Development of an Organizational Memory Scale
Study 2 (b) Retiree Sample Method
Sampling
Seven groups, all under the umbrella of a social organization for fully-retired and partly-retired persons (generally with a professional/business work background) were approached for voluntary participants. There was at the time, a combined total of 798 members listed with the groups approached, although there was not a 100% attendance rate at the meetings where participation was sought. Typically there was a 20-25%
response rate of those attending to take away questionnaires. Out of the 200 questionnaires disseminated at the meetings, 115 were completed and returned representing a response rate of 57.5%.
Participants
Of those who completed demographic details, there were 75 (65.2%) males and 39 (34.2%) females, with an overall mean age of 72.80 years (SD=6.55). In terms of their past employment (specified as the participant‟s workplace immediately prior to their retirement decision), participants reported a mean organizational tenure of 23.29
(SD=13.64) years. There were 30 (26.1%) individuals who classified themselves as team members, 32 (27.8%) as team leaders/middle managers; 39 (33.9%) as team
leaders/senior managers, and 9 (7.8 %) who identified as having been the chief executive officer of their organization just immediately prior to retirement.
Materials
The questionnaire can be found in Appendix D, and a full description of the study measures can be found in Study 5. There were a number of questions pertaining to the participants‟ prior work and retirement process, followed by several scales, including the organizational memory scale. Participants were also asked about their experiences as a mentor and/or protégé and were asked to complete the intention to mentor scale and some demographic details. Responses for the organizational memory items were via a 7-point Likert-type scale anchored with 1= strongly disagree and 7 = strongly agree. Scale scores were obtained by summing the items in each, and mean item scores calculated by
dividing sum scores by the number of items in each scale.
Procedure
A pilot study involving seven retired or partly retired individuals involved in a business mentoring network was conducted to evaluate the questionnaire initially, and there were no changes made to the questionnaire as a result. Responses from the pilot study were not included in this sample. All participants were assured of the anonymity and confidentiality of their responses, and no incentive was given for participation.
Participants gave informed consent by completing and returning the questionnaire. A post-paid envelope was provided and participants were encouraged to return the survey within two weeks.
Results
The CFA results for the retiree sample can be seen in Table 2.7. Model 1 treated the 21 organizational memory scale items as all loading onto a single factor. The results showed poor fit with a positive x² value, a CFI value well below .90, and a RMSEA well above .08. Model 2 consisted of five unrelated first-order factors. Again this model exhibited poor fit although not to the extent of the first model, with the x², CFI, and RMSEA values still outside acceptable bounds.
Model 3 (see Figure 2.5) consisted of the first-order factors (allowed to correlate with one another). This resulted in an improvement in fit over Models 1 and 2. While the x² value was still significant, there was modest support for fit with acceptable CFI and RMSEA values. Model 4 (see Figure 2.6) consisted of the five first-order factors with one second-order factor of organizational memory. As with Model 3 there was an
improvement in fit indices over Models 1 and 2. Model 4 again resulted in a significant x², but also achieved modest fit with acceptable CFI and RMSEA statistics. While Model 4 had the higher x² value (indicating less fit), the difference between the values for Model 3 and Model 4 were not significant, and these models also shared acceptable RMSEA and CFI values. While the confirmatory factor analysis for the retiree sample did not achieve the same level of fit as the employee sample, it was still acceptable. Models 3 and 4 succeeded in achieving significantly more fit than Models 2 and 3.
Table 2.7. Confirmatory Factor Analysis Fit Indices for Organizational Memory scale (Five Factors) with Retiree Sample (N=115)
x² (df ) p CFI RMSEA Lo Hi PCLOSE Comparison
x² Change
df Change
p for x² Change
Model 1
N=115 694.63 (189) .00 .58 .15 .14 .17 .00 Models 1 & 3 396.26 10 <.001 Model 2
N=115 431.98 (189) .00 .79 .11 .09 .12 .00 Models 2 & 3 133.61 10 <.001 Model 3
N=115 298.37 (179) .00 .90 .08 .06 .09 .00 Models 3 & 4 5.67 5 ns Model 4
N=115 304.04 (184) .00 .90 .08 .06 .09 .00 Models 1 & 4 Models 2 & 4
390.59 127.94
5 5
<.001
<.001
Figure 2.5. Confirmatory factor analysis diagram of five correlated first-order factors of organizational memory (Model 3) for the retiree sample (N=115).
Figure 2.6. Confirmatory factor analysis diagram of five first-order factors and one second-order factor of organizational memory (Model 4) for the retiree sample (N=115).
Descriptive Statistics
The mean item scores, means, standard deviations, coefficient alphas, and intercorrelations of the study variables for the retiree sample can be found in Table 2.8.
For the retirees, as with the employee sample, job knowledge, external network, and socio-political knowledge had the three highest mean item scores. While these scores are high, they should be interpreted with some perspective. Participants in this study had greatest mean age and tenure, and a higher percentage of them had achieved advanced positions within their organizations, compared with the two previous samples. Their organizational memory results should reflect this level of experience.
The intercorrelations of the study variables (see Table 2.8) showed organizational tenure correlated only with history and external network, while job tenure correlated with the history subscale alone. There were negative correlations between age and all of the organizational memory subscales, although only the correlation between age and industry knowledge was significant. There were also negative correlations (although not
significant) between organizational memory (except job knowledge) and length of retirement. These results may suggest some memory decay over time. The coefficient alphas (also see Table 2.8) for the organizational memory scale and subscales ranged from .73 to .90 for the retirement sample, indicating acceptable to excellent internal consistency estimates.
