Research Method

The unit of analysis for this research is CEO or delegate in local government in Egypt. The sampling frame included the directory of DSS units in the Egyptian local governments issued by the Information and Decision Support Centre (IDSC). A package that was mailed to senior executive officers contained two items: a cover letter explaining the importance of the study, and the questionnaire with a stamped return address on the back. The covering letter requested the respondent to return the completed questionnaire within two weeks. The respondents were assured of the confidentiality of their responses.

Of the 309 questionnaires that were returned from sample, 294 (about 73.5 percent) were valid, 12 incomplete and three returned by the post-office due to incorrect addresses. To ensure that the valid responses were representatives of the larger population, a non-response bias test was used to compare the early and late respondents. χ² tests show no significant difference between the two groups of respondents at the 5 percent significance level, implying that non-response bias is not a concern.

Cronbah's coefficient ± was used to assess the reliability of all multi-item scales. All scales showed reasonable reliability (Ramaprasad, 1987) as indicated in Table 2.

Table 2: Cronbah's coefficient ± for constructs

Factors	α
DSS usage (3 items)	0.70
PEU (6 items)	0.69
PU (7 items)	0.72
Task characteristics (5items)	0.65
Cultural characteristics (4 items)	0.78
DSS characteristics (12 items)	0.68
Environmental characteristics (4 items)	0.71
Organizational characteristics (7 items)	0.78
Internal support characteristics (5 items)	0.74
External support characteristics (3 items)	0.81
Decision maker characteristics (12 items)	0.68
Top management support (6 items)	0.79

Following Taylor and Todd (1995), because of sample size limitations, multi-item constructs for the external variables were measured using a summated scale derived as the average value of all items pertaining to these constructs.

SEM techniques are a second-generation multivariate technique and have gained increasing popularity in management sciences, notably marketing and organizational behaviour, in the last decade (Chau, 1997). AMOS 4.0 program (Arbuckle and Wothke, 1999) was used to test the hypothesized linear effect of each group of variables on PEU, PU and DSS usage. There are a number of measures generated by AMOS to evaluate the goodness of fit of the model, as with other commercial statistical software packages that adopt the structural equation modelling approach.

The most popular index is perhaps the chi-square statistic. This statistic tests the proposed model against the general alternative in which all observed variables are correlated. It measures the distance (difference, discrepancy, deviance) between the sample covariance or correlation matrix and the fitted covariance or correlation matrix (Joreskog and Sorbom, 1993). With this index, significant values indicate poor model fit while insignificant values indicate good fit. This is why it is also called a "badness-of-fit" measure. Hartwick and Barki (1994) pointed out a major shortcoming of this index. They noted "in large samples, the chi-square statistic will almost always be significant, since chi-square is a direct function of a sample size. In small samples, the statistic may not be chi-square distributed, leading to inaccurate probability values." In their study, Hartwick and Barki used four other measures of overall model goodness of fit: chi-square/degree of freedom, Non-Normed Fit Index (NNFI), Comparative Fit Index (CFI), and Average Absolute Standardised Residual (AASR). In another study, Segars Grover (1993) included several other measures of model fit: goodness-of-fit Index (GFI), Adjusted Goodness-of-fit Index (AGFI), fit criterion, and Root Mean Square Residual. Table 3 lists the recommended values of various measures of model fit as suggested by these authors. Many researchers recommend that multiple fit criteria be used (Bollen and Long, 1993; Breckler, 1990; Tanaka, 1993) in order to attenuate any measuring biases inherent in different measures.

Table 3: Recommended values of goodness-of-fit measures

Goodness of fit measure	Recommended Value
Chi-square	p ≥ .05
Chi-square/degree of freedom	≤ 3.0
Goodness-of-fit Index (GFI)	≥ .90
Adjusted Goodness-of-fit Index (AGFI)	≥ .80
Normed Fit Index (NFI)	≥ .90
Non-Normed Fit Index (NNFI)	≥ .90
Comparative Fit Index (CFI)	≥ .90
Root Mean Square Residual (RMSR)	≤ .10
Incremental Fit Index (IFI)	≥ 0.90
Adapted with modifications from Hartwick and Barki (1994) and Segars and Grover (1993)

The hypothesized research model is shown in Figure 2. The goodness of fit measures for this model are summarized in Table 4 indicating a significant χ² = 246.58,df =225,p =154. This result indicated a good fit, as the probability level was above the generally accepted critical value p = .05, which supported the research hypotheses.

Table 4: Fit measures for task characteristics model

Fit Measure	Task Characteristics model
Discrepancy (CMIN)	246.58
Degrees of freedom	225
P	0.15
Number of parameters (NPAR)	100
Discrepancy / df (CMINDF)	1.10
RMR	0.06
GFI	0.94
Adjusted GFI	0.91
Parsimony-adjusted GFI	0.65
Normed fit index (NFI)	0.68
Relative fit index (RFI)	0.57
Incremental fit index (IFI)	0.96
Tucker-Lewis index (TLI)	0.94
Comparative fit index (CFI)	0.95
Parsimony ratio (PRATIO)	0.75
Parsimony-adjusted NFI (PNFI)	0.51
Parsimony-adjusted CFI (PCFI)	0.72
RMSEA (PCLOSE)	0.02
P for test of close fit	1.00

Figure 2: DSS adoption model for SDM

The methodological problems in cross-cultural research (Ercan et al., 1991) were noted, and mindful of the ongoing debate (Boyd, 1993) in the field, the authors acknowledged and attempted to integrate current developments into their research design. As with the major part of cross-cultural research (Leung and Bond, 1989), this was a cross-sectional static study. In particular, it is important to note that the number of cultures generally included in cross-cultural research needs to be low (Nath, 1968; Sekaran, 1983), and that much of the research as yet should be viewed as being primarily developmental in nature (Adler, 1983). Moreover, it is necessary to appreciate that language is not a neutral vehicle, and that our thinking is affected by the words, phrases and categories available in our language (Hofstede, 1980). Consequently, equivalence of meaning is more important than direct translation, and research study needs to be designed, executed and interpreted from each participating culture's perspective and not from a single culture. Such points were borne in mind in the process of questionnaire design and analysis of results.

The discussion and analysis that follows examines and defines the relevance of the factors that influence DSS usage in making strategic decisions in local governments in Egypt.