This chapter outlines the results from the pilot sample and descriptive statistics of the research sample that contains 535 respondents. Factor analysis is used with corporate social responsibility (CSR) dimensions and its' results are discussed in this chapter and its' validation is illustrated too. The chapter also contains the analysis of data collected and tested for research hypotheses.

**5.1 Analysis of Pilot Survey**

The pilot survey was distributed using a sample of cases. Two methods of sampling were thought for consideration. The first is convenience sampling (non-probability based sampling) and the second is systematic random sampling (probability based sampling). A decision was in favor in selecting the second method (systematic based sampling) rather than the first method due to its lack of an upper boundary on the number of cases coming out from a specific mobile outlet. A random starting point was selected from the range of (1 to 9) and the starting point was randomly selected as each 6th customer coming out of the mobile outlet. A balanced sample size was selected (20 customer from each mobile operator) from the population under study to examine the psychometric properties of the constructs. In this situation, the population was represented by mobile outlets distributed in major Alexandrian malls (Carrefour and San Stefano). This was achieved by means of calculating the Cronbach's reliability coefficient for each construct and the items-to-total correlation (Nunnally and Bernstein, 1994) to assess the internal reliability of each construct and whether items are related to a given construct.

Table 5.1: Measure of variables reliability for pilot survey

Constructs/Items

Cronbach's for construct

ER: Economic Responsibility (4 items)

0.662

LR: Legal Responsibility (3 items)

0.645

ETR: Ethical Responsibility (4 items)

0.683

PR: Philanthropic Responsibility (3 items)

0.804

**Total items of CSR (14)**

**0.777**

PSQT: Technical (3 items)

0.693

PSQF: Functional (6 items)

0.721

PSQP: Price (2 items)

0.765

**Total items of PSQ (11)**

**0.741**

CL: Behavioral (4 items)

0.747

CL: Attitudinal (4 items)

0.759

**Total items of CL (8)**

**0.854**

**Total items (33)**

**0.863**

The Cronabach's for each construct is displayed in Table 5.1. For items comprising economic responsibility construct (4 items), coefficient of item to total correlation were all above 0.3 (see appendix 2) indicating that no items were subject for deletion (Hair et al., 2000). Additionally, the Cronabach' alpha value for the economic responsibility construct was 0.662 that indicates the homogeneity and consistency of the items comprising the scale, since it is larger than 0.6. This indicates that all items of the economic construct collectively contribute in building the construct and any items deletion would result negatively in building that construct (Sakaran, 2003).

Likewise, for items comprising legal responsibility construct (3 items) the coefficients of item to total correlation all above 0.3 (see appendix 2) as recommended by Hair et al. (2000), with corresponding Cronbach' value of 0.645 indicating that all items contribute to building the legal construct of CSR. Similar conclusions can be drawn for ethical responsibility construct (4 items) with corresponding Cronbach' value of 0.683, and for philanthropic responsibility construct (3 items) with corresponding Crobach' alpha value of 0.804. The Cronbach' alpha value was calculated also for the total items comprising CSR (14 items) and its value was 0.777 indicating that these items are reliable for measuring the CSR in an exploratory setting (see section 5.3 for EFA analysis on the observed constructs).

For items comprising perceived technical service quality construct (3 items), coefficient of item to total correlation were all above 0.3 (see appendix 2) indicating that no items were subject for deletion, (Hair et al., 2000). Additionally, the Cronabach' alpha value for the technical construct was 0.693 indicting the homogeneity and consistency of the items comprising the scale, since it is larger than 0.6. This indicates that all items of the technical construct for perceived service quality collectively contribute in building the construct and deletion of any items would result negatively in building that construct (Sakaran, 2003).

Similar conclusions can be drawn for perceived functional service quality construct (6 items) with corresponding Cronbach' alpha value of 0.721, and for price construct (only 2 items) the values of item to total correlation can't be calculated with corresponding Crobach' alpha value of 0.765. The Cronbach' alpha value was calculated also for the total items comprising perceived service quality (11 items) and its value was 0.741 indicating that these items are reliable for measuring perceived service quality.

For items comprising behavioral customer loyalty construct (4 items), coefficient of item to total correlation were all above 0.3 (see appendix 2) indicating that no items were subject for deletion (Hair et al., 2000). Additionally, the Cronbach's alpha value for the behavior construct was 0.747 indicating the homogeneity and consistency of the items comprising the scale, since it is larger than 0.6. This indicates that all items of the behavioral customer loyalty construct collectively contribute in building the construct and deletion of any items would result negatively in building that construct (Sakaran, 2003).

Same conclusion can be drawn for items comprising attitudinal customer loyalty construct (4 items), and the Cronbach's alpha value for the attitudinal construct was 0.759. The Cronbach' alpha value was calculated also for the total items comprising customers loyalty (8 items) and its value was 0.854 indicating that these items are reliable for measuring customer loyalty. The Cronbach' alpha value was calculated also for the whole questionnaire with (33 items) and its value was 0.863 indicating that these items are reliable for measuring the relations among the variables.

**5.2 Descriptive Statistics**

The respondents in this study were 535 representing mobile users of 3 mobile operators in Egypt (see Table 5.2 for the characteristics of sample respondents) of which 212 of them (39.6%) were Mobinil users, 234 of them (43.7%) were Vodavone users, while only 89 (16.7%) of them were Etisalat users.

