Employee involvement and the middle manager: saboteur or scapegoat?

Summary of re s u l t s D o w n s i z i n g ±+D e l a y e r i n g +N o r m a t i v e+Experience of + +E I + S e l f - e f® c a c y + Intent toManagerial ± + support EI e m p o w e r m e n t ± Belief that EIis eff e c t i v e±Management ±l e v e l + EI leads to loss of powerCareer p r o s p e c t s+ Curvilinear re l a t i o n s h i p DISCUSSION B e f o re considering the implications of these results we should consider the possiblelimitations of the present study. First, the sample under study is not a random sample ofmanagers; the respondents are all members of the Institute of Management. It is possiblethat members of the institute are more favourably disposed than average to involvementinitiatives. Secondly, the response rate is not high (25 per cent). It is possible that there issome systematic bias in that perhaps respondents are more interested in and morefavourably disposed to employment initiatives than non-respondents. (We can drawsome reassurance from the ® nding that respondents and non-respondents show the samedistribution in terms of organisation size and sector.) The possibility of sampling bias ismost damaging to the validity of descriptive results, such as mean levels of attitudes.H o w e v e r, the main concern here is that the sample could be systematically biased insome way that led to a spurious correlation between variables. This is much less likelythan a simple bias in the mean value of a variable.A second possible problem is that of common method bias. Limitations on access tothe sample of managers in the study meant that all information had to be collectedt h rough a single questionnaire. There is some evidence that where a single questionnaireis used to collect data on both independent and dependent perceptual variables thatsome degree of correlation can arise simply as a result of the measurement process. Forthis reason it is sensible to disre g a rd small correlations between variables even where therelationship is statistically signi® cant. However, both the size of reported effects and thed i ff e rential relationships between independent variables and the various dependentvariables suggest the ® ndings are not simply a product of common method variance.With these cautions in mind we can consider the implications of the re s u l t s .