CHAPTER 4: DATA PRESENTATION AND ANALYSIS
4.8 Quantitative data Analysis
4.8.2 Multinomial Logistics Regression
Multinomial logistics regression was utilised in the study for testing the causal relationship between the research variables. Multinomial logistics regression is ideal for analysing phenomena which have more than two possible outcomes. The multinomial logistic regression method is applied or utilised for determining the predictability of the dependent variable using the independent variables. According to Bayaga (201 0), the probabilities of different possible outcomes can be predicted by using the multinomial logistic regression. Petrucci (2009) noted that the method is known in different names such as the softmax regression, multiclass LR, polytomous LR, maximum entropy etc. in simple terms the method can be defined as the extension of the binary logistic regression. The major difference from the binary regression is that this method allows more than two categories of the outcome variable.
According to Hosmer et al. (201 3), the multinomial logistic regression is one of the attractive methods as it does not infer the linearity, normality or homoscedasticity. It can be recognised that the absence of the assumptions such as the independence among the dependent variable
strengthens the capability of the multinomial logistic regression. In addition, the method uses the diagnostic statics which can be interpreted easily. Pal (2012) has pointed out that the method avoids the assumptions of the covariance matrices and equal variances across the groups. Furthermore, for the analysis using the multinomial logistic regression, the independent variable does not need to be an interval. According to Huttunen et al. (2013) under the multinomial logistic regression, the distributed error terms will be avoided. Hence, it does not occupy a position under the assumptions. As the data collected in the current study qualifies as multiclass (with more than two possible outcomes), multinomial logistics regression is deemed ideal. Multinomial logistics regression, in this case, was used for identifying the predictors of customer satisfaction in London Overground.
Passenger attitude towards the effectiveness of service of London Overground as a predictor of passenger satisfaction
Multinomial logistics regression was used for assessing whether passenger attitude towards effectiveness of service of London Overground could be used as a predictor of passenger satisfaction. The following results were generated from the multinomial logistics regression.
► The passengers who found the services of London Overground to be effective are more likely to very satisfied with the service than passengers who found the service ineffective.
► The passengers who found the services of London Overground to be highly effective, effective and moderate are less likely to be dissatisfied with the service of London Overground when compared to passengers to who found the service ineffective. A clear causal relationship between passenger attitude towards the effectiveness of service of London Overground and passenger satisfaction in London Overground is evident from the results of the multinomial logistics regression. This result indicates that passenger attitude towards the effectiveness of London Overground is a predictor of passenger satisfaction. The implication of this result is that improvement in effectiveness of service of London Overground is critical for improving the customer satisfaction in the service.
Association between issues faced London Overground and overall customer satisfaction Based on the information collected from passengers regarding the key issues faced in London Overground, the effect of these issues on customer satisfaction was determined with the help of multinomial logistics regression (Refer to Appendix 2, table 7). The results of the multinomial logistics regression revealed the following results;
► Passengers who found high crime rate to be a major issue in London Overground were less likely to be satisfied with the service than passengers who found the late arrival of trains to be a major issue
► Passengers who found profit-oriented strategies of London Overground to be a major issue were less likely to be satisfied with the service than passengers who found the late arrival of trains to be a major issue
► Passengers who found customer complaints to be a maJor issue m London Overground were less likely to be satisfied with the service than passengers who found the late arrival of trains to be a major issue
The implication of the above result is that though the late arrival of trains in London Overground was identified as a major issue by passengers, this issue does not have as much effect on passenger satisfaction as other issues such as high crime rate, profit-oriented strategies of London Overground and customer complaints. This means that London Overground needs to prioritise reducing crime rates and resolving customer complaints about the late arrival of trains as they are more influential on customer satisfaction.
Passenger period of use of service of London Overground as a predictor of passenger satisfaction
Multinomial logistics regression was used for estimating how passenger period of use of services of London Overground influenced passenger satisfaction. The results of multinomial logistics regression revealed that passenger period of use of had a significant effect on passenger satisfaction. The following findings were inferred from the results of the multinomial regression analysis.
► Passengers who have been using London Overground for less than 3 years are less likely to be dissatisfied with London Overground than a passenger who has used the service for over 10 years.
► Passengers who have been using London Overground for 3-6 years are less likely to be dissatisfied with London Overground than a passenger who has used the service for over 10 years.
► Passengers who have been using London Overground for 6-10 years are less likely to be dissatisfied with London Overground than a passenger who has used the service for over 10 years.
The above results clearly indicate that passenger satisfaction levels are comparatively lower among customers who have been using the service for a longer time period have lower satisfaction with the services offered by London Overground. The implication of this result is that it sheds light on the ineffectiveness of the existing service of London Overground in satisfying customers who have been using the service for the longest period of time. This can also explain the very low retention rate in London Overground as the survey only 15% of passenger belonged to 10+ year group.
Passenger attitude towards rail privatisation as a predictor of passenger satisfaction This test focused on evaluating how passenger attitude towards rail privatisation influenced their satisfaction with London Overground. Some of the main results emerging from this analysis are:
► The results of the test revealed that passengers who found rail privatisation to be ineffective were more likely to be dissatisfied with London Overground than passengers who had a positive outlook towards rail privatisation.
► Passengers who were neutral to rail privatisation were also likely to be less satisfied with London Overground than passengers who had a positive outlook towards rail privatisation.
This result shows that customer attitude towards rail privatisation has clear impact on customer satisfaction. This point to the fact that customer satisfaction is an outcome of customer perception of service and that passengers who have a negative perception of rail privatisation are therefore likely to be less satisfied with the service as they are persuaded by their negative outlook. On the other hand, individuals who have a positive outlook towards rail privatisation are influenced by their positive perception when they articulate their satisfaction towards London Overground.