standard deviation increase in the number of Sophisticated Investors that have already invested in a project (i.e., an increase from 20 to 35 Sophisticated Investors) leads to a decrease in the likelihood that Casual Investors will choose the same project by 61.84%. Similarly, for Sophisticated Investors, the likelihood to invest decreases by 21.87% (
β = − 0.009,
p < 0.001) if the variable Sophisticated Investors_before is increased by one standard deviation. In contrast, the coefficient of the interaction effect for Crowd Enthusiasts is positive and statistically significant (see Table
5;
β = 0.034,
p < 0.001) and even large enough to turn the negative coefficient of the baseline group from
β = − 0.033 (
p < 0.001) to
β = 0.002 (
p < 0.001). In economic terms, Table
6 suggests that a one standard deviation increase of Sophisticated Investors_before leads to an increase in the likelihood that Crowd Enthusiasts will choose the project by 4.75%. Supporting hypothesis 4, Crowd Enthusiasts seem to be more likely to follow the investment decision of their more experienced and knowledgeable peers than the groups of Sophisticated Investors and Casual Investors, even though the economic magnitude of this response is rather small. Another important caveat for this result is the
first-come,
first-serve mechanism adopted by Companisto. As the funding mechanism induces quick investments at the early stages of a funding campaign, it may even impede herding behavior. Investors might feel inclined to invest early, forgoing the possibility to wait and observe the decision of others first.