6 Analysis
6.2 Analysis of data from Step I
The data in Step I are limited. Despite this, I carry out not only a general analysis but also apply both qualitative (QCA) and quantitative (LISREL) tools to the available data.
6.2.1 General analysis of Step I information
The existence of implementation problems is confirmed by the executives (see tables 5 and 6 in 5.1.3). The varying implementation success in individual companies is also obvious according to the information in these tables. The picture is not very positive. Seven companies of 30 (24 %) have an implementation index
≤50 and the median is only 67. Unfortunately, there are few open comments linked directly to the individual distribution, which could help to understand the reasons for the low implementation index. So far it seems that there is an important improvement potential in implementation efficiency.
The implementation index varies between companies. In table 7, the companies have been arranged in three groups according to implementation efficiency level.
It is not possible to observe any evident or systematic tendencies regarding different variables. However, the profit situation is certainly interesting (see 3.4.2.1). Figure 8 gives a hint about a U-curve. Regression analysis on the data including all information (NB! In figure 8, some extreme values are excluded) produces a significant equation solution
INDEX = 63.7 + 0.32*PROFIT (7)
with R2=0.20. Adding PROFIT2 to the equation does not improve R2 and gives furthermore a non-significant solution: the U-curve is not confirmed. As already discussed in 5.1.2 there seems to be a slight correlation between poor profit and a newly appointed CEO, confirmed by computation results in table 8. The combination of these observations is analyzed with QCA in 6.2.2.
0 20 40 60 80 100
-30 -20 -10 0 10 20
PROFIT % IE index
Figure 8. Plotting implementation efficiency and company profitability (three year average;
some extreme profit values have been excluded due to diagram layout reasons)
Does size of company influence implementation efficiency? Some economic variables are listed under the corporate profile headline in table 7, which may be used as a measurement of the size of a company as well as the number of employees under the headline corporate culture in the same table. In this study,
“people” is focused. Therefore it seems reasonable to use number of employees as a measurement of the size of company.
In figure 9 the relation between implementation efficiency and size of company measured as number of employees is plotted excluding the four biggest companies with a range of 3000 to 14000 employees. The correlation is r=-0.18 and the average implementation efficiency index is 64. The four largest companies have an index range of 63 to 85, the later for the biggest one. However, all these figures together indicate that size does not matter regarding implementation efficiency.
The picture is unchanged if other variables are used as measurements of size.
0 20 40 60 80 100
0 500 1000 1500
Number of employees IE index
Figure 9. The relation between implementation efficiency and size of company
The CEO answers to the open questions about reasons for good versus poor implementation efficiency contain much interesting information (table 9). It seems that corporate culture factors play an important role; 12 CEOs say that formulated and shared values and needs are important for good implementation and 8 CEOs claim that internal resistance and cultural conflicts have the opposite effect. Also, communication and clarity are important for good implementation, as well as lacking resources have relevance for poor implementation. Contrarily, participating in the decision making process and a low complexity of the decision do not matter in terms of successful implementation; the follow-up system and insufficient anchoring are also not of importance.
There are no evident differences in CEO opinions when looking at different implementation efficiency levels. That means that even if the CEO estimates quite a poor status of implementation efficiency in his own company, he shares the opinions of his colleagues in more successful companies about reasons for good versus poor implementation efficiency.
A provocative reflection is that “well implemented” seems to be explained mainly in words of good top management and “poorly implemented” by problems
related to subordinates! It is, however, not possible to analyze this observation further due to shortage of information.
To summarize, I make the following conclusions based on the general analysis of Step I:
CC1. There is a potential for important improvements in implementation efficiency (S)
CC2. Corporate factors as formulated and shared values and needs as well as internal resistance and culture conflicts have impacts of good and poor implementation respectively (M)
CC3. Decision factors such as communication and clarity as well as available resources have impacts of good and poor implementation respectively (M)
CC4. Size of company does not matter with regard to implementation efficiency (M)
CC5. Extreme profit situations (very poor or very good) do not lead to high implementation efficiency (W)
CC6. Executives estimate that successful implementation mainly depends on themselves and unsuccessful implementation mainly depends on subordinates (-)
6.2.2 Step I QCA
The purpose of the QCA approach is to find causal relations between corporate factors and implementation efficiency in the preliminary model. The QCA technique is briefly described in 4.4.2.
The dataset includes almost 40 variables, of which six are economic variables from each of the last five years (that is 30 of the total 40). They are continuous with two exceptions: leadership style, which is trichotomous, and the presence of minutes of Top Management Team meetings, which is dichotomous. Evaluated variables such as turn over pro employee, profit margin and years as CEO in the company have been calculated from the original variables (see table 14).