When comparing the results of both the employee and retiree samples, the high correlation between the history and industry knowledge factors is of some note (see Figure 2.3 for employees, and Figure 2.5 for the retirees). This may be an issue for the stability of the factor structure of the scale in future studies.
Table 2.8. Study 2(b) Means, Mean Item Scores (Scales), Standard Deviations, Coefficient Alphas, and Correlation Matrix (Retiree Sample, N=115)
M SD α 1 2 3 4 5 6 7 8 9 10
1. Age 72.31 8.09 1.00
2. Job tenure 14.42 10.86 .11 1.00
3. Organizational tenure 23.29 13.64 .31** .39*** 1.00
4. Length of retirement 11.37 7.57 .82 -.01 .20 1.00
5. Socio-political knowledge 5.67 1.05 .90 -.07 .19 .05 -.15 1.00
6. Job knowledge 6.49 0.59 .85 -.08 .11 .15 .05 .34*** 1.00
7. External network 5.68 1.21 .80 -.06 .14 .25* -.10 .51*** .39*** 1.00
8. History 4.89 1.27 .78 -.19 .22* .28** -.11 .36*** .25* .46*** 1.00
9. Industry knowledge 5.14 1.13 .73 -.21* .19 .06 -.18 .42*** .29** .58*** .56*** 1.00
10. Organizational memory 5.59 0.76 .90 -.15 .21* .20 -.14 .81*** .54*** .76*** .73*** .75*** 1.00
* p < 0.05 level; **p < 0.01 level, *** p < 0.001 level, (two-tailed).
Discussion
Study 2 (a) and (b) sought to confirm the factor structure of the organizational memory scale established in the EFA in Study 1. Models 3 and 4 both achieved modest fit with both the employee and retiree samples. As hypothesised, Model 3 achieved benchmark status out of the four models proposed for both samples as suggested by Marsh and Hocevar (1985). There was a significant difference in fit between the Models 3 and 4 for the employee sample, but not for the retiree sample. As hypothesised, Models 3 and 4 also achieved greater fit than Models 1 and 2 in both samples. The CFA analyses show support for the correlated five-factor solution of organizational memory established in the EFA in both the employee and retiree samples, as well as confirmation of a proposed second-order factor of
organizational memory.
Study 2 also gave the opportunity to assess the construct validity of the subscales through the inclusion of several demographic variables. Of particular importance is the relationship between the organizational memory scale and subscales, and the variable organizational tenure, as Walsh and Ungson (1991) have stated:
The most important individual attribute that is relevant to the study of organizational memory may be the length of service in the organization.... The absolute length of service in the tenure profile of an organization is critical to the effective retrieval of information. (p.78)
The employee sample showed promising results with organizational tenure correlating significantly with all the organizational memory subscales. Job tenure correlated with only two of the subscales, job knowledge as expected, and history. Industry tenure correlated with four subscales, two of these relationships much stronger (industry knowledge as expected,
and history) than the others (socio-political knowledge and job knowledge). It had also been expected that industry tenure would correlate with external network, when actually this was the only organizational memory subscale it did not correlate with. However, workers may have moved around geographically, and in that case, industry tenure may not necessarily correlate with external network. All three tenure variables correlated significantly with the job knowledge, history and overall organizational memory scales.
In the retiree sample, variables that had previously shown significant correlations with the organizational memory scales did not fare as well. The links between both job and
organizational tenure and the organizational memory subscales were more tenuous.
Organizational tenure correlated only with external network and history while job tenure correlated with history alone. Perhaps the incremental value of organizational tenure (in terms of gains in organizational memory) plateaus beyond an optimal time (e.g., like the ten years required for expertise suggested by Leonard & Swap, 2005b), and beyond that stage other variables (i.e., hierarchical position) become more influential. It is interesting to note that the history subscale, not surprisingly, had the most consistent relationship with tenure variables across the samples in Study 1, Study 2(a), and Study 2(b).
While the job knowledge subscale related well with the tenure variables, it also related least well with the combined organizational memory scale in both employee and retiree samples which may suggest the utility of developing a fully separate job knowledge and expertise scale. The high mean score on this subscale across samples may also suggest the need for further development of the subscale, to provide further differentiation between individuals of varying expertise. The two items of the discarded astuteness scale that did correlate well in Study 1 show some potential with the addition of other items to create a scale of job expertise or “deep smarts” (e.g., Leonard & Swap, 2005a), and future research may use these as the basis for future scale development.
With regards to reliability there are positive signs with the organizational memory scale and subscales achieving acceptable to excellent internal consistency estimates. However the history and industry knowledge scales seem to rate consistently less well than the other three subscales, and there may be issues with some similarity between these subscales as indicated with the removal of one item that loaded on both factors in the EFA. There is potential to add items to each subscale to improve internal consistency which despite achieving respectable levels look vulnerable compared with the other subscales.
Conclusion
This chapter chronicled the development of the organizational memory scale. The scale was designed to enable employees to estimate their resources across a number of job and organizational knowledge domains which are relevant to transfer by mentoring. The knowledge categories proposed for the study, and the resulting subscales do not represent an exhaustive list of the domains that make up organizational memory, and there is potential to add to these, with priority perhaps to be given to further development of a separate job-related knowledge and expertise scale. Overall, the scale shows promise for use in future research with a consistent factor structure, good internal consistency estimates for the overall scale and subscales, and support for the expected relationships between the subscales and the tenure variables often used as “proxy” variables for the acquisition of job, organizational, and industry knowledge.