Table 5.2: Sample Characteristics

Variable

Mobinil

Vodafone

Etisalat

Total

212*

**%**

234*

**%**

89*

**%**

100* (%)

Gender

Male

113

53.3

120

51.3

56

62.9

289 (54)

Female

99

46.7

114

48.7

33

37.1

246 (46)

Age

16-25

91

42.9

79

33.8

48

53.9

218 (40.7)

26-35

66

31.1

99

42.3

25

28.1

190 (35.5)

36-45

34

16.0

34

14.5

9

10.1

77 (14.3)

46-55

12

5.7

13

5.6

4

4.5

29 (5.4)

> 55

9

4.2

9

3.8

3

3.4

21 (3.9)

Marital Status

Single or Engaged

130

61.3

139

59.4

62

69.7

331 (61.9)

Married

72

34.0

84

35.9

21

23.6

177 (33.1)

Divorced or Widowed

10

4.7

11

4.7

6

6.7

27 (5.0)

Education

Before University

18

8.5

15

6.4

12

13.5

45 (8.4)

University Student

40

18.9

30

12.8

22

24.7

92 (17.1)

University Graduate

108

50.9

121

51.7

41

46.1

270 (50.5)

Post Graduate

46

21.7

68

29.1

14

15.7

128 (23.9)

Occupation

Admin. & Execute.

40

18.9

70

29.9

15

16.9

125 (23.4)

Prof. & Technical

43

20.3

41

17.5

12

13.4

96 (17.9)

Academic

56

26.4

66

28.2

26

29.2

148 (27.7)

Employer & Worker

22

10.4

13

5.5

8

9

43 (7.9)

Other

51

24.1

44

18.8

28

31.5

123 (22.9)

Monthly Income

< 1000

87

41.0

71

30.3

43

48.3

201 (37.6)

1000-3000

65

30.7

88

37.6

27

30.3

180 (33.6)

3001-5000

24

11.3

38

16.2

8

9.0

70 (13.1)

5001-10000

33

15.6

33

14.1

8

9.0

74 (13.8)

> 10000

3

1.4

4

1.7

3

3.3

10 (1.86)

Note: Values given in parentheses for the last column are calculated as percentage.

*: Represent respondents count for each mobile operator.

Figure 5.1 depicts the classification of gender and age group, in which 289 of them (54%) were male, 246 of them (46%) were female. The vast majority of age range of the respondents was 16-55 years with a mean of 30.12 years for the whole sample.

Figure 5.1: Classification of gender, age group, marital status, and education by categories

Additionally, the mean age for Mobinil users was 30.21 years, while 30.84 years for Vodavone, and 28.03 years for Etisalat users. Age wise, 40.7% (218 respondents) were aged 16-25 years, 35.5% (190 respondents) were 26-35 years, 14.3% (77 respondents) were 36-45 years, 5.4% (29 respondents) were 46-55 years, and only 4% (21 respondents) were above 55 years. According to the marital status of the respondents, 331 of them (62%) were single or engaged, 177 of them (33%) were married, and only 27 of them (5%) were divorced or widowed (see Figure 5.1 for more details).

Among all respondents, 8.4% (45 respondents) had an education indicated as before university, 17% (92 respondents) were university student, 50.5% (270 respondents) had a university degree, and 23.9% (128 respondents) were post graduate students studying either Master or PhD degree.

Figure 5.2: Classification of occupation and monthly income by categories

Figure 5.2 displays the occupation level of the respondents, 23.4% (125 respondents) were classified as administration or executives, 17.9% (96 respondents) were professional or technical, 27.7% (148 respondents) were working in academic fields, 7.9% (43 respondents) were employer and worker, and 23% (123 respondents) were classified as other. The last category represented by other include student, unemployed, and all other relevant occupation.

According to the monthly income of the respondents, 37.6% (201 respondents) of them had monthly income less than 1000 Egyptian Pound, 33.6% (180 respondents) of them had monthly income 1000-3000 Egyptian Pounds, 13.1% (70 respondents) of them had 3001-5000 Egyptian Pounds, 13.8% (74 respondents) had 5001-10000 Egyptian Pounds, and only 1.86% (10 respondents) had more than 10000 Egyptian Pounds as monthly income. Other counts and percentages for each demographic variable classified for each mobile operator are presented in Table 5.2.

**5.3 Factor Analysis of CSR**

First, the psychometric properties of the constructs for exploring the three constructs of the whole sample () were assessed by calculating the Cronbach's reliability coefficient (Nunnally and Bernstein, 1994). These coefficients are represented for each of the constructs in Table 5.3. All scales have reliability coefficients ranging from 0.677 to 0.816, which exceed the cut-off value of 0.60 set for basic research (Nunnally and Bernstein, 1994).

Table 5.3: Measure of constructs' reliability for CSR of Mobile users

Constructs

Number of Items

Economic responsibility

4

0.712

Legal responsibility

3

0.699

Ethical responsibility

4

0.677

Philanthropic responsibility

3

0.816

Total

14

0.790

Second, an exploratory factor analysis was performed within the six-stage model-building framework introduced by Hair et al. (2006). The main objective using factor analysis was to examine if the items for a construct share a single underlying factor, or whether items can be deleted. As a secondary objective, whether a new structure for the constructs comprising the four scales of corporate social responsibility can be revealed by examining the correlation structure in the data. The IBM statistical package for social science (IBM SPSS V19) was used to conduct the factor analysis and analyze the scale structure. Table 5.4 presents the correlation matrix for 14 items of the scale.

Table 5.4: Correlation matrix of the CSR constructs

Item No.

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

1.

1

2.

**.122****

1

3.

**.164****

**.282****

1

4.

.029

**.191****

**.257****

1

5.

**-.098***

**.232****

**.158****

**.171****

1

6.

**-.193****

.079

.005

.039

**.294****

1

7.

**-.108***

**.141****

.068

.082

**.402****

**.236****

1

8.

**-.210****

**.263****

**.177****

**.119****

**.503****

**.239****

**.433****

1

9.

**-.180****

.037

.068

**.090***

**.321****

**.156****

**.290****

**.330****

1

10.

-.043

**.115****

**.148****

**.113****

**.347****

**.239****

**.427****

**.372****

**.415****

1

11.

