**4. Methodology**

To identify the most appropriate scales for measuring the relationships proposed with a high degree of reliability, validity, and dimensionality, it developed a process with several stages [51]. The first stage consisted of constructing scales with a good degree of content validity. It made an exhaustive literature review, taking as reference the scales constructed in various prior studies of affective evaluation [52, 53], symbolic benefit [54, 55], functional benefit [56], hedonic benefit [56, 57], and satisfaction [58–60].

A study of critical incidents was done, in which people were asked to describe the factors that formed part of the constructs analyzed. Fifty people chosen by nonprobabilistic convenience sampling participated in this study, from which was obtained the prior scales of the study constructs. A second procedure was done to purify these scales, following the recommendations of [61]. There was also a series of "focus groups" composed of regular clients from different banks in Chile, as well as interviews with business experts and executives from the banking industry. These analyses permitted us to identify the indicators that best reflect the dimensions needed for this study, reformulating and/or eliminating indicators that were problematic or repetitive using a modification of the [62]. Each expert was asked to classify items relative to their dimension, considering three alternatives: clearly, somewhat, or not at all representative. Finally, it was decided to preserve the items in which there was a high level of consensus [63]. From these analyses, it obtained the scales that enabled us to develop the questionnaire. In the second stage, after constructing the questionnaire, it performed a quantitative pre-test with a random sampling of 50 people. Using these data, it calculated an exploratory factor analysis and the Alpha Cronbach for each of the resulting dimensions. This initial analysis enabled us to confirm the existence of each dimension resulting from the preceding analyses. The items were written as statements (see **Table 1**) to be evaluated using a 7-point Likert scale.

It used nonprobabilistic quota sampling proportional to the market participation of the first three leading banks in the private banking sector in Chile: Santander Bank (18%); Chile Bank (17%), and BCI Bank (15%)<sup>1</sup> . The surveys were administered outside the main branches of each bank, especially in the downtown of Santiago.

that did not adhere well to their dimension, it made various principal components factor analyses with Varimax rotation [64]. This procedure did not indicate the need to eliminate any indicator from the scales analyzed (see **Table 2**); all scales showed a good degree of one-

Taking into account all of the scales constructed, it developed a CFA using structural equations modeling (SEM) to confirm whether the indicators or variables were appropriate for achieving good model fit. It used the three criteria proposed by [66]. The first criterion consists of eliminating indicators that had a weak convergence condition with their corresponding latent variable. It took as limit a t-student value higher than 2.28 (p = 0.01). The second criterion consisted of eliminating from the analysis variables whose loadings translated into standardized coefficients lower than 0.5. Finally, it also had to eliminate the indicators that showed a linear relationship

22. Following this analysis, it was not necessary to eliminate any indicator. The fit indices for the confirmatory factor analysis were acceptable: IFI = 0.948, CFI = 0.948, RMSEA = 0.078, Normed χ<sup>2</sup> = 5.81. Once verified the optimal model, it confirmed the reliability of each scale by applying three tests: Alpha Cronbach (limit = 0.7), composite reliability of the construct (limit = 0.7) [67],

lower than 0.3. In this process, it used the statistical package AMOS SPSS

Antecedents of Satisfaction with Financial Services: Role of Perceived Benefits

http://dx.doi.org/10.5772/intechopen.74653

101

dimensionality, with factor loadings well over 0.4 [65].

with goodness of fit, R<sup>2</sup>

**Table 1.** Measurement scales.

In the third stage, the data was collected. The survey was applied to a total sample of 850 people. It has, however, to eliminate 64 forms because they were incomplete and/or incorrectly answered, leaving a definitive sample of 786 people. For the responses, the last visit the bank was taken as a reference at which the respondents were customers. Of the total interviewees, 64% were men and 49% were married (83% were 25–54 years old, 84% held university degrees, 89% worked full time, and 81% had monthly incomes over 750,000 Chilean pesos (1.114 USD)<sup>2</sup> .

With the data obtained, it has been made a psychometric analysis to develop scales with a good degree of reliability, validity, and dimensionality. It applied exploratory factor analysis, confirmatory factor analysis (CFA), and various reliability analyses with the Alpha Cronbach statistics, reliability construct, and average variance extracted (AVE). To identify the items

<sup>1</sup> Risk rating ICR. Knowledge & Trust, December 2016.

<sup>2</sup> Exchange rate december 23, 2016.


**Table 1.** Measurement scales.

**4. Methodology**

100 Marketing

hedonic benefit [56, 57], and satisfaction [58–60].

statements (see **Table 1**) to be evaluated using a 7-point Likert scale.

each bank, especially in the downtown of Santiago.

Risk rating ICR. Knowledge & Trust, December 2016.

Exchange rate december 23, 2016.

