pdf - Nyenrode Business Universiteit

pdf - Nyenrode Business Universiteit pdf - Nyenrode Business Universiteit

30.08.2013 Views

2.6. CONTRACTIBILITY MIMIC MODEL (APPENDIX A) 53 described by a sample of interchangeable indicators (Bisbe et al. 2007:792). With reflective indicators, one or more indicators are often excluded from the construct. This exclusion can be the result of a factor analysis and/or scale reliability analysis if a poorly fitting survey item is left out of the construct to enhance scale reliability. When measuring with formative indicators, this approach is not recommended because it alters the meaning of the construct. The second group of constructs consists of reflective indicators. Whereas formative constructs cause the latent construct, the reflective measures are its result. The reflective measures in this model are the following: difficulty to assess work (Q25, reversed coding), difficulty to reach agreement on output quality (Q26, reversed coding), and knowledge about the business unit manager’s job performance (Q27). In this case, the reflective indicators are all single-item constructs 34 . Table 2.16 presents the descriptive statistics for the formative and reflective contractibility indicators. Table 2.16: Descriptive Statistics for Contractibility Indicators Mean Min Max Std. Dev. SkewnessKurtosis Goal clarity 2.34 1.00 5.00 0.57 0.43 1.19 Measurability of outputs 4.48 2.00 6.00 0.69 -0.45 0.38 Transformation process 4.95 2.50 7.00 0.93 -0.48 -0.04 Difficulty to assess work 2.58 1.00 5.00 0.78 0.38 0.08 Difficulty to reach agreement on output quality 2.58 1.00 5.00 0.66 0.18 0.12 Knowledge about job performance 3.42 1.00 5.00 1.05 -1.06 1.62 Formative constructs, such as contractibility, can be calculated by using various approaches. For one, it is possible to calculate a formative construct as a summated scale of the underlying formative indicators only. This calculation is achieved by summating the unweighted scores of the formative indicators. The resulting scale includes all of the underlying indicators, which is important because all of the indicators function as necessary conditions for the formative construct. Because the scores on the formative indicators do not need to covary, factor analysing the components of the formative construct is meaningless, as is the Cronbach-type analysis of scale reliability (Speklé & Verbeeten 2008). Because oftentimes no theoretical suggestions exist concerning how the underlying indicators should be weighted, this unweighted approach is justifiable. 34 Because the reflective indicators are all single-item constructs, no factor analyses and/or scale reliability measures are reported.

54 CHAPTER 2. RPE AT THE BUSINESS UNIT MANAGER LEVEL I calculate contractibility not by using an unweighted summated scale but by using a more sophisticated MIMIC model. MIMIC also incorporates reflective indicators to calculate the relative weights of each of the formative indicators when forming the latent construct. As the relative weights of the three causal dimensions (goal clarity, measurability of outputs and knowledge of transformation process) cannot be derived theoretically, one may resort to structural equation techniques, such as MIMIC, to determine the appropriate weights (Bisbe et al. 2007). In the absence of a theoretical basis for the weights of the indicators, MIMIC at least provides an empirical basis for the index. MIMIC estimates the latent formative construct contractibility, as depicted in figure 2.4. The MIMIC model calculates contractibility by using the formative (causal) indicators on the left-hand side of the figure and the reflective (outcome) indicators on the right-hand side of the figure. The numbers on the arrows are standardized regression coefficients. The numbers above the reflective measures (boxes) and the latent construct (contractibility) indicate the amount of variance explained. The model explains 47% of the variance in the latent construct contractibility. Goal Clarity Measurability of Outputs Knowledge of Transformation Process Figure 2.4: Contractibility MIMIC Model .40 Difficulty to Asses Work (reversed) ✬ .47✩ .33 Agreement Contractibility Output Quality (latent) (reversed) ✫ ✪ .17 ✂ ✂ Job Performance ✂ ✂ ✂ ✂ ✂ ✂ ✂✍ .34 ❇ ❇ ❇ ❇ ❇ ❇ ❇ ✂ .33 ❇❇◆ ✂ ✲ ✂ .18 ✂ ✂ ✂ ✂ ✂ ✂✍ .63 .57 ✲ ❇ ❇ ❇ ❇ ❇ ❇ ❇ .41❇ ❇◆

2.6. CONTRACTIBILITY MIMIC MODEL (APPENDIX A) 53<br />

described by a sample of interchangeable indicators (Bisbe et al. 2007:792). With reflective<br />

indicators, one or more indicators are often excluded from the construct. This exclusion<br />

can be the result of a factor analysis and/or scale reliability analysis if a poorly fitting<br />

survey item is left out of the construct to enhance scale reliability. When measuring with<br />

formative indicators, this approach is not recommended because it alters the meaning of<br />

the construct.<br />

The second group of constructs consists of reflective indicators. Whereas formative constructs<br />

cause the latent construct, the reflective measures are its result. The reflective<br />

measures in this model are the following: difficulty to assess work (Q25, reversed coding),<br />

difficulty to reach agreement on output quality (Q26, reversed coding), and knowledge<br />

about the business unit manager’s job performance (Q27). In this case, the reflective indicators<br />

are all single-item constructs 34 . Table 2.16 presents the descriptive statistics for<br />

the formative and reflective contractibility indicators.<br />

Table 2.16: Descriptive Statistics for Contractibility Indicators<br />

Mean Min Max<br />

Std.<br />

Dev.<br />

SkewnessKurtosis<br />

Goal clarity 2.34 1.00 5.00 0.57 0.43 1.19<br />

Measurability of outputs 4.48 2.00 6.00 0.69 -0.45 0.38<br />

Transformation process 4.95 2.50 7.00 0.93 -0.48 -0.04<br />

Difficulty to assess work 2.58 1.00 5.00 0.78 0.38 0.08<br />

Difficulty to reach agreement on output<br />

quality<br />

2.58 1.00 5.00 0.66 0.18 0.12<br />

Knowledge about job performance 3.42 1.00 5.00 1.05 -1.06 1.62<br />

Formative constructs, such as contractibility, can be calculated by using various approaches.<br />

For one, it is possible to calculate a formative construct as a summated scale of the underlying<br />

formative indicators only. This calculation is achieved by summating the unweighted<br />

scores of the formative indicators. The resulting scale includes all of the underlying indicators,<br />

which is important because all of the indicators function as necessary conditions<br />

for the formative construct. Because the scores on the formative indicators do not need<br />

to covary, factor analysing the components of the formative construct is meaningless, as is<br />

the Cronbach-type analysis of scale reliability (Speklé & Verbeeten 2008). Because oftentimes<br />

no theoretical suggestions exist concerning how the underlying indicators should be<br />

weighted, this unweighted approach is justifiable.<br />

34 Because the reflective indicators are all single-item constructs, no factor analyses and/or scale reliability<br />

measures are reported.

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

Saved successfully!

Ooh no, something went wrong!