-.073

**.115****

**.130****

.066

**.319****

**.207****

**.276****

**.388****

**.237****

**.348****

1

12.

**-.170****

.058

**.118****

.082

**.345****

**.156****

**.341****

**.390****

**.639****

**.455****

**.263****

1

13.

**-.140****

**.096***

**.108***

**.133****

**.325****

**.131****

**.257****

**.314****

**.587****

**.390****

**.252****

**.613****

1

14.

-.069

**.160****

.080

.074

**.345****

**.172****

**.296****

**.336****

**.523****

**.447****

**.237****

**.526****

**.655****

1

** Correlation is significant at the 0.01 level (two-tailed).

* Correlation is significant at the 0.05 level (two-tailed).

A review of the correlation matrix reveals that 70 of the 91 correlations (approximately 77%) are significant at the 0.01 level while only 5 of the 91 (approximately 5%) are significant at the 0.05 level, which provide adequate basis to perform a factor analysis for each item and for the overall basis.

An additional step in conducting an exploratory factor analysis (EFA) involves examining the suitability of the data to support the analysis. The Kaiser-Myer-Olkin (KMO) measure of sampling adequacy (see Table 5.5) is found to be 0.882, higher than the acceptable value of 0.6 suggested by Pallant (2010), indicating that the sample size is large enough to factor analyze 14 variables and the KMO test is judged to be Marvelous (Hair et al., 2006).

Table 5.5: KMO and Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy

.882

Bartlett's Test of Sphericity

Approx. Chi-Square

3630.204

Df

190

Sig.

.000

Additionally, the Chi-square value of Bartlett's Test of Sphericity shows a statistically significant correlation (). This indicates the suitability of the intercorrelation matrix of the 14 items for factor analysis. Thus, the sample size and the nature of the data are both fit for the exploratory factor analysis.

Table 5.6: Total variance explained and un-rotated factor loading matrix

No.

Items

Factor

Commonalities

Factor 1

Factor 2

Factor 3

1

Achieve maximum profit

-.216

.454

.421

0.430

2

Maximize stock holder's wealth

.267

.637

.018

0.477

3

Maximize long term success

.234

.643

.254

0.532

4

Minimize its cost

.218

.479

.175

0.308

5

Commitment to all rules and regulations

.648

.213

-.268

0.537

6

Doesn't commit with some rules and regulations

.374

-.019

-.542

0.434

7

Works under strict social laws

.593

.067

-.312

0.453

8

Ethical principals over making profit

.673

.175

-.320

0.587

9

Helps government to solve social problems

.267

.637

.018

0.675

10

Always does what is right

.673

.175

-.320

0.466

11

Deals honestly with its clients

.513

.127

-.270

0.353

12

Engages in R&D to improve level of society

.750

-.250

.260

0.693

13

Sufficient monetary contribution to charities

.724

-.229

.388

0.727

14

Encourages customers to participate in charities

.716

-.168

.316

0.641

Total

Sum of squares (eigenvalues)

4.419

1.586

1.307

7.312

Percentage of trace

31.565

11.327

9.338

52.229

The data were analyzed through principal components factor analysis using VARIMAX rotation method. Table 5.6 shows the total variance explained and un-rotated factor loading matrix regarding the 14 items comprising CSR with three possible factors and their relative explanatory powers. In the table, it is possible to assess the importance of each component and select the ideal number of factors while using the eigenvalues at the same time.

The three factors capture 52.23% of the variance of the 14 items, which can be deemed sufficient in terms of explained total variance. The table shows that factor (1) accounts for 31.565% of the variance (eigenvalue 4.419), factor (2) for 11.327% (eigenvalue 1.586) and factor (3) for 9.338% (eigenvalue 1.307). It is also worth noting that no cross-loading problem was observed in the table for any items to be deleted and assessing other possible options (decreasing or increasing the number of factors and using another rotation technique), it was decided to keep all items and conclude three factor solutions.

The rotated factor matrix for the whole set of 14 items of CSR with the corresponding three factor solutions are shown in Table 5.7. It is evident that the economic responsibility as an observed variable comprising 4 items has been rotated into a new latent variable with the same 4 items as before. Thus, the economic responsibility construct has remains unchanged with 4 items contributing to explaining the perceptions of mobile users of the three telecommunication organizations with corresponding items maximize long term success with loading 0.722, maximize stockholder's wealth with loading 0.640, minimize its cost with loading 0.541, and finally achieve maximum profit with loading 0.525.

Table 5.7: Total variance explained and rotated factor loading matrix (VARIMAX)

No.

Items

Factor

Commonalities

Factor 1

Factor 2

Factor 3

13-1

Sufficient monetary contribution to charities

.842

0.430

9-2

Helps government to solve social problems

.800

0.477

12-3

Engages in R&D to improve level of society

.799

0.532

14-4

Encourages customer to participate in charitable activities

.774

0.308

10-5

Always does what is right

.513

0.537

6-6

Doesn't commit with some rules and regulations

.696

0.434

5-7

Commitment to all rules and regulations

.646

0.453

8-8

Ethical principals over making profit

.644

0.587

7-9

Works under strict social laws

.621

0.675

11-10

Deals honestly with its clients

.549

0.466

3-11

Maximize long term success

.722

0.353

2-12

Maximize stock holder's wealth

.640

0.693

4-13

Minimize its cost

.541

0.727

1-14

Achieve maximum profit

.525

0.641

Total

Sum of squares (eigenvalues)

3.119

2.545

1.648

7.312

Percentage of trace

22.278

18.182

11.769

52.229

Note: Factor loading less than 0.4 have not been reproduced and items have been sorted by size on each factor.

