3. CAPÍTULO III MÉTODO, METODOLOGÍA Y CARACTERIZACIÓN DEL
3.1. Fases del trabajo de campo
3.1.1. Fase I Exploración y descubrimiento: Una aproximación al habitus del
5.3.1 The CSR Database
We take the opportunity presented by Black’s (2004) doctoral research thesis of a large Australian commercial organisation to test the DEA model. The organisation she studied was a bank which consistently scored highly on an Australian national Social
to establish the existence of CSR which could be measured and attributable to certain antecedent factors. Black was able to develop an empirically supported construct for CSR from an investigation of 39 variables in six organisational business units with usable data from 231 managerial units from a sample of 243. All were given individual and aggregated CSR scores based on responses to psychometric testing of the variables. A breakdown of the variables is presented in Table 5.2. The first category provides demographic data such as: age, gender, managerial role, years of service, business division, etc. The next category has an overall performance score for CSRCM based on the results of two categories, dimensions and indicative factors. The following categories are the antecedents of CSRMC and the consequences of good CSR, and finally one for variables that were found unusable.
Table 5.2 Breakdown of CSRCM variables in the study
Category Variables Measure Sample Size
Demographic Data 8 Nominal 243
CSR Management Capacity (Overall Score)
1 Scale (7) 228
Dimensions of CSRMC 5 Scale 227
Indicative Factors (of CSRMC) 8 Scale 227
Antecedents of CSRMC 7 Scale 231
Consequences of CSRMC 7 Scale 231
Other Miscellaneous 4 Scale 232
Black’s model was supported by traditional hypothesis-testing methodology with correlations between the indicative factors for the CSRMC all significant at the 0.01 level (2-tailed tests). There was strong support for the hypothetical model of antecedents and outcomes of CSRCM as proposed, and significant ANOVA results for its eight sub- dimensions as related to the employee stakeholder group. Qualitative results through comprehensive interviews further attested to the significance of the CSRMC construct. It was the positive results for a CSR model that invited an analysis of the data from the performance perspective of a DEA algorithm. No analysis of the type possible with DEA was attempted by Black.
5.3.2 The DEA Justification
Little seems to have been reported on studies of CG based on a diagnostic tool such as DEA. Perhaps this is because of the mathematical nature of the operations research based technique, as explained earlier, or because commercial firms have had CG traditionally examined from a non-quantifiable perspective. There also may have been little awareness of a need to use DEA because of its obscurity to researchers in the CG discipline, or because other established and research-rich investigative frameworks existed, e.g. board composition, independence and remuneration, stock holding participants, government, regulatory and other stakeholder influences.
This thesis, however, attempts to legitimize the use of DEA within the firm because of its attraction as a multi-criteria decision analysis technique with the benefit of weightings to assign a ranked position to the measured units. Adopting and applying the Golany and Roll (1989) procedure (see figures 3.3 and 3.4), DEA can perform well in the analysis of non-commensurate multiple inputs and outputs to give a measure of efficiency which reflects some aspect of organisational performance. The procedure for DEACSR is:
establish the population of DMUs as managerial units within the organization;
set the goals for analysis as efficiency;
select the number DMUs for comparison (231 from a mail response pool of 245);
define the input and output factors: the antecedent variables and the output the aggregated CSRMC variable;
examine those factors:
◦ by subjective judgment;
◦ by correlations; and
◦ by trial runs;
formalise the final model;
reach conclusions.
The output-input relationship that subsumes the DEA approach is grounded in the mathematical ratio form. This allows it to be generalized into a broader multiple criteria control model by using the Greenberg and Nunamaker (1987) transformation. Here the output-input factors themselves can be expressions of a ratio form as surrogates for the (difficult to obtain) exact measures of quantified tangible inputs and outputs. For example, scores from attitudinal surveys using instruments such as the popular Likert scale may be used by converting these scores into individual or aggregated ratios. In effect the supplanting of traditional metrics by surrogate factors expressed in a quantified form allows an articulation of those indicators that managers consider constitute good performance. Managers are often able to elicit what factors are contributors to overall good performance, and to rate these hierarchically but not absolutely. For example, two managers may score ‘motivation’ differently on some scale say 8/10 and 6/10, yet these scores are moot if both rate motivation ahead of ‘punctuality’.
Managers also know that some factors have a greater impact on performance than others but often feel that they are unable to gauge the weightings of importance for these factors. But:
…ignoring the interrelationships among performance measures and limitations on their possible combination may result in specification of weights vastly different from those which would have been specified had these aspects been considered. (Greenberg and Nunamaker 1987, p. 333)
Since the optimum set of weights can vary depending on a number of issues (even simple ones such as which measurement scales are used), managers may be unable, rather than unwilling, to assimilate the information in the specification of weights for different indicators.
The cognitive complexity of this task, together with managerial inexperience or lack of data may render it unassailable for these decision makers, and thus the task is not attempted or attempted rudimentarily.
DEA does not require the a priori assignment of relative weights to individual measures because the technique itself identifies the subset of performances that are Pareto optimal9 and these can be regarded by managers as those performance benchmarks which are indicators of good performance. Should the current factors not be Pareto optimal, the model shows where improvement is needed to achieve this condition. The degree of sub- optimality is also the amount of improvement possible and is referred to as slack, discussed previously.
This slack is only displayed by a non-efficient DMU. Improvement is possible by various alternate strategies such as achieving more from the same resources or by maintaining a stable output and reducing the resources required, and thus removing slack. DEA will be used to investigate the relationship between inputs and outputs for CSR where the eight indicative factors of the five dimensions would be considered in the aggregated CSRMC score as output and the antecedents as input factors. Black’s results were all expressed as inferential statistics thus allowing the transformation possible by DEA.
Each of Black’s sample managerial units is regarded as a DMU for the DEA computation.