3.5 Una personalidad sin relieves: representación negativa del héroe
3.5.2 Atributos negativos. Más malo que el diablo
The results of testing hypotheses H1a–H7a, H8a and H8b, and H9a and H9 are given in Table 25. The first column shows the independent, control and moderated variables entered in the regression model. The next five columns show different regression models with standardized beta coefficients and their significance levels for each variable. For the hypothesized paths, the significance test is one-tailed. For the control variables, the significance test is two-tailed. All the independent variables were entered simultaneously in the regression models when testing predicted direct
relationships. Predicted moderated relationships were tested one at a time. Entering the base line model (only control variables) into the regression analysis gives the following results: R2 = .16, adjusted R2 = .10, F = 2.84, p ≤ .05.
Table 25. Results of regression analyses for product concept superiority.
Variables entered Model 1 Model 2 Model 3 Model 4 Model 5
Independent variables
Input control -.07 -.08 -.07 -.07 -.06
Front-end process formalization .20** .17** .21** .20** .22**
Outcome-based rewarding -.09 -.08 -.10 -.09 -.08
Strategic vision .23*** .21*** .23*** .23*** .24***
Informal communication -.04 -.09 -.04 -.04 -.04
Participative planning -.01 .00 -.01 .01 -.03
Intrinsic task motivation .14* .14** .14* .14* .12*
Control variables
Market uncertainty .11 .11 .12 .10 .09
Technology uncertainty .06 .11 .06 .07 .06
Size -.26** -.24** -.25** -.25** -.27**
R&D intensity .24** .28** .23** .23** .23**
Piece good industry .25* .29** .26** .25* .23*
Process-based production industry .27** .27** .27* .26* .26*
Objectives of the development project -.12 -.10 -.12 -.11 -.12
Definition of the front-end process -.06 -.04 -.07 -.06 -.05
Moderated variables
Front-end process formalization x
Market uncertainty .29***
Front-end process formalization x
Technology uncertainty .05
Outcome-based rewarding x Market
uncertainty .02
Outcome-based rewarding x Technology
uncertainty -.10 R2 .25 .32 .26 .25 .26 Adjusted R2 .16 .23 .15 .15 .16 F 2.64** 3.52*** 2.48** 2.46** 2.55** Sig. of F change .00*** .55 .79 .27 * p ≤ .10; ** p ≤ .05; *** p ≤ .01 Standard coefficient betas are shown
Dependent variable: product concept superiority
Hypotheses H1a–H7a were tested with Model 1 as in Table 25. Hypothesis H1a, that input control has a positive association with product concept superiority, is not supported. Non-significant negative association was found in testing with Model 1. IH2a hypothesized that front end process formalization is negatively associated with product concept superiority. This hypothesis can be rejected since significant positive association was found (beta = .20, p ≤ .05). In other words, the opposite hypothesis would have gained support in the light of this data.
Hypothesis H3a, which predicted a negative association between outcome-based rewarding and product concept superiority, is not supported either. The association is indeed negative but non-significant. Hypothesis H4a is supported; strategic vision is positively associated with product concept superiority (beta = .23, p ≤ .05). Hypothesis H5a stated that informal communication is positively associated with product concept superiority. This is not supported since a negative and non-significant relationship was found. Hypothesis H6a, which predicted a positive association between participative planning and product concept superiority, needs to be rejected. A negative and non-significant relationship was found in Model 1. Finally, hypothesis H7a stated a positive association between intrinsic task motivation and product concept superiority. This hypothesis gets marginal support and can be accepted with caution (beta = .14, p ≤ .1). In addition, four significant control variable effects can be found in Model 1: firm size has a significant negative effect on product concept superiority (beta = -.26 p ≤ .05); R&D intensity has a significant positive effect on product concept superiority (beta = .24, p ≤ .05); the piece good industry has a marginally significant positive effect on product concept superiority (beta = .25, p ≤ .1); and the process-based production industry a significant positive effect on product concept superiority (beta = .27, p ≤ .05).
R and F values, indicating overall explanatory power of the model and the adequacy of the model respectively, are reported at the end of Table 25. The adjusted R2 value, which takes into account the number of independent variables and the sample size, indicates reasonable explanatory power (R2 = .16), i.e. that the variables reasonably explain the variation in the product concept superiority variable. An F-test was used to test the significance of the overall regression model. The F value of Model 1 in Table 25 is statistically significant. Predictor value centering was applied to tackle problems
of multicollinearity. The highest VIF values were with the piece good industry (2.66) and the process-based production industry (2.81), therefore multicollinearity was not a problem in Model 1.
Hypotheses H8a and H8b were tested with Model 2 and Model 3 respectively, presented in Table 25. Model 1 in the table represents a comparison model where the influence of the moderating effect is compared. Hypothesis H8a stated that the more market uncertainty, the more negative the association between front end process formalization and product concept superiority. Model 2 does not suffer from high multicollinearity (highest VIF 2.81). Model 2 also has good explanatory power (R2 = .23). The standardized coefficient beta for the moderated variable (front end process formalization x market uncertainty) is .29 with strong statistical significance. Both the F value and change in F value (.00) have strong statistical significance. Since hypothesis H8a predicted more negative association, it needs to be rejected. Yet the opposite hypothesis would have been strongly supported.
Hypothesis H8b, which was tested with Model 3 as in Table 25, stated that the more technology uncertainty, the more negative the association between front end process formalization and product concept superiority. The VIF values indicate that multicollinearity was not a problem in Model 3 (highest VIF 2.81). The adjusted R2 value in turn indicates that Model 3 has reasonable explanatory power (R2 = .15). The standardized coefficient beta for the moderated variable (front end process formalization x technology uncertainty) is .05, which is not statistically significant. The F value of Model 3 is statistically significant. However, the F value change (compared to Model 1) is not (.55), and thus hypothesis H8b needs to be rejected.
Hypothesis H9a was tested with Model 4 and hypothesis H9b with Model 5 as in Table 25. Model 1 again represents a comparison model where the influence of the moderating effect is compared. Hypothesis H9a, that under high market uncertainty the association between outcome-based rewarding and product concept superiority is more negative, is not supported. The VIF values indicate that multicollinearity was not a problem in Model 4 (highest VIF 2.87). The adjusted R2 value indicates that Model 4 has reasonable explanatory power (R2 = .15). The standardized coefficient beta for the moderated variable (outcome-based rewarding x market uncertainty) is
.02, which is not statistically significant. The F value of Model 4 is statistically significant. However, the F value change is not statistically significant (.79). Thus hypothesis H9a can be rejected. Hypothesis H9b stated that the more technology uncertainty, the more negative the association between outcome-based rewarding and product concept superiority. The highest VIF value is 2.81, which again indicates that multicollinearity was not a problem in Model 5. Model 5 has reasonable explanatory power (R2 = .16). The standardized coefficient beta for the moderated variable (outcome-based rewarding x technology uncertainty) is -.10. This is not statistically significant. The F value of Model 5 is statistically significant but the F value change is not (.27). Based on this, hypothesis H9b should be rejected.