Table 14. Transformation of the continuous variables into dichotomous variables in Step I
Estimation proposal Abbr 1 0
CEO's self-estimation of leadership style L Value Directive
Presence of minutes of meetings M YES NO
Year(s) as CEO Y >4 years <4 years
Age of CEO A <40 years >40 years
Turn over last year T >500 MSEK <500 MSEK
Solidity S <30% >30%
Number of employees last year E >1000 <1000 Profit, three year average P <0 alt >10% 0-10%
Growth, three year average G <0 alt >10% 0-10%
CEO's estimation of IE, evaluated I >66% <66%
Pivot value Selected variables
I have selected both original and evaluated variables, suitable for a QCA analysis. The criteria for selection are mainly the formulated research questions but my personal experiences have also been taken into account. The selection and the transformation from the original variable value into a dichotomous variable are shown in table 14. I have not grouped the variables under factor groups of the preliminary implementation model (see figure 3) as I have no direct information about corporate culture. However, the variables T, S, E, P and G constitute corporate profile.
The developed truth table is shown in table 15.
Table 15. Truth table of selected variables in Step I (abbreviations see table 14)
L M Y A T S E P G I
1 0 1 1 0 1 1 1 0 1 0
2 1 1 1 0 1 0 1 0 1 1
3 1 1 1 0 1 1 1 0 1 1
4 1 0 0 1 1 1 0 1 1 1
5 1 1 1 0 1 0 1 0 1 0
6 0 1 1 1 1 0 0 0 0 1
7 0 1 0 0 0 0 0 1 1 0
8 0 1 0 1 0 0 0 1 1 1
9 1 1 1 0 0 1 0 1 1 1
10 0 1 0 0 1 0 0 0 1 0
11 1 1 1 0 1 0 0 0 0 0
12 1 1 0 0 0 1 0 1 0 0
13 1 1 1 1 0 0 0 1 0 1
14 0 1 1 1 1 0 0 0 1 0
15 1 0 1 0 0 1 0 1 0 1
16 0 0 0 0 1 1 1 0 0 1
17 0 0 0 0 1 0 0 0 1 1
18 1 1 0 1 0 0 0 1 0 1
19 0 1 1 0 0 0 0 1 0 1
20 1 1 1 0 0 0 0 0 1 1
21 1 1 1 0 0 0 0 1 1 0
22 1 1 0 0 1 0 0 1 0 0
23 0 0 1 0 0 0 0 1 0 0
24 0 1 0 0 1 1 1 1 0 0
25 1 1 0 1 1 0 0 1 1 1
26 1 1 0 0 1 0 1 1 0 1
27 1 1 0 0 1 0 1 1 0 1
28 1 1 0 0 1 0 0 0 1 0
29 0 1 0 1 0 0 0 0 1 0
30 0 1 1 1 0 1 0 0 0 0
Nb of 1 17 25 15 9 17 9 8 16 16 16
Nb of 0 13 5 15 21 13 21 22 14 14 14
Transformed variables Row
1
The pivot values have been decided with one eye on the dataset (“reasonable 1/0 distribution”) and the other eye on the research questions. L was easy to manage as no respondent has characterized himself as a political leader. P has been treated according to RQ6 (see 3.4.2.1). The transformation contains some subjectivity. I have tested different pivot values before coming to the results in table 14. The main problem is that some variables are unbalanced in 1/0 cases even for small changes in pivot values.
When creating a truth table (see 4.4.2), there can be a problem with missing values. This occurs in my data. The matrix generated by the selected variables is 10x30. There are 9 missing values of the total potential of 300 observations, giving 24 rows to compute instead of 30. To solve this problem I have applied the imputation method described in 4.4.4. All missing values are imputed and the matrix is complete with 30 rows as shown in table 15.
The first analysis is carried out with three variables as this approach permits me to select one variable from each factor group. L(eadership style) is undisputable.
For corporate profile, P(rofit) is selected according to RQ6. Corporate culture was not covered by any questions to the CEOs. In table 14 I propose years as acting CEO, Y, even if it is discussed in 4.2.2.3 how far the CEO influence reaches on corporate culture. Alternatively the age of CEO, A, is a possible estimation of corporate culture. I choose Y, as Y has indicated some influence on the estimation of implementation efficiency (see tables 7 and 8). If the CEO has acted 4 years or more, he has had time to affect the corporate culture if possible at all. Therefore the QCA is carried out with L, P and Y as independent variables and I(mplementation efficiency) as dependent.
