Ethical Sales Behavior, Relationship Quality, and Customer Loyalty
Research Methodology
This study is exploratory, quantitative in nature, aiming to develop a better understanding of the relationships
among the ethical sales behavior, relationship quality and customer loyalty to the bank. More specifically, the
study intends to empirically investigate the direct and indirect effect of ethical sales behavior as perceived by
customer on his loyalty to the bank through customer trust and customer commitment as mediators.
Selection of Sample and Respondents Demographics
The proposed research model is tested in the context of bank industry because financial services are highly abstract services characterized by credence attributes and consequently difficult for consumers to fully understand. Therefore, the customer must rely on the bank salesperson for correct information and proper guidance. Consequently, bank industry can adequately reflect the actual characteristics of ethical sales behavior relationships. Accordingly, the study is empirical based on the primary data collected from a sample of bank customers in Amman – Jordan.
The study was carried out in
2010 on a convenience sample from four main Jordanian commercial banks: Arab Bank, Jordan-Kuwait Bank,
Housing Bank for Trade and Finance and Ahli National Bank. Questionnaires were administered to a total of 320
bank customers within the area of Amman. A total of 226 respondents returned surveys, of which nine
questionnaires were rejected due to the lack of some information. Thus, 217 valid questionnaires were finally
obtained, giving response rate of 67.8 per cent, a rate that is regarded as good. Since the questionnaire was
administered in Arabic, the questionnaire was drafted in English and translated into Arabic thereafter.
The participants in this study were predominantly males (54.4 percent), majority (53.4 percent) of the
respondents were ages up to 40 years old. In terms of education, most of them (65.4 percent) were university
graduates (possessed a bachelor degree) and the majority of the respondents work in the private sector (49.3
percent). Finally (55.1 percent) of the respondents were married.
Data Analysis
The statistical package SPSS (version 19.0) was used for data analysis. A two-step detailed statistical analysis of data was involved. First, factor analysis was performed to extract the underlying factor of study variables. Second; a structural equation modeling was conducted using AMOS 7 to test the hypotheses in order to understand the direct and indirect effect of ethical sales behavior on - customer loyalty, customer trust and customer commitment.
Measures
All constructs in this study (i.e. ethical sales behavior, commitment, and trust and customer loyalty) were measured using multi-item Likert scale adapted from prior research previously validated in other contexts to ensure content validity. All items were measured using seven-point Likert-type scales. The respondents were asked to indicate their agreement or disagreement with the statements provided, with anchors of 1 to indicate 'strongly disagree' and 7 to indicate 'strongly agree'. The five items used to measure ethical sales behavior were adapted from those originally developed by Román to assess ethical sales behavior from the customer perspective in the Spanish financial sector.
This scale was also adapted next by Román and Ruiz; Chen
and Mau; Hansen & Riggle. All items were negatively worded and subsequently reverse coded.
The measure for trust were adapted from Morgan and Hunt; Karande et al.; Sanzo et al, 2003;
Hansen and Riggle; Kennedy et al.; Chen and Mau; Sirdeshmukh et al.;
Eisingerich and Bell.The eight items measure of customer trust designed to capture the major dimensions
of trust: confidence, reliability, and integrity. These items were utilized as well, focusing specifically on those
that capture respondents' perceptions as to the integrity and reliability of the bank/salesperson.
Measure Reliability
Different procedures were carried out using SPSS 19 to examine the psychometric properties of the proposed
measurements. As the first step in analysis of the scale, internal reliability for the adapted scale was compared
to that reported in the developmental literature. Reliability analysis, understood as internal consistency, was
carried out by calculating Cronbach's α coefficient .However, all the individual scales exceeded the
recommended standards level of .70 proposed by Nunnally and Hair et al. As depicted in Table 1,
the Cronbach alphas coefficients demonstrating the reliability of the scale used for measuring the constructs are
at acceptable levels and fall between 0.865 for ethical sales behavior and 0.942 for customer loyalty, which
suggests a high internal consistency among the items in each construct.