Remarkably, the items of the ethical construct (4 items) have been fused into both the legal and philanthropic construct resulting in new latent variables namely philanthropic-ethical and legal-ethical constructs. The naming of these new constructs follows the suggestion by Hair et al. (2006) in that the items with higher loading contributes to naming the new latent variable. Thus, the new latent variable "philanthropic-ethical" comprises (5) items of which (3) items are philanthropic and (2) items are ethical. The items sufficient monetary contributions to charities have the highest loading (0.842) of the first factor (originally philanthropic construct) and the item helps government to solve social problems have the second highest loading (0.800) contributing to this factor (originally ethical construct) service for the new dimension namely philanthropic-ethical construct.

Similar finding have been concurred for the second factor solution resulting in a new latent variable that can be named "legal-ethical ". The legal-ethical construct comprises (5) items of which (3) are legal while (2) are ethical items. Thus, the new latent variable legal-ethical comprise the item with the highest loading (0.696) namely "doesn't commit with some rules and regulations" originally legal construct and the item "commitments to all rules and regulation" as the second highest loading (0.646) arising from legal construct too, while the item "ethical principals over making profit" for loading (0.644) arising from ethical construct.

**5.3.1 Validation of Factor Analysis**

As a final assessment step to the factor solutions is the degree of generalizability of the results to the population in which the sample was drawn. A direct approach to this assessment is to apply the same scale to new samples from the population (holdout sample). Due to time limitation of this research it was deemed somewhat difficult to replicate the analysis on an entirely new sample. Consequently, split sample analysis was chosen for the validation assessment.

Two randomly split samples were obtained from the original sample (265 respondents in each) and the factor analysis procedure was performed to compare the final factor solutions for each. The two factor solutions from each split sample were quite similar to each other in terms of loadings and commonalities for all of the items. It is worth noting that the VARIMAX rotation method had been used as per the original factor solutions. The results for both split samples have shown stability within the two samples. An alternative approach comparable to the split sample is to use a cross validation technique in which case the sample is divided into a ratio of (2/3 and 1/3). The first sample might be used for training purposes while the second sample is used for validation purpose. A confirmatory factor analysis will be required to authenticate the generalizability of this result across the population.

**5.4 Relation of New CSR and Perceived Service Quality**

In order to examine the positive relationship of corporate social responsibility new dimensions on perceived service quality, an ordinary least square regression technique (OLS) were employed. As a first step, a correlation matrix examining the relationship between each separate scale of the perceived service quality (technical, functional, and price) and the new dimension of corporate social responsibility obtained using a summated score of the factor solutions outlined in Section 5.3. The objective of this step is to observe if large correlation among the new dimension of corporate social responsibility exists and also with respect to the scale of perceived service quality. Table 5.8 displays the correlation matrix of the new CSR dimensions and scales of perceived service quality.

Table 5.8: correlation matrix of the new CSR scale and perceived service quality

EC

Leg-E

Ph-E

PSQT

PSQF

PSQP

Economic (EC)

1

Legal-Ethical (Leg-E)

.137**

1

Philanthropic-Ethical (Ph-E)

.085*

.533**

1

PSQ Technical

.216**

.331**

.212**

1

PSQ Functional

.205**

.488**

.361**

.462**

1

PSQ Price

.193**

.430**

.295**

.418**

.447**

1

**. Correlation is significant at the 0.01 level (2-tailed).

*. Correlation is significant at the 0.05 level (2-tailed).

As was expected a weak correlation between economic and both legal-Ethical and philanthropic-ethical (). On the contrary, a moderate significant correlation between legal-Ethical and philanthropic-ethical scale was observed (). Additionally, enough correlation between each scale of perceived service quality and the rest of the new CSR were also observed to run the multiple regressions modeling technique.

**5.4.1 Relationship Between new CSR and Perceived Technical Service Quality**

An ordinary least square multiple linear regression was performed to examine whether each scale of the new CSR is positively related with perceived technical service quality. Table 5.9 displays overall model significance individual variable significance relating new corporate social responsibility dimensions (economic, legal-ethical, philanthropic-ethical responsibilities) to average response of technical aspect of perceived service quality.

Overall the relationship classed as being statistically significant, and 14.1% of the variation in perceived service quality can be accounted for using a regression equation with the new CSR dimensions (i.e. EC, Leg-E, and Ph-E) as predictors (). However, examining the individual significance of each variable shows that philanthropic-ethical is classed as being insignificant predictor ().

Table 5.9: Overall and individual variable significance of new CSR and perceived technical service quality.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.375a

.141

.136

.86164

NA

**ANOVAb**

Model

Sum of Squares

df

Mean Square

Sig.

1

Regression

64.588

3

21.529

28.998

.000a

Residual

394.230

531

.742

Total

458.817

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Std. Error

Beta

1

(Constant)

1.439

.271

5.310

.000

EC

.254

.059

.174

4.283

.000

Leg-E

.351

.060

.281

5.884

.000

Ph-E

.049

.049

.048

.999

.318

a. Predictors: (Constant), Ph-E, EC, Leg-E

b. Dependent Variable: PSQT

A regression analysis was repeated excluding the philanthropic-ethical dimension and keeping the other two predictors variables only. In this case, the regression assumptions can be fully checked. The results of the final regression analysis relating average response of perceived technical service quality as the dependent variable to both new dimensions of CSR namely economic and legal-ethical being predictors variables are shown in Table 5.10 after excluding philanthropic-ethical scale.

Table 5.10: Final regressions summaries relating both legal-ethical and economic responsibilities of new CSR to average response for technical aspect of PSQ.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.373a

.139

.136

.86164

1.808

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

63.846

2

31.923

42.998

.000a

Residual

394.971

532

.742

Total

458.817

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

1.486

.267

5.568

.000

Economic

.255

.059

.175

4.298

.000

.981

1.019

Legal-Ethical

.383

.051

.307

7.553

.000

.981

1.019

a. Predictors: (Constant), Legal-Ethical, Economic

b. Dependent Variable: Average response for technical

It can be concluded that the overall estimated regression equation using economic and legal-ethical responsibilities as predictors of the technical aspect of perceived service quality is classed as being statistically significant relation and both predictor accounts for 13.9% of the variation in perceived technical service quality ().

