In the second step of the analysis, multivariate analysis was performed to find out to what extent expected consequences of growth could explain growth willingness, and what specific expectancy variables had smaller or larger effect on growth willingness. The contingencies from the previous analyses were used to divide the sample, since it is possible that the pattern of relationships differ between industries, size brackets and age groups. The results from the regressions are displayed in Table 5.
Comparison Of Means For Growth Willingness And Expected Consequences Of Growth For Firms Of Different Industries.
|Service n=340||Manufacturing n=571||Retail n=246|
|Survival of crises||2.44||2.36||2.48|
Note: One-way ANOVA with Bonferroni test is used in the analysis. a= growth willingness is measured on a 7-point scale, 1 indicating a strongly negative and 7 a strongly positive attitude. b= expected consequences of growth is measured on a 5-point scale, 1 indicating a strongly negative and 5 a strongly positive attitude. c= p< 0.05 for difference to lowest group.
The adjusted explained variance, ranging from 0.16 to 0.28, indicates that expected consequences have an influence on growth willingness and that the proposed model is relevant.
The results reveal that non-economic concerns are very important determinants of growth willingness. Personal income is not the most important variable in any regression, suggesting that money is not the most important motivator. In all regressions, employee well-being is the most important explanatory variable, giving a remarkably consistent result. Workload is relatively unimportant in all regressions as well as work-tasks, with the exception of the smallest size bracket. The pattern concerning independence is the opposite, it has an effect in all regressions except in the smallest size bracket and the service industry, where the largest proportion of small firms are found.
Significant standardised regression coefficients for the remaining explanatory variables have the same signs across all analyses which indicates that the regressions are stable. However, their rank order and magnitude vary depending on how the sample is divided. Interpretation of these coefficients should be restrictive. Considering the moderate explained variance and the magnitude of the employee well-being coefficient in all regressions, relatively little is explained by other variables in all regressions.
Linear Regression Results For The Effect Of Expected Consequences Of Growth On Growth Willingness When The Sample Is Divided Based On The Three Contingencies Industry, Size And Age.
|Manuf. n=571||Service n=340||Retail n=246||5-9 emp n=326||10-19 emp n=479||20-49 emp n=353||Old firms n=771||Young firms n=372|
|Survival of crises||.09*||.10||.04||.07||.07||.13*||.09**||.06|
Note: Forced entry of independent variables is used. Standardised regression coefficients are displayed in the Table. *= p< 0.05; **= p< 0.01; ***= p< 0.001
Further support for the latter result is provided when the regression is run for the full sample, which is displayed in Table 6. While significant effects are obtained for all expected consequences but work tasks, the magnitude of the coefficients for variables other than employee well-being are small in magnitude. Due to the large number of cases in the full sample, significant results are easier to obtain. When the contingencies are added to the equation as dummy variables they alter the equation only to a small extent. Albeit significant on two instances, their standardised regression coefficients are generally low and the explained variance is not increased. In all, the conclusion is that explanatory variables are not dramatically different in different industries, size brackets or age groups.
As mentioned earlier, data were collected from three different samples during a ten year period. There are reasons to analyses the samples separately and compare the results. First, this makes it possible to check the stability of the results, if the results are the same for all three samples, conclusions will be more valid. Second, data were collected during different stages of the business cycle which may affect the attitudes of the respondents. Different explanatory variables may be important during different phases of the business cycle. A pure trend effect over time is also conceivable.
The results of the analyses of the three different
samples are displayed in Table 7.
Employee well-being is by far the most important explanatory
variable in all samples, whereas the magnitude and rank order of
all other explanatory variables vary. In all, the relationships
are relatively stable over-all, but not in detail. No clear
cyclical or trend pattern emerges over this time period.
© 1997 Babson College All Rights Reserved
Last Updated 06/01/98