Table 1. Cronbach's alpha coefficient of study variables
| No. | Dimension | Items Number | Coefficient |
|---|---|---|---|
| 1 | Ethical sales behavior | 5 | 0.865 |
| 2 | Customer trust in the sales person | 8 | 0.931 |
| 3 | Customer Commitment | 7 | 0.898 |
| 4 | Customer loyalty | 6 | 0.942 |
| All Dimensions | 26 | 0.960 |
As can be seen in Table 1 these Cronbach alphas indicate that the scales used in the questionnaire satisfactorily
measured the constructs, what means that the ability of the set of items employed to represent each of the latent
constructs is satisfactory. However, unidimensionality of each construct must be checked. Therefore, items in
each multi-item scale were factor analyzed separately using principal component factor analysis with Varimax
rotation. The criteria for choosing variables are based on Kaiser's suggestions: an eigenvalue greater than
1 after Varimax rotation, absolute values of factor loadings greater than .50. As shown in Table 2, results
indicate that in all case a single factor emerged, i.e. there is one factor derived from each variable. Ethical sales
behavior (eigenvalue = 3.264); customer loyalty (eigenvalue = 4.664); customer trust (eigenvalue = 5.393); and
customer commitment (eigenvalue = 4.355). Furthermore, all of the items in the research model had factor
loadings greater than 0.70 (0.73 – 0.91).
Table 2. Construct measurement summary
| Statements | Factor loading | Variance Explained % |
|---|---|---|
| Ethical sales behavior ( Cronbach Alpha: 0.865 ) | 65.284 | |
| 1. The Bank salesperson lies about availability in order to make a sale (R). | .784 | |
| 2. The Bank salesperson lies about competition in order to make the sale (R). | .788 | |
| 3. The Bank salesperson gives answers when he/she does not really know the answers (R). |
.841 | |
| 4. The Bank salesperson applies sales pressure even though he/she knows the product is not right for me (R). | .861 | |
| 5. The Bank salesperson paints rosy pictures of the products to make them sound as good as possible (R). |
.762 | |
| Customer trust ( Cronbach Alpha:0.931 ) | 67.408 | |
| 6. I believe that the Bank has my interest in mind. | .806 | |
| 7. I feel that the Bank is trustworthy. | .819 |
|
| 8. I have confidence of the products and services of this Bank. | .819 | |
| 9. Generally speaking, you do not have to worry at all in dealing with this Bank. |
.778 | |
| 10. The Bank can be relied on to keep their promises. | .857 | |
| 11. T11. I believe the information that the Bank provides. | .839 | |
| 12. This Bank is honest and truthful. | .873 | |
| 13. This Bank can be counted on to do what is right. | .772 | |
| Customer Commitment ( Cronbach Alpha: 0.898 ) | 62.216 |
|
| 14. I want to remain a customer of this Bank because I genuinely enjoy my relationship with. |
.730 | |
| 15. I have a strong sense of loyalty towards this Bank. | .807 | |
| 16. My relationship with this Bank is one that I really care about. | .801 | |
| 17. I feel a strong sense of belonging to this Bank. | .780 | |
| 18. I am very committed to the relationship with this Bank. | .788 | |
| 19. My relationship with this Bank is very important to me. | .807 | |
| 20. My relationship with this Bank is something I intend to maintain Indefinitely. | .806 |
|
| Customer loyalty( Cronbach Alpha: 0.942 ) | 77.734 | |
| 21. I am willing to place my future business with this Bank. | .882 | |
| 22. This Bank is my first choice when I want to re buy or re patronize a preferred product or service in the future. |
.820 | |
| 23. I will recommend the Bank to my friends and relatives. | .915 | |
| 24. I will introduce the Bank with positive evaluation. | .916 | |
| 25. I feel satisfied with placing a contract with this Bank. | .885 | |
| 26. I am willing to encourage others to do business with this Bank. | .867 |
However, both tests of principal component analysis and Cronbach's alpha support unidimensionality and
therefore were accepted as being reliable for the path analysis. In addition, a correlation matrix of the variables
in the model was constituted and fall between 0.44 and 0.89, as shown in Table 3, suggesting several highly
significant correlations between the variables included in the conceptual model. Descriptive statistics and
correlations among variables are presented in Table 3.
Table 3. Correlation matrix and descriptive statistics of study variables (n= 217)
| Ethical sales behavior | Trust | Commitment | Loyalty | |
|---|---|---|---|---|
| Ethical sales behavior | 1 | |||
| Trust | .663(**) |
1 | ||
| Commitment | 659(**) |
.891(**) | 1 | |
| Loyalty | .440(**) |
.748(**) | .716(**) | 1 |
| Mean | 5.2009 | 4.8992 | 5.000 | 4.472 |
| Std. deviation | 1.051 | 1.034 | 0.949 | 1.337 |
Thereafter, Following Anderson and Gerbing, to assess construct validity of the multi-item scales
confirmatory factor analysis (CFA) was conducted in order to determine if the extracted dimensions in step 1
offered a good fit to the data (Table 4).