The estimated regression equation can be described as follows:

Technical Aspect of PSQ = 1.486 + 0.255*Economic + 0.383*Legal-Ethical

To interpret the effect of economic responsibility, it can be said that under the fitted model a one unit increase in the economic responsibility is associated with an average increase in the technical aspect of perceived service quality by 0.255 unit () assuming the legal-ethical responsibility remains fixed. Similar conclusion can be made to describe the effect of legal-ethical responsibility on the technical aspect of perceived service quality. A final remark on the aforementioned estimated regression equation is to check the assumptions of regression.

As a first step, a one sample Kolmogorov-Smirnov test was performed to test whether the residual was normally distributed. The application of the one sample shows that the residual are not normally distributed (). As a second step, the multicollinearity has been assessed using the variance inflation factor (VIF) for each independent variable. Table 5.10 indicates the VIF value of either independent variable is (1.019) and that this value is less than 3. It can be concluded that the estimated regression equation does not suffer from multicollinearity, (Freund et al., 2006).

Other tests for the residuals had been conducted such as independence and equal variance; none of these tests have been violated. It can be concluded that there is a violation of the normality assumption for the residuals; however this violation does not render the ability of using the estimated regression equation to relate the empirically derived new constructs of CSR namely economic responsibility and legal-ethical to the technical aspect of perceived service quality but with due care for estimation.

Therefore, in regard to hypotheses (H1a) it can be concluded that we can reject null hypotheses that (ï¢ economic = ï¢ legal-ethical = ï¢ philanthropic-ethical = 0) and conclude that a positive relation exists between economic and legal-ethical responsibility and affecting the perceived technical service quality. However, we fail to reject the null hypotheses and conclude that there is no relation exists between philanthropic-ethical responsibility and perceived technical service quality.

**5.4.2 Relation Between new CSR and Perceived Functional Service Quality**

To investigate the positive relationship between the new dimensions of corporate social responsibility and the functional aspect of perceived service quality, an ordinary multiple linear regression was performed. Table 5.11 demonstrates the final regression summaries relating economic, legal-ethical, and philanthropic-ethical responsibilities of new CSR to the average response for functional aspect of PSQ.

It can be concluded that the overall estimated regression equation using economic, legal-ethical, and philanthropic-ethical responsibilities as predictors of the functional aspect of perceived service quality is classed as being statistically significant relation and these predictors accounts for 27.2% of the variation in the functional aspect of PSQ (). Moreover, examining the individual variables significant indicate that all predictor variables account for explaining the functional aspects of perceived service quality (all individual p-values < 0.05).

Table 5.11: Regressions summary of new CSR constructs (economic, legal-ethical, and philanthropic-ethical responsibilities) with average response for functional aspect of PSQ.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.521a

.272

.267

.63769

1.945

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

80.479

3

26.826

65.968

.000a

Residual

215.933

531

.407

Total

296.411

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

1.344

.201

6.699

.000

Economic

.163

.044

.139

3.719

.000

.981

1.019

Legal-Ethical

.396

.044

.395

8.971

.000

.707

1.414

Philanth.-Ethical

.114

.036

.139

3.170

.002

.716

1.398

a. Predictors: (Constant), Philanthropic-Ethical, Economic, Legal-Ethical

b. Dependent Variable: Average response for perceived service quality functional

For testing the assumptions of regression, a one sample Kolmogorov-Smirnov test was performed to test whether the residual was normally distributed. The application of the one sample K-S test shows that the residual are normally distributed (). The multicollinearity issue seems not to differ from the conclusion drawn from the previous section (see Table 5.11 for values of VIF). Other tests for the residuals had been conducted such as independence and equal variance; none of these tests have been violated. It appears that the regression assumptions underpinning the relation between the new dimensions of CSR and functional aspect of perceived service quality are partially satisfied. This implies that the estimated regression equation can be concluded as follows:

Functional Aspect of PSQ = 1.344 + 0.163*Economic + 0.396*Legal-Ethical + 0.114*Philanthropic-Ethical

To interpret the positive effect of legal-ethical responsibility, it can be said that under the fitted model a one unit increase in the legal-ethical responsibility is associated with an average increase in the functional aspect of perceived service quality by 0.396 unit () assuming that the economic and philanthropic-ethical responsibilities remain fixed. Likewise, similar conclusion can be made to describe the positive effect of economic and philanthropic-ethical responsibilities on the functional aspect of perceived service quality.

Thus, in regard to hypotheses (H1b) it can be concluded that we can reject the null hypotheses that (ï¢ economic = ï¢ legal-ethical = ï¢ philanthropic-ethical = 0) and conclude that a positive relation exists between the perception of corporate social responsibility represented by economic, legal-ethical, and philanthropic-ethical and perceived functional service quality.

**5.4.3 Relation Between new CSR and Perceived Price Service Quality**

To investigate the positive relationship between the new dimensions of corporate social responsibility and the price aspect of perceived service quality, an ordinary multiple linear regression was performed. Table 5.12 displays the final regression summaries relating economic, legal-ethical, and philanthropic-ethical responsibilities of new CSR to the average response for price aspect of PSQ. It can be concluded that the overall estimated regression equation using economic, legal-ethical, and philanthropic-ethical responsibilities as predictors of the price aspect of perceived service quality is classed as being statistically significant relation and these predictors accounts for 20.9% of the variation in the price aspect of PSQ ().

