Het assortiment van de winkel heeft veel variatie
M = 2,60 M = 3,88 + 1,28
De winkel biedt een ruime keus aan producten
M = 2,78 M = 3,90 + 1,12
De kleding die de winkel verkoopt, wordt aangeboden in veel kleuren
M = 2,36 M = 3,80 + 1,44
De kleding die de winkel verkoopt wordt aan geboden in veel maten
M = 2,25 M = 3,64 + 1,39
De winkel heeft een groot assortiment
M = 2,04 M = 4,07 + 2,03
TABLE 5. Manipulation check
5.3 Descriptive results
5.3.1 Influence of a large assortment on store choice likelihood
To measure the effects of a large assortment and a small assortment on consumers’ store choice likelihood, two separate linear regressions has been performed. A linear
regression is an approach to modeling the relationship between the dependent variable, store choice likelihood and one or more explanatory variables. In this research the two independent variables are a small and a large assortment and are called the explanatory variables. The case of one explanatory variable in one test is called simple regression.
Taking a look at table 6, it becomes clear that a large assortment is a significant predictor of store choice likelihood. The F value tests the overall significance of the regression model. The first regression model (table 6) shows a good estimation of store choice likelihood, since the F value is 29.29 and the P is .00 (p < .01). The F value should normally range from zero to a large number. The adR2 (.23) indicates the percentage variation in store choice likelihood by a large assortment: 23% of the variance of store choice likelihood is explained by a large
assortment. The regression coefficient (B) of the variable large assortment indicates an increase of the effect on store choice likelihood (.51). The β should not be taken into consideration since this model concerns a single regression and not a multiple regression.
Model R R square Adjusted R square Std. Error of the estimate
.49 .24 .23** .83
Model Sum of Squares df Mean Square F Sig. Regression Residual Total 20.50 62.99 83.50 1 90 91 20.50 .70 29.29** .00 Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta (Constant) Large assortment 1.41 .51 .38 .09 .49 3.71 5.41** .00 .00 ** p < .01 ; *p < .05 ; ° p < .1
TABLE 6. Linear regression analysis large assortment on consumers’ store choice likelihood H&M
5.3.2 Influence of a small assortment on store choice likelihood
Taking a look at table 7 it becomes clear that a small assortment is no significant predictor of store choice likelihood. This second regression model shows no good estimation of store choice likelihood because the F value is below 1(.05) and the P is .81 (> .1).
A negative value of adR2 did occur (-.01) because the fit between a small assortment and store choice likelihood is actually worse than just a horizontal line. In this case, R-square cannot be interpreted as the square of a correlation. The regression coefficient (B) of the variable small assortment indicates a negative effect on store choice likelihood (-.02). The β should again not be taken into consideration since this model concerns a single regression.
These marks show that a large assortment will have a strong and positive effect on
consumers’ store choice likelihood while a small assortment will have a negative effect on consumers’ store choice likelihood.
Model R R square Adjusted R square Std. Error of the estimate
.02 .00 -.01 .85
Model Sum of Squares df Mean Square F Sig. Regression Residual Total .04 65.77 65.81 1 90 91 .04 .73 .05 .81 Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta (Constant) Small assortment 1.77 -.02 .24 .09 -.02 7.39 -.23 .00 .81 ** p < .01 ; *p < .05 ; ° p < .1
TABLE 7. Linear regression analysis small assortment on consumers’ store choice likelihood Mango
To find out if there are more striking findings, the small and large assortments will be tested in one single test with both store choice likelihood variables. For this analysis a MANOVA test will be conducted. MANOVA stands for multivariate analysis of variance and is a statistical test procedure for comparing means of more dependent variables; in this case two. The regression analysis used before, can only measure one dependent variable at one time. The MANOVA helps to answer: 1. do changes in the independent variable(s) have significant effects on the dependent variables; 2. what are the interactions among the dependent variables and 3. what are the interactions among the independent variables. Essentially, MANOVA takes scores from the multiple dependent variables and creates a single dependent variable giving the ability to test for the above effects. Statistical reports however will provide individual p-values for each dependent variable, indicating whether differences and interactions are statistically significant. In comparing with the regression analyses done before, the MANOVA analysis should result in more findings among the different variables.
