Modelling satisfaction with ATMs: Data collection

The data set was collected from 380 consumers in a busy shopping area of a major UK city (ie Cardiff) in 1997, and 200 consumers in a major Hungarian city (ie Budapest) also in 1997 (however only 102 of the Hungarian sample used ATMs). A small team of researchers interviewed respondents over a four-day period. Table 1 shows the descriptive statistics for all the variables collected. As is fairly typical for marketing data on customer satisfaction, the frequency distribution on satisfaction levels is skewed towards the top end of the rating scale (fours and fives) with a fairly small tail in respect of the distribution for ones and twos.

EMPIRICAL RESULTS

The hypothesised model has been tested in two stages; first, differences in the explanatory and dependent variables were tested using a student t-test framework (see Table 1). Secondly, a regression model utilising a dummy variable for each country was estimated (see Table 2); the dummy variable took the value of zero for the UK sample and one for the Hungarian sample.

Table 1 Means and standard deviations


UK


Mean


SD


Hungary Mean
SD


student t-test

Overall satisfaction

3.85

0.87

3.50

0.64

3.83***

Recommend to others

3.97

0.93

3.89

0.93

0.71

Full use of services

2.62

1.14

2.48

1.18

1.12

Frequency of use

3.50

1.14

2.17

1.18

10.08***

Expectations

3.79

0.91

3.72

0.96

0.69

Confidence

3.81

1.04

3.32

0.95

4.54***

Charges

3.21

1.33

3.04

0.67

1.30

Perceived risk

2.41

1.09

2.35

0.96

0.50

Number of observations

380

102

Broadly, the two data sets have fairly similar means as shown by the student t statistics indicating no statistical difference; however, there are three striking exceptions. First, Hungarian customers appear to display a much lower overall level of satisfaction than UK customers.

Secondly, Hungarian customers on average use ATMs much less frequently than UK customers. Hungarian customers appear to use an ATM 2-4 times per month, whereas UK customers appear to use an ATM 5-10 times per month. Thirdly, Hungarian users appear to be less confident than UK users when using an ATM service. This is probably due to the different banking systems in each country.

Table 2 Regression results for the hypothesis model

Variables

Global coefficient

Dummy variable

Overall expectations

0.14***

-0.11

(3.51)

(-1.42)

Perceived risk

-0.07**

0.18**

(-2.34)

(2.54)

Confidence

0.09**

-0.06

(2.46)

(-0.70)

Recommendations to others

0.37***

-0.141

(8.60)

(-1.59)

Charges

0.06**

-0.07

(2.07)

(-0.65)

Frequency of use

0.09**

-0.02

(2.46)

(-0.35)

Full use of services

0.02

-0.01

(0.62)

(-0.09)

Constant

1.16***

0.74

(5.72)

(1.41)

n

482

Adjusted R2

0.404

F

24.32 +

Table 2 shows the results from applying the hypothesised model and has been estimated using ordinary least squares.

Overall, the model works fairly well with an adjusted R2 of over 40 per cent and a number of explanatory variables that appear to be important in explaining variation in overall satisfaction. Expectations, perceived risk, confidence, level of charges and the frequency of use are the prominent variables in the global model. The dummy variables for the Hungarian model display little difference from the global model with the exception of perceived risk. Therefore, it can be concluded that the hypothesised model can be applied to different countries, and although the importance of each variable may change in different countries this model appears to be fairly robust. Furthermore, the hypothesis H1 can be accepted; that the model appears to explain the variation in people’s overall satisfaction with ATM services across countries. Finally, the model was tested for multi- collinearity and hetroscedasticity problems using the appropriate statistical tests, which the estimated model passed easily.