Table 5.12: Regressions summary of new CSR constructs (economic, legal-ethical, and philanthropic-ethical responsibilities) with average response for price aspect of PSQ.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.457a

.209

.205

.88773

1.921

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

110.769

3

36.923

46.853

.000a

Residual

418.463

531

.788

Total

529.232

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

.746

.279

2.672

.008

Economic

.212

.061

.135

3.475

.001

.981

1.019

Legal-Ethical

.488

.062

.364

7.933

.000

.707

1.414

Philanthropic-Ethical

.099

.050

.090

1.969

.049

.716

1.398

a. Predictors: (Constant), Philanthropic-Ethical, Economic, Legal-Ethical

b. Dependent Variable: average response for perceived service quality for price

Moreover, examining the individual variables indicate that all predictor variables classed as being statistically significant and they account for explaining the price aspects of perceived service quality (all individual p-values < 0.05).

A one sample Kolmogorov-Smirnov test was performed to test whether the residual were normally distributed in order to test the regression assumption. The application of the one sample K-S test shows that the residual are not normally distributed (). Also, the multicollinearity has been assessed using the variance inflation factor (VIF) for each independent variable. Table 5.12 indicates the VIF values of all predictors variables are (1.019, 1.414, and 1.398; respectively) and these values are less than 3. This concludes that no serious correlation among predictors (Freund et al., 2006). Other tests for the residuals had been conducted such as independence and equal variance; none of these tests have been violated. It appears that the regression assumptions underpinning the relation between the new dimensions of CSR and price aspect of perceived service quality are all satisfied. This implies that the estimated regression equation can be concluded as follows:

Price Aspect of PSQ = 0.746 + 0.212*Economic + 0.488*Legal-Ethical + 0.099*Philanthropic-Ethical

To interpret the positive effect of legal-ethical responsibility, it can be said that under the fitted model a one unit increase in the legal-ethical responsibility is associated with an average increase in the price aspect of perceived service quality by 0.488 unit () assuming that the economic and philanthropic-ethical responsibilities remain fixed. Likewise, similar conclusion can be made to describe the positive effect of economic and philanthropic-ethical responsibilities on the price aspect of perceived service quality.

Thus, in regard to hypotheses (H1c) it can be concluded that we can reject the null hypotheses that (ï¢ economic = ï¢ legal-ethical = ï¢ philanthropic-ethical = 0) and conclude that a positive relation exists between perception of corporate social responsibility represented by economic, legal-ethical, and philanthropic-ethical responsibility and perceived price service quality.

**5.5 Relation of New CSR and Customer Loyalty**

In order to examine the positive relationship of CSR new dimensions on customer loyalty (behavioral and attitudinal) an ordinary regression modeling technique were employed. A preliminary step is to examine the correlation between each separate aspects of the customer loyalty and the new dimension of CSR obtained using a summated score of the factor solutions outlined in Section 5.3. Table 5.13 displays the correlation matrix of the new CSR dimensions and aspects of customer loyalty.

Table 5.13: correlation matrix of the new CSR scale and customer loyalty

EC

LegE

PhE

CLB

CLA

EC

1

LegE

.137**

1

PhE

.085*

.533**

1

CLB

.215**

.515**

.398**

1

CLA

.142**

.495**

.387**

.769**

1

**. Correlation is significant at the 0.01 level (2-tailed).

*. Correlation is significant at the 0.05 level (2-tailed).

A weak correlation was detected between economic and both behavioral and attitudinal aspects of customer loyalty (). On Contrary, moderate and statistically significant correlation between legal-Ethical scale and both behavioral and attitudinal aspects of customer loyalty were detected (). Likewise, moderate significant correlation between philanthropic-Ethical scale and both behavioral and attitudinal aspects of customer loyalty were detected.

**5.5.1 Relation Between new CSR and Behavioral Customer Loyalty**

To investigate the positive relationship between the new dimensions of corporate social responsibility and the behavioral aspect of customer loyalty, an ordinary multiple linear regression was performed. Table 5.14 displays the final regression summaries relating economic, legal-ethical, and philanthropic-ethical responsibilities of new CSR to the average response of behavioral aspect of customer loyalty.

Table 5.14: Regressions summary of new CSR constructs (economic, legal-ethical, and philanthropic-ethical responsibilities) with average response for behavioral aspect of CL.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.555a

.308

.304

.69430

1.960

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

113.662

3

37.887

78.597

.000a

Residual

255.968

531

.482

Total

369.630

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

.830

.218

3.802

.000

Economic

.191

.048

.146

3.994

.000

.981

1.019

Philant.-Ethical

.156

.039

.170

3.982

.000

.716

1.398

Legal-Ethical

.453

.048

.405

9.423

.000

.707

1.414

a. Predictors: (Constant), Legal-Ethical, Economic, Philanthropic-Ethical

b. Dependent Variable: Average CL Behavior

It can be concluded that the overall estimated regression equation using economic, legal-ethical, and philanthropic-ethical responsibilities as predictors of the behavioral aspect of customer loyalty is classed as being statistically significant relation and these relation accounts for 30.8% of the variation in the behavioral aspect of customer loyalty ().

Moreover, examining the individual variables indicate that all predictor variables classed as being statistically significant predictor variables and they account for explaining the behavioral aspect of customer loyalty (all individual -values < 0.05). A one sample Kolmogorov-Smirnov test was performed to test whether the residual were normally distributed in order to test the regression assumption. The application of the one sample K-S test shows that the residual are not normally distributed (). In addition, the collinearity issue was deemed not to affect the estimated regression outlined next.