The SPSS outcome shows three different tables. The second table shows the actual
the Wilks Lambda needs to be regarded. The Wilks Lambda is the measurement of the
difference between groups of means on the independent variables; the smaller the lambda, the greater the differences. In table 8 the test results are depicted. It resulted in a Wilks Lambda of .00 and F = 4.49 for large assortment and a Wilks lambda of .11 and F = 1.59 for small
assortment. The two assortment variables differ on the two Y’s (store choice likelihood H&M and store choice likelihood Mango). According to the third outcome table, the large
assortment shows to have a significant effect on both store choice likelihood with P HM is .04 (< .05) and P Mango is .00 (< .05). The small assortment has only a significant effect on one of the dependent variables: P Mango is .02 (< .05). This last finding shows a different result
as regards the outcome of the regression analysis in paragraph 5.3.2 where the small
assortment shows no significant effect on store choice likelihood at all. After all, this slightly different outcome does not support the first hypothesis, since the large assortment has still a more positive effect on store choice likelihood than the small assortment does. There is not enough evidence for hypothesis 1, thus the first hypothesis must be rejected.
Large and Small assortment on two store choice likelihood
Sig Wilkson Lambba Value
F
Large assortment Small assortment
Between subjects effects
Small assortment Y choice large Y choice small Large assortment Y choice large Y choice small .00** .11 .72 .02* .04* .00** .02 .11 4.49** 1.59 .73 2.98* 2.00* 9.73** ** p < .01 ; *p < .05 ; ° p < .1
TABLE 8. MANOVA analysis large and small assortment on both consumers’ store choice likelihood
5.3.3 Mediation effect of loyalty on the relationship between a small assortment and store choice likelihood
To measure the effect of the mediator store loyalty on the relationship between a small assortment and consumers’ store choice likelihood, more than one regression analysis is needed. The causal steps methods developed by Baron and Kenny (1986) is the most
three conditions necessary for mediation, and the updated version by Kenny (1998) describes four steps to infer mediation.
If loyalty (Z) mediates, there can be said that Z is able to fully or partial explain the relation between small assortment (X) and store choice likelihood (Y). As Baron and Kenny
suggested, partial mediation is a more realistic expectation than complete mediation. There are four steps needed to demonstrate a mediator effect. The first step is a regression analysis with Y as dependent variable and X as predictor. This analysis needs to demonstrate that X has a significant effect on Y. Table 8 step 1 shows that there is no relationship between a small assortment and store choice likelihood.
The second step is a regression analysis with loyalty (Z) as dependent variable and small assortment (X) as predictor. This analysis needs to demonstrate that X (small assortment) has a significant effect on Z (loyalty). This is actually true (see step 2).
The mediator variable Z can only explain a relationship between X and Y if Z is linked with Y as well. Punctually said; a third regression analysis must be performed with Y as dependent variable and Z as predictor. This analysis needs to demonstrate that Z has a significant effect on Y. Step 3 shows that loyalty has no significant effect on store choice likelihood.
If loyalty can explain the relation between a small assortment and store choice likelihood, this would imply that store choice likelihood is no longer in relation with a small assortment as soon as loyalty (Z) will be held constant. To check this, a last regression analysis will be performed with Y as dependent variable and both X and Z as predictors. This analysis needs to demonstrate that the partial regression coefficient of X, keeping Z constant is not
significant. Step 4 shows that store choice likelihood is not in relationship with small assortment as soon as loyalty is held constant, which is good. But since there is no
relationship between a small assortment and store choice likelihood at all and the fact that loyalty has no significant effect on store choice likelihood, the mediator role of loyalty cannot be measured properly. The mediator variable loyalty is not able to fully or even partial explain the relation between a small assortment (X) and store choice likelihood (Y). With this
information there is not enough evidence to support H2.
Sig Adjusted R square Unstandardized coefficient (B)
STEP 1 X Y .81 -.01 -.02
STEP 3 Z Y .95 -.01 .00 STEP 4 X and Z Y Small assortment (X) Loyaly (Z) .77 .85 -.02 -.03 .02
TABLE 9. Step 1-4 small assortment (** p < .01 ; *p < .05 ; ° p < .1)
5.3.4 Mediation effect of loyalty on the relationship between a large assortment and store choice likelihood
For the final hypothesis the same four steps had been followed, to test the mediator role of store loyalty on the relationship between a large assortment and store choice likelihood. The third hypothesis predicts that there is no mediated effect of loyalty in this relationship.
Step 1 shows that there is a relationship between a large assortment and store choice likelihood. The second step measures if X (large assortment) has a significant effect on Z (loyalty). Step 2 shows that a large assortment has indeed a significant effect on loyalty (p = .00). The third step measures if loyalty also has a significant effect on store choice likelihood just has a large assortment has a significant effect on loyalty. Table 9 step 3 shows that loyalty has a significant effect on store choice likelihood. The last step of measuring the value of the mediator variable loyalty is to check if loyalty can explain the relation between large
assortment and store choice likelihood. This would be the case if store choice likelihood is no longer in relation with large assortment as soon as loyalty (Z) will be held constant. This analysis needs to demonstrate that the partial regression coefficient of X, keeping Z constant is not significant.