(17%), and BCI Bank (15%)<sup>1</sup>

1

2

To identify the most appropriate scales for measuring the relationships proposed with a high degree of reliability, validity, and dimensionality, it developed a process with several stages [51]. The first stage consisted of constructing scales with a good degree of content validity. It made an exhaustive literature review, taking as reference the scales constructed in various prior studies of affective evaluation [52, 53], symbolic benefit [54, 55], functional benefit [56],

A study of critical incidents was done, in which people were asked to describe the factors that formed part of the constructs analyzed. Fifty people chosen by nonprobabilistic convenience sampling participated in this study, from which was obtained the prior scales of the study constructs. A second procedure was done to purify these scales, following the recommendations of [61]. There was also a series of "focus groups" composed of regular clients from different banks in Chile, as well as interviews with business experts and executives from the banking industry. These analyses permitted us to identify the indicators that best reflect the dimensions needed for this study, reformulating and/or eliminating indicators that were problematic or repetitive using a modification of the [62]. Each expert was asked to classify items relative to their dimension, considering three alternatives: clearly, somewhat, or not at all representative. Finally, it was decided to preserve the items in which there was a high level of consensus [63]. From these analyses, it obtained the scales that enabled us to develop the questionnaire. In the second stage, after constructing the questionnaire, it performed a quantitative pre-test with a random sampling of 50 people. Using these data, it calculated an exploratory factor analysis and the Alpha Cronbach for each of the resulting dimensions. This initial analysis enabled us to confirm the existence of each dimension resulting from the preceding analyses. The items were written as

It used nonprobabilistic quota sampling proportional to the market participation of the first three leading banks in the private banking sector in Chile: Santander Bank (18%); Chile Bank

In the third stage, the data was collected. The survey was applied to a total sample of 850 people. It has, however, to eliminate 64 forms because they were incomplete and/or incorrectly answered, leaving a definitive sample of 786 people. For the responses, the last visit the bank was taken as a reference at which the respondents were customers. Of the total interviewees, 64% were men and 49% were married (83% were 25–54 years old, 84% held university degrees, 89% worked full time, and 81% had monthly incomes over 750,000 Chilean pesos (1.114 USD)<sup>2</sup>

With the data obtained, it has been made a psychometric analysis to develop scales with a good degree of reliability, validity, and dimensionality. It applied exploratory factor analysis, confirmatory factor analysis (CFA), and various reliability analyses with the Alpha Cronbach statistics, reliability construct, and average variance extracted (AVE). To identify the items

. The surveys were administered outside the main branches of

.

that did not adhere well to their dimension, it made various principal components factor analyses with Varimax rotation [64]. This procedure did not indicate the need to eliminate any indicator from the scales analyzed (see **Table 2**); all scales showed a good degree of onedimensionality, with factor loadings well over 0.4 [65].

Taking into account all of the scales constructed, it developed a CFA using structural equations modeling (SEM) to confirm whether the indicators or variables were appropriate for achieving good model fit. It used the three criteria proposed by [66]. The first criterion consists of eliminating indicators that had a weak convergence condition with their corresponding latent variable. It took as limit a t-student value higher than 2.28 (p = 0.01). The second criterion consisted of eliminating from the analysis variables whose loadings translated into standardized coefficients lower than 0.5. Finally, it also had to eliminate the indicators that showed a linear relationship with goodness of fit, R<sup>2</sup> lower than 0.3. In this process, it used the statistical package AMOS SPSS 22. Following this analysis, it was not necessary to eliminate any indicator. The fit indices for the confirmatory factor analysis were acceptable: IFI = 0.948, CFI = 0.948, RMSEA = 0.078, Normed χ<sup>2</sup> = 5.81. Once verified the optimal model, it confirmed the reliability of each scale by applying three tests: Alpha Cronbach (limit = 0.7), composite reliability of the construct (limit = 0.7) [67],


**Table 2.** Exploratory factorial analysis of the scales.

and analysis of the variance extracted (limit = 0.5) [68]. The results show that the minimum values established for these reliability parameters are fulfilled in all cases (see **Table 3**).

**Table 3.** Reliability analysis of scales.

Antecedents of Satisfaction with Financial Services: Role of Perceived Benefits

http://dx.doi.org/10.5772/intechopen.74653

103

**Table 4.** Discriminant validity.

The validity of the SEM model was confirmed using content and construct validity. The scales show a good degree of content validity, since it made a thorough analysis of the literature, a study of critical incidents with customers, and then a purification of the scale using different "focus groups" with customers and in-depth interviews with commercial experts and executives from different banks. To fulfill construct validity, it analyzed if the proposal, purified scale fulfilled the conditions for convergent and discriminant validity. Convergent validity was confirmed by observing that all of the standardized coefficients of the CFA were statistically significant at 0.01 and higher than 0.5 [69]. To confirm the presence of discriminant validity, it was used the confidence interval test [70], which consists of constructing the confidence intervals resulting from the correlations between the different latent variables that compose the CFA model (see **Table 4**). This test indicates the model's discriminant validity, since no confidence interval contained the value of 1 [71].

With all of these antecedents, it concluded that the proposes model shows a good degree of general validity.


**Table 3.** Reliability analysis of scales.

and analysis of the variance extracted (limit = 0.5) [68]. The results show that the minimum val-

The validity of the SEM model was confirmed using content and construct validity. The scales show a good degree of content validity, since it made a thorough analysis of the literature, a study of critical incidents with customers, and then a purification of the scale using different "focus groups" with customers and in-depth interviews with commercial experts and executives from different banks. To fulfill construct validity, it analyzed if the proposal, purified scale fulfilled the conditions for convergent and discriminant validity. Convergent validity was confirmed by observing that all of the standardized coefficients of the CFA were statistically significant at 0.01 and higher than 0.5 [69]. To confirm the presence of discriminant validity, it was used the confidence interval test [70], which consists of constructing the confidence intervals resulting from the correlations between the different latent variables that compose the CFA model (see **Table 4**). This test indicates the model's discriminant validity, since no

With all of these antecedents, it concluded that the proposes model shows a good degree of

ues established for these reliability parameters are fulfilled in all cases (see **Table 3**).

confidence interval contained the value of 1 [71].

**Table 2.** Exploratory factorial analysis of the scales.

general validity.

102 Marketing


**Table 4.** Discriminant validity.