Other tests for the residuals had been conducted such as independence and equal variance; none of these tests have been violated. It appears that the regression assumptions have been violated when testing the normality of residuals. This implies that the estimated regression equation can be used with due care and it is described as follows:

Behavioral Aspect of Customer Loyalty = 0.830 + 0.191*Economic + 0.453*Legal-Ethical + 0.156*Philanthropic-Ethical

To interpret the positive effect of legal-ethical responsibility, it can be said that under the fitted regression model a one unit increase in the legal-ethical responsibility is associated with an average increase in the behavioral aspect of customer loyalty by 0.453 unit () assuming that the economic and philanthropic-ethical responsibilities remain fixed. Likewise, similar conclusion can be made to describe the positive effect of economic and philanthropic-ethical responsibilities on the behavioral aspect of customer loyalty.

Therefore, in regard to hypotheses (H2a) it can be concluded that we can reject the null hypotheses that (ï¢ economic = ï¢ legal-ethical = ï¢ philanthropic-ethical = 0) and conclude that a positive relation exists between the perception of corporate social responsibility represented by economic, legal-ethical, and philanthropic-ethical responsibility and behavioral customer loyalty.

**5.5.2 Relation Between new CSR and Attitudinal Customer Loyalty**

To investigate the positive relationship between the new dimensions of corporate social responsibility and the attitudinal aspect of customer loyalty, an ordinary multiple linear regression was performed. Table 5.15 displays the final regression summaries relating economic, legal-ethical, and philanthropic-ethical responsibilities of new CSR to the average response of attitudinal aspect of customer loyalty. It can be concluded that the overall estimated regression equation using new CSR constructs as predictors of the attitudinal aspect of customer loyalty is classed as being statistically significant relation and these relation accounts for 27.2% of the variation in the attitudinal aspect of customer loyalty ().

Moreover, examining the individual variables indicate that all predictor variables classed as being statistically significant predictor variables and they account for explaining the attitudinal aspect of customer loyalty (all individual -values ). A one sample Kolmogorov-Smirnov test was performed to test whether the residual were normally distributed in order to test the regression assumption.

Table 5.15: Regressions summary of new CSR constructs (economic, legal-ethical, and philanthropic-ethical responsibilities) with average response for attitudinal aspect of CL.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.521a

.272

.267

.82874

1.910

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

135.958

3

45.319

65.986

.000a

Residual

364.694

531

.687

Total

500.652

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

.460

.261

1.766

.078

Economic

.112

.057

.073

1.965

.050

.981

1.019

Philant.-Ethical

.183

.047

.171

3.913

.000

.716

1.398

Legal-Ethical

.513

.057

.394

8.935

.000

.707

1.414

a. Predictors: (Constant), Legal-Ethical, Economic, Philanthropic-Ethical

b. Dependent Variable: Average Response of CL Attitudinal

The application of the one sample K-S test shows that the residual are normally distributed (). In addition, the collinearity issue was deemed not to affect the estimated regression equation outlined next:

Attitudinal aspect of Customer Loyalty = 0.460 + 0.112*Economic + 0.513*Legal-Ethical + 0.183*Philanthropic-Ethical

To interpret the positive effect of legal-ethical responsibility, it can be said that under the fitted regression model a one unit increase in the legal-ethical responsibility is associated with an average increase in the attitudinal aspect of customer loyalty by 0.513 unit () assuming that economic and philanthropic-ethical responsibilities remain fixed. Likewise, similar conclusions can be made to describe the positive effect of economic and philanthropic-ethical responsibilities on attitudinal aspect of customer loyalty.

Therefore, in regard to hypotheses (H2b) it can be concluded that we can reject the null hypotheses that (ï¢ economic = ï¢ legal-ethical = ï¢ philanthropic-ethical = 0) and conclude that a positive relation exists between the perception of corporate social responsibility represented by economic, legal-ethical, and philanthropic-ethical responsibility and attitudinal customer loyalty.

**5.6 Effect of Perceived Service Quality on Customer Loyalty**

In order to examine the positive relationship between the perceived service quality (technical, functional, and price) and customer loyalty (behavioral and attitudinal) an ordinary least square regression modeling technique were employed. A preliminary step is to examine the correlation between each separate aspects of the perceived service quality and the aspects of customer loyalty. Table 5.16 demonstrates the correlation matrix of aspects of perceived service quality and aspects of customer loyalty.

Table 5.16: correlation matrix between perceived service quality and customer loyalty

PSQT

PSQF

PSQP

CLB

CLA

PSQT

1

PSQF

.462**

1

PSQP

.418**

.447**

1

CLB

.497**

.565**

.483**

1

CLA

.539**

.493**

.479**

.769**

1

**. Correlation is significant at the 0.01 level (2-tailed).

A moderate and statistically significant correlation between technical aspect of perceived service quality and both behavioral and attitudinal aspects of customer loyalty were detected (). Similar moderate statistically significant correlations were detected between functional and prices aspects of perceived service quality and both aspects of customer loyalty.

**5.6.1 Effect of Perceived Service Quality on Behavioral aspect of Customer Loyalty**

To examine the positive relationship between aspects of perceived service quality and the behavioral aspect of customer loyalty, an ordinary multiple linear regression was performed. Table 5.17 demonstrates the final regression summaries relating average aspects of PSQ (technical, functional, and price) to the average response of behavioral aspect of customer loyalty. The overall estimated regression equation using aspects of perceived service quality as predictors of the behavioral aspect of customer loyalty is classed as being statistically significant relation and these relation accounts for 42.8% of the variation in the behavioral aspect of customer loyalty ().

Moreover, all predictor variables seem to be statistically significant predictor variables and they account for explaining the behavioral aspect of customer loyalty (all individual -values ). A one sample Kolmogorov-Smirnov test was performed to test whether the residual were normally distributed in order to test the regression assumption.

Table 5.17: Regressions summary of the effect of perceived service quality (technical, functional, price) on the average response for behavioral aspect of Customer loyalty.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.654a

.428

.425

.63102

1.916

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

158.195

3

52.732

132.431

.000a

Residual

211.435

531

.398

Total

369.630

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

.643

.148

4.348

.000

Av. Resp. Tech.