For all rows except 000 I use the technique of majority. That means that all cases are put into the outcome column, where the majority is already found (see table 16). Row 000 has two cases in each 1/0 column. All rows including 00 have a majority of 0-cases. Therefore it seems relevant to start the analysis by putting row 000 into outcome 0.
The truth table from the first analysis step is shown in table 17. All rows are covered by cases. But the contradictory row results are obvious. It is, however, possible to see a structure, which I use when handling the problem.
Table 16. Original truth table of L Y P > I in Step I
L Y P 1 0
1 1 1 2 1
1 1 0 4 2
1 0 1 5 2
1 0 0
0 1 1 1
0 1 0 1 4
0 0 1 1 2
0 0 0 2 2
I cases Independent var
3 6 7 1 1 5 3 4
The next step in the analysis is to formulate the equation. From table 17, the following equation is designed
LYP + LYp + LyP + lYP > I (8) Minimizing gives:
LY + LP + YP > I (9)
Equation (9) says that value leadership in combination with either a long period of CEO regime (at least four years) or a challenging economic situation (negative profit or above 10 %) cause high implementation efficiency, as well as a long period of CEO regime in combination with a challenging economic situation.
There are three pathways to high implementation efficiency and they are all a combination of two of the three variables.
Table 17. Adapted truth table of L Y P > I in Step I
L Y P
1 1 1 1
1 1 0 1
1 0 1 1
1 0 0 0
0 1 1 1
0 1 0 0
0 0 1 0
0 0 0 0
Cases I
Independent var
How robust is the presented solution? I answer this question by addressing three aspects. First, does adding more independent variables improve the explanation?
The model contains three corporate factor groups. If I use two variables of each instead of one, the number of rows increases from 8 to 64. The data has 30 sets.
Adding variables is obviously impossible. Picking up just one more variable, giving 16 rows, is technically possible. But doing so, the weight of a certain factor group is doubled giving a risk of misleading conclusions. Nevertheless I have tried. It is simplest to look at corporate profile. Adding growth, G, causes one empty row (no case observed) and a very high degree of contradictory row results.
So adding one more variable does not contribute to a better solution and thus G is unlikely to affect the outcome.
A second aspect is the choice of variables. I have separately tested age of CEO, A, instead of Y and growth, G, instead of P. A gives approximately the same truth table but with slightly higher contradictory row results. G gives higher contradictory row results. If I manage the contradictory row results with the same technique as earlier, G gives the following equation:
LYG + LYg + LyG + Lyg > I (10) which can be minimized to
L > I (11)
That means that value-driven leadership would be necessary and enough for high implementation efficiency, which does not seem convincing.
Finally I have tested changing the pivot values of Y and P. Y has been changed from 4 to 3 years. P has been changed to <3%/>3% for 1/0. The truth table is changed very little not forcing any changes in basic equation (9) if I manage the contradictory row results as before.
Row 000 is initially regarded as a 0-outcome. If it is managed as a 1-outcome, we get the following solution:
LYP + LYp + LyP + lYP + lyp > I (12) Minimizing gives
LY + LP + YP + lyp > I (13)
Equation (13) assigns a fourth pathway to high implementation efficiency compared with equation (9): a combination of directive leadership, short CEO regime and no profit challenge. The result shows the solution sensitivity of calculation conditions rather than a convincing fourth way to high implementation efficiency
In total, the sensitivity analysis indicates that solution (9) is reasonably consistent to changed conditions.
Tentative QCA-tests with alternative trichotomous combinations of variables in table 15 have not given any promising results.
CC7. Value-driven leadership, a long period of CEO regime and a challenging economic situation in pair-wise combinations lead to a high implementation efficiency (W)
The purpose of the QCA approach is achieved resulting in some interesting findings.
6.2.3 LISREL analysis of Step I
The purpose of the LISREL analysis is to find causal relations between corporate factors and implementation efficiency in the preliminary model. The LISREL tool is shortly described in 4.4.3. The variables used in the LISREL analysis here are the same as in table 14 (of course in their original form as continuous variables) completed with age of CEO when appointed and size of TMT (corporate culture).
Some of the variables have been transformed to logarithms before computing in order to approximate normal distribution characteristics of the variables (Jöreskog
& Sörbom, 1996).
I have made several computations but have not been successful in finding model fitness. The main reason seems to be a limitation of the dataset. However, some observations are made. It seems that profit, growth, solidity, number of employees and turn over in different combinations are useful variables for measuring corporate profile. Contrarily, it does not seem as CEO characteristics as age, years
of action as CEO and age at appointment are good estimations of corporate culture.
The purpose of the LISREL modeling approach to find causality in Step I information failed.