Table 4. Confirmatory factor analysis: measurement property
| Estimate | S.E. | C.R. | P | Label | |||
|---|---|---|---|---|---|---|---|
| E5 | <--- | Ethical Behavior | .707 | ||||
| E4 | <--- | Ethical Behavior | .833 | .114 | 11.158 | *** | par_1 |
| E3 | <--- | Ethical Behavior | .811 | .115 | 10.906 | *** | par_2 |
| E2 | <--- | Ethical Behavior | .711 |
.119 | 9.658 | *** | par_3 |
| E1 | <--- | Ethical Behavior | .699 | .124 | 9.504 | *** | par_4 |
| T1 | <--- | Trust | .778 | ||||
| T2 | <--- | Trust | .771 | .076 | 12.324 | *** | par_5 |
| T3 | <--- | Trust | .776 | .077 | 12.416 | *** | par_6 |
| T4 | <--- | Trust | .725 | .078 | 11.427 | *** | par_7 |
| T5 | <--- | Trust | .842 | .079 | 13.768 | *** | par_8 |
| T6 | <--- | Trust | .828 | .076 |
13.475 | *** | par_9 |
| T7 | <--- | Trust | .866 | .083 | 14.299 | *** | par_10 |
| T8 | <--- | Trust | .748 | .078 | 11.872 | *** | par_11 |
| C7 | <--- | commitment | .779 | ||||
| C6 | <--- | commitment | .788 | .084 | 12.444 | *** | par_12 |
| C5 | <--- | commitment | .712 | .088 | 11.006 | *** | par_13 |
| C4 | <--- | commitment | .691 | .084 | 10.623 |
*** | par_14 |
| C3 | <--- | commitment | .760 | .082 | 11.912 | *** | par_15 |
| C2 | <--- | commitment | .793 | .082 | 12.553 | *** | par_16 |
| C1 | <--- | commitment | .692 | .094 |
10.645 |
*** |
par_17 |
| L6 | <--- | Loyalty | .844 | ||||
| L5 | <--- | Loyalty | .861 | .063 | 16.391 | *** | par_18 |
| L4 | <--- | Loyalty | .906 | .059 | 17.944 | *** | par_19 |
| L3 | <--- | Loyalty | .899 | .058 | 17.721 | *** | par_20 |
| L2 | <--- | Loyalty | .770 |
.061 | 13.647 | *** | par_21 |
| L1 | <--- | Loyalty | .856 | .054 |
16.218 | *** | par_22 |
According to Bagozzi and Yi, and Bearden et al., the CFA result provide overall fit indices: Chi-square was highly significant ( p < 0.001), the goodness of fit index (GFI) was 0.760 and adjusted goodness of fit index (AGFI) was 0.713; incremental fit index (IFI) = 0.882; the normed-fit index (NFI) = 0.829 and the root mean square residual (RMR) = 0.097. All indices indicating that the measurement models provide good support for the factor structure determined through the exploratory factor analysis, which suggests the unidimensionality of the measures. According to Anderson and Gerbing, the validity of the structural model is assessed using three key validity dimensions: nomological, discriminant and convergent.
The nomological validity is the validity of the entire model and indicates whether the model fits the data; the discriminant validity refers to the extent of separation between the constructs; and the convergent validity verifies the homogeneity of the indicators and their constructs. However, convergent validity can be stated as all its variables are associated to significant and high loadings at least at be significant at 95%. Therefore, it can be stated that scales have convergent validity.Thus, evidence of convergent validity was provided by the fact that all measurement items loaded on the appropriate constructs. Discriminant validity means that each factor represents a separate dimension.
According to Anderson and Gerbing, discriminant validity is the degree of correlation among the constructs, and the correlation between different constructs. It refers to the principle that the indicators for different constructs should not be so highly correlated as to indicate they are measuring the same thing. As mentioned above, correlations were run between the items of several constructs used in this study (Table 3).
However, in all instances the item correlations satisfied the R= 0:85 cut-off often used by researchers for this assessment, thus suggesting the model demonstrated good discriminant validity. Furthermore, nomological validity refer to the ability of a scale to behave as expected with respect to some other constructs to which it is related. As mentioned above, ethical sales behavior can improve customer trust, customer commitment and customer loyalty. Therefore nomological validity would be demonstrated if the scale were positively and significantly correlated with customer trust, customer commitment and customer loyalty.
As stated in Table 3, all correlation coefficients
between the dimensions of ethical sales behavior, customer trust, customer commitment and customer loyalty are
positive and significant (at p < 0.01). Thus nomological validity of the scale is demonstrated. Considering the
reliability, nomological, convergent, and discriminant validity tests collectively, it's conclude that the
measurement model satisfies all the psychometric property requirements. Table 3 shows the means, standard
deviations, and intercorrelations for the constructs operationalized in this study.