A Relationship Model Of Loyalty, Delinquency And Profitability In Personal Loan Portfolio

Abstract

Loyalitas, Konsumtif, Risiko Kredit, Kemampuan Membayar, Bank, Loyalty, Consumer, Credit Risk, Capacity to Pay, Bank The second analysis was done by developing a Loyalty Index Score through 4 variables. The level of loyalty for each customer in the sample was developed through those variables. By looking at data distribution and normality, the scoring method was done using criteria based scaling. The third analysis was done by using regression. Regression was used to analyze the xssociation of Loyalty Index Score to Profitability. Regression is choosen because of its practicability. Regression in here was divided into 2 parts, to test the regression value between loyalty to profitability, level of delinquency to profitability and loyalty to level of delinquency. This analysis shows the direct and indirect effect of loyalty to profitability by way of path analysis approach. This approach was done without any segmentation or grouping, it is purely direct regression between Loyalty Index Score, level of delinquency and profitability. To have a better model, next regression was done between loyalty to revenue, loyalty to credit loss and loyalty to profitability. This regression was done after grouping the same Loyalty Index Score with its revenue, credit loss and profitability in that group. The group was sorted from the lowest one into the highest one where revenue, credit loss and profitability numbers was filled in according to its Loyalty Index Score group following the sequence. By doing this, it was expected that each score group will show definite revenue, credit loss and profitability numbers which can be used for portfolio development. Further analysis was done by doing clustering. Clustering gives much better Figure on what segment to be approach, what segment to be further developed and what segment to be left. To do this, the sample will be clustered or classified into 3 different loyalty categories: Class 1, Class 2 and Class 3. The cluster was developed by using k-means algorithm. The last analysis was done by using ANOVA to see the significance impact of loyalty on profitability based on credit risk segments or capacity to pay. Based on the clusters developed, accounts in the sample were classified based on its level of delinquency: non delinquent, early delinquent, late delinquent and restructuring segment. Profitability value on clusters within the same level of delinquency was compared to see the effect of loyalty on profitability in every level of delinquency. In summary, sample distribution were dominated by male customer, married, no dependence, has Bachelor degree, income below 5 million, age 30 – 39 years and owned the house. This is a typical of low to mid level of customer segment which can be someone who just started their job (but not a fresh graduate) and family, need the cash to strengthen their future plan either short or long term. Normally, this segment is not price sensitive but more into cash flow sensitive or in other words, very sensitive to the total instalment that they need to pay on a monthly basis. The development of loyalty index score was continued by analysing the accounts in the sample by its Period of membership (POM), Number of Product Ownership (NOPO), Loan Completion Rate (LCR) and Referral (RF). The information on those 4 variables was then scored by considering its central tendency value based on the overall sample distribution on each variable. The value of Mean, Median, Mode, Max and Min was used as main considerations in determining the scale and score to be given to the accounts in the sample. After all samples were scored, data normality was checked to ensure that the next analysis come from a strong data foundation. The data normality test was done and the result shows that the data is acceptable to be used for further analysis as can be proved from the data distribution value, mean (86.87), median (90), skewness (-0.20) and kurtosis (0.17). This information confirmed that the data is quite centred as can be seen from the slight different value of mean and median value whereas skewness and kurtosis value were very close to zero. Path analysis result shows that loyalty index score does influence level of delinquency and profitability. However, the sensitivity value of this model through path analysis will need to be further improved and therefore the next analysis through simple regression between loyalty index score and profitability was done one more time by grouping the sample into its loyalty index score and compare that with the percentage of credit loss, average revenue collected and profitability within the same group. This approach was able to produce much stronger result with R square value equal to 0.81 (credit loss), 0.87 (revenue collected) and 0.84 (profitability). This R square value was much better than the first analysis through path analysis approach with R square equal to 0.16 (loyalty and level of delinquency to profitability). ANOVA as the next analysis tools used in this research was further confirmed the effect of loyalty to profitability. The cluster development by K-means algorithm approach produce 3 different clusters with various profitability level in each level of delinquency. The difference of profitability in Non Delinquent sample was clearly significant especially between Cluster 1 & 2 (average profitability IDR 1.8 million) with Custer 3 (average profitability IDR 2.4 million). In Early Delinquent sample, the difference was also significant between Cluster 1 (average loss IDR 6.0 million) with Cluster 2 and 3(average loss IDR 1.3 million and IDR 0.9 million). Late Delinquent sample showed a significant different between Cluster 1 (average loss IDR 23.0 million) with Cluster 2 and 3 (average loss IDR 9.7 million and 10.1 million). Restructuring segment also showed significant profitability result among clusters (Cluster 1 equal to IDR 1.2 million, Cluster 2 equal to IDR 0.17 million and Cluster 3 equal to IDR 0.02 million). The main difference in the restructuring segment versus the other three group is on the association of loyalty index score with the profitability result. Restructuring segment was the only segment with no dependency from loyalty index score to profitability as can be seen from the low R square value at 0.043. Cross tabulation analysis was done to further see the effect of demography on loyalty cluster. Based on chi2 result, it was confirmed that out of 7 demography variables (Gender, Marital status, Age, Dependent, Income Group, Education and House Ownership), there were 4 variables proven to be significant. The 4 variables are Age, Dependent, Income Group and Education. Further analysis by using cramer’s v was done to see the dependency of loyalty cluster with those 4 demography variables and it was concluded that only Income Group is proven to be significant. Based on the analysis result, it was concluded that loyalty does have effect on profitability, which shows that loyalty index score is proven to be the key determinant for level of delinquency and profitability in consumer credit portfolio. This result suggest that to have a long term growth, banks should focus on loyalty, keep on offering the right service to the customer for long term relationship and in the end will produce stable growth for the portfolio.