.215

.034

.240

6.253

.000

.731

1.368

Av. Resp. Func.

.395

.044

.354

9.078

.000

.709

1.411

Av. Resp. Price

.188

.032

.225

5.906

.000

.744

1.345

a. Predictors: (Constant), average response for technical, functional, price aspects of perceived service quality

b. Dependent Variable: Average behavioral customer loyalty

The application of the one sample K-S test shows that the residual are normally distributed (). All VIF values for each predictor variable are all less than 3, which indicate that no serious colinearity among the predictors. Other assumptions for the regression equation have been performed, and it can be concluded that all assumptions have been fulfilled. Thus, the estimated regression equation can be described as follows:

Behavioral aspect of Customer Loyalty = 0.643 + 0.215*Technical + 0.395*Functional + 0.188*Price

Under the fitted regression model a one unit increase in the functional aspect of perceived service quality is associated with an average increase in the behavioral aspect of customer loyalty with a positive effect amounted to be 0.395 unit () assuming that technical and price aspects of perceived service quality remain fixed. Likewise, similar conclusions can be made to describe the positive effect of technical and price aspects of perceived service quality on behavioral aspect of customer loyalty.

Thus, in regard to hypotheses (H3a) it can be concluded that we can reject the null hypotheses that (ï¢ technical = ï¢ functional = ï¢ price = 0) and conclude that a positive relation exists between the perceived service quality represented in three aspects technical, functional, and price and behavioral customer loyalty.

**5.6.2 Effect of Perceived Service Quality on Attitudinal aspect of Customer Loyalty**

To examine the positive relationship between aspects of perceived service quality and the attitudinal aspect of customer loyalty, an ordinary multiple linear regression was performed. Table 5.18 displays the final regression summaries relating average aspects of PSQ (technical, functional, and price) to the average response of attitudinal aspect of customer loyalty. The overall estimated regression equation using aspects of perceived service quality as predictors of the attitudinal aspect of customer loyalty is classed as being statistically significant relation and these relation accounts for 40.7% of the variation in the attitudinal aspect of customer loyalty (). Moreover, all predictor variables are classed as being statistically significant predictor variables and they account for explaining the attitudinal aspect of customer loyalty (all individual -values ).

Table 5.18: Regressions summary of the effect of perceived service quality (technical, functional, price) on the average response for attitudinal aspect of Customer loyalty.

**Model Summaryb**

Model

Square

Adjusted Square

Std. Error of the Estimate

Durbin-Watson

1

.638a

.407

.404

.74745

2.074

**ANOVAb**

Model

Sum of Squares

Df

Mean Square

Sig.

1

Regression

203.993

3

67.998

121.711

.000a

Residual

296.659

531

.559

Total

500.652

534

**Coefficientsa,b**

Model

Unstandardized Coefficients

Standardized Coefficients

Sig.

Collinearity Statistics

Std. Error

Beta

Tolerance

VIF

1

(Constant)

-.063

.175

-.358

.720

Av. Resp. Tech.

.347

.041

.333

8.512

.000

.731

1.368

Av. Resp. Func.

.304

.052

.234

5.897

.000

.709

1.411

Av. Resp. Price

.229

.038

.236

6.082

.000

.744

1.345

a. Predictors: (Constant), average response for technical, functional, price aspects of perceived service quality

b. Dependent Variable: Average CL Attitudinal

A one sample Kolmogorov-Smirnov test was performed to test whether the residual were normally distributed in order to test the regression assumption. The application of the one sample K-S test shows that the residual are normally distributed (). All VIF values for each predictor variable are all less than 3, which indicate that no serious colinearity among the predictors. Other assumptions for the regression equation have been performed, and it can be concluded that all assumptions have been fulfilled.

Thus, the estimated regression equation can be described as follows:

Attitudinal aspect of Customer Loyalty = 0.347*Technical + 0.304*Functional + 0.229*Price

Under the fitted regression model a one unit increase in the technical aspect of perceived service quality is associated with an average increase in the attitudinal aspect of customer loyalty with a positive effect amounted to be 0.347 unit () assuming that functional and price aspects of perceived service quality remain fixed. Likewise, similar conclusions can be made to describe the positive effect of functional and price aspects of perceived service quality on attitudinal aspect of customer loyalty.

Hence, in regards to hypotheses (H3b) it can be concluded that we can reject the null hypotheses that (ï¢ technical = ï¢ functional = ï¢ price = 0) and conclude that a positive relation exists between perceived service quality represented in three aspects technical, functional, and price and attitudinal customer loyalty.

**5.7 Chapter Summary**

In this chapter, the use of exploratory factor analysis and ordinary least square regression have been utilized to examine the stated hypotheses outlined in Chapter (4), Section (4.5). Both methods were able to derive a valid and concise decision regarding each hypotheses and new structure for the scale of corporate social responsibility has been proposed.

The exploratory factor analysis procedure outlined in Section 5.3 provides a new structure for corporate social responsibility (CSR) comprising three new responsibilities constructs namely; economic, legal-ethical responsibility, and philanthropic-ethical responsibility. The economic responsibility has (4) items, while newly established legal-ethical construct has (5) items and newly established construct philanthropic-ethical has also (5) items. In light of the first proposed question in this research (see section 4.3).

These constructs are relatively new in the literature and believed to be a novel contribution. In summary, while the results of examining the structure of CSR through factor analysis outlined in Section (5.3) presented a reasonable structure for the CSR scale, there is surely a need for more confirmatory research based on these results. In particular, exploratory and confirmatory studies conducted in different sectors apart from telecommunication.

The ordinary least square (OLS) regression assessment outlined