30.08.2013 Views

pdf - Nyenrode Business Universiteit

pdf - Nyenrode Business Universiteit

pdf - Nyenrode Business Universiteit

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

2.7. ANALYSES FOR FOR-PROFIT BUS (APPENDIX B) 57<br />

2.7 Analyses for For-Profit BUs (Appendix B)<br />

This appendix presents robustness checks of the models presented in section 2.4.3.1 (page<br />

42) and 2.4.3.2 (page 43). The robustness checks analyse the base models using a for-profit<br />

subsample of the data. Although I do not have a priori different theoretical expectations<br />

for RPE antecedents among not-for-profit business units, it is possible that the empirical<br />

results are qualitatively or quantitatively different for not-for-profit business units from forprofit<br />

sectors. To control for the influence of not-for-profit business units on my results,<br />

I estimate both the RPE-Use and the RPE-based-Targets models with for-profit business<br />

units only.<br />

However, leaving out not-for-profit business units has one important drawback; it reduces<br />

the sample size. This is potentially harmful for the significance of the models, because the<br />

models contain interaction effects that require larger sample sizes. The testing of hypotheses<br />

using interaction effects with relatively small datasets can easily result in statistical null<br />

findings (see: Aguinis 1995:1142, Cohen et al. 2003:297). This risk is especially relevant<br />

in the case of the RPE-Use model, since the full-sample model (table 2.11 on page 42) already<br />

has limited explanatory power (ANOVA F-statistic = 1.718, p = 0.063, R 2 = 0.075,<br />

adjusted R 2 = 0.031). As the results of the for-profit analyses show, the RPE-Use model<br />

is insignificant. The ANOVA F-statistic of this model, presented in table 2.19 - panel A is<br />

1.461 (p = 0.148). As argued above, this may be the result of the decreased sample size:<br />

the full-sample model has 267 included observations, the subsample analysis contains 224<br />

observations. Due to the insignificance of the model, no inferences can be drawn from this<br />

analysis.<br />

Unlike the RPE-Use model, the for-profit robustness analysis of the RPE-based-Targets<br />

model is significant (F-statistic = 1.884, p = 0.044). The results of this analysis are shown<br />

in table 2.19 - panel B. Similar to the RPE-Use model, the RPE-based-Target model has<br />

a weaker F and R 2 -statistic than the full-sample analysis. Again, this may be the result of<br />

the reduced number of observations. Whereas the original full-sample RPE-based-Targets<br />

model (table 2.13 on page 45) contains 242 observations, the subsample analysis includes<br />

200 observations. The full-sample analysis explained 12.2% of the variance in RPE-based-<br />

Targets (adjusted R 2 = 0.076). The current subsample analysis results in an R 2 of 0.099<br />

(Adjusted R 2 = 0.047). Qualitatively, the subsample analysis yields the same results as the<br />

full-sample analysis. The results in table 2.19 - panel B support hypotheses H1 (the effect<br />

of common uncertainty on RPE use) and H3 (the interaction effect of uncertainty and<br />

information asymmetry on RPE use), but do not find siginficant support for the second<br />

hypothesis (the effect of the interaction between information asymmetry and comparability<br />

on RPE use).<br />

Concluding, the for-profit subsample analyses provide additional support for the hypotheses,<br />

for as far as the models themselves are significant. The insignificance of the RPE-Use<br />

model is probably the result of reducing the samplesize of the dataset.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!