Table 9 step 4 demonstrates that the regression coefficient of X (large assortment), keeping Z (loyalty) constant is not significant since p=.13 (p > .01) which is the rule of thumb in step 4. Since there is a relationship between a large assortment and store choice likelihood, the fact that a large assortment has a significant effect on loyalty, the fact that loyalty has a significant effect on store choice likelihood and the fact that steps 4 shows no significant effect there can be said that loyalty has a mediator effect on the relation between a large assortment and store choice likelihood. Since there is a mediating effect found, there must be discriminated between full and partial mediation of loyalty (Z). If the effect of X on Y is totally explained by Z there is full mediation of Z. The effect of X on Y disappears when Z is held constant. Partial mediation means that the effect of X on Y gets weaker but does not disappear completely. Tables 9 shows that loyalty has a partial mediation since the effect of a large assortment on store choice likelihood gets weaker but does not disappear (step 1 = .51 and
step 4 =.19). According to step 4, 33% of the total effect on store choice likelihood is
mediated by loyalty. This outcome is in contradiction with the last hypothesis, predicting the opposite effect that there is no mediator effect at all. With these conclusions, H3 is not supported.
Sig Adjusted R square Unstandardized coefficient (B) STEP 1 X Y .00** .23** .51** STEP 2 X Z .00** .49** .73** STEP 3 Z Y .00** .32** .58** STEP 4 X and Z Y Large assortment (X) Loyaly (Z) .13 .00** .33 .19 .44
TABLE 10. Step 1-4 large assortment (** p < .01 ; *p < .05 ; ° p < .1)
5.3.5 Mediation effect of loyalty on the relationship between assortment variety and store choice likelihood
Two stepwise regression analyses have been performed to test the mediation effect of loyalty on the relationship between a small assortment and store choice likelihood and the relationship between a large assortment and store choice likelihood. To find out if there are more striking findings, the small and large assortments will be tested in one single test. To combine both small assortment and large assortment a different type of test will be used. The MANOVA will be used because, this time, there are two dependent variables (store choice likelihood H&M and store choice likelihood Mango), two independent variables (large assortment and small assortment) and two mediators (store loyalty H&M and store loyalty Mango). Since the mediator role will be tested again, the same four steps as before, will be taken. The second table shows the actual MANOVA outcomes. For the most exact values the Wilks Lambda needs to be regarded.
The first step shows that the two different assortment sizes, small and large assortment, did differ on the two different Y’s (store choice likelihood). Only for the large assortment, the MANOVA resulted in a significant result (Wilks Lambda = .76, F = 13.7, P = .00). The small assortment resulted in no significant results. (Wilks Lambda = .96, F = 1.67, P = .19). From the third table, the between subjects effects we can see that a large assortment has a
statistically significant effect on only store choice likelihood with a large assortment (see table 11).
The second step shows that both a large and a small assortment have a significant effect on store loyalty. The between subjects effects show that a large assortment has a significant effect on both loyalty large assortment and loyalty small assortment (see table 11).
STEP 1 X1, X2 Y1, Y2 Sig Wilkson Lambba Value
F
Large assortment Small assortment
Between subjects effects
Small assortment Y choice large Y choice small Large assortment Y choice large Y choice small .00** .19 .08 .74 .00** .37 .76 .96 13.7** 1.67 3.1 .10 27.6** .79
STEP 2 X1, X2 Z1, Z2 Sig Wilkson Lambda Value
F
Large assortment Small assortment
Between subjects effects
Small assortment Z loyaltylarge Z loyaltysmall Large assortment Z loyaltylarge Z loyaltysmall .00** .00** .00** .32 .01** .00** .50 .73 44.16** 16.47** 31.35** 1.01 6.05** 86**
TABLE 11. Steps 1 and 2 (** p < .01 ; *p < .05 ; ° p < .1)
The third step is performed with the two Y’s as dependent variables and the two Z’s as predictors. The outcome shows that loyalty large assortment has a significant effect on store choice likelihood while loyalty of small assortment shows no significant effect at all (see table 11). From the between subjects effects there can be concluded that there is only a significant effect of loyalty large assortment on store choice likelihood of the store with a large
assortment. The last step will be performed with the two Y’s as dependent variables and all X’s and Z’s as predictors. Table 11 step 4 shows that only loyalty of a large assortment has a significant mediator effect on the relationship between a large assortment and store choice likelihood large assortment. These are the same results has steps 1, 2 and 3 demonstrated.
STEP 3 Z1, Z2 Y1, Y2 Sig Wilkson Lambba Value
F
Loyalty large Loyalty small
Between subjects effects Loyalty small Y choice large Y choice small Loyalty large Y choice large Y choice small .00** .95 .88 .79 .00** .09