DETERMINANTS OF LOAN RECOVERY IN LENDING INSTITUTIONS, LAGOS METROPOLIS, NIGERIA

Abstract

Credit is a significant input in the making of any investment, the availability and accessibility of which measure the performance of economy across countries of the world. However, the continued increase in loan default and bad debt with low corresponding recovery of same has reached a worrisome dimension globally. The current study accesses the determinants of loan recovery in Nigerian lending institutions. The target populations for the study comprise of all Commercial Banks (CB) and Primary Mortgage Institutions (PMIs) practising in Lagos Metropolis. Three thousand, one hundred and ninety - seventy loan accounts in the categories of fully and partially recovered loans were randomly drawn from the databases of Fourteen(14) and Forty (40) Commercial Banks (CB) and Primary Mortgage Institutions (PMIs) respectively from 2000 - 2012. Collected data were analysed using t - test statistics, Analysis of Variance (ANOVA), Logistic Regression Model (LRM) and Artificial Neural Networks. The result of the t- test reveals that growth in Gross Domestic Product, year of borrower relationship with lenders, age of real estate collateral assets are significant characteristics of the fully recovered loans while inflation growth, increase in interest rate, loan size, and loan - to - value are significantly characteristics of partially recovered loans. The result of the Logistic Regression Model (LRM) also reveals that Gross Domestic Product, borrowers' history of default, value of collateral and location of real estate collateral assets have positive significant effects on loan recovery and that a unit increase in them raises the probability of loan recovery by 55%, 69%, 10% and 75% respectively. The result further reveals that growth in inflation rate, change in interest rate and loan duration have negative effects on loan recovery while loan size, loan-to -value indicate positive but non- significant effects on loan recovery. Moreover, the results of Analysis of Variance (ANOVA) reveals; that the total amount of loan recovered from real estate sector is significantly higher than all other sectors in the economy, there is significant variation in the recovery performance of loan recovery strategies employed in lending institutions and that the amount of loan recovered when a combined collateral assets are taken as security for loan is significantly higher than when a single collateral asset is used as a secondary tool of loan recovery. Furthermore, the overall loan classification efficiency of Logistic Regression Model (LRM) and Artificial Neural Networks of ascertaining the creditworthiness of loan applicants is adequate and falls within the recommended acceptable standard of 65% and above, however, the predictive performance of ANN is significantly better than that of LRM in terms of early detection of non-credible loan applicants incorrectly classified as credible loan applicants (Type l error) and credible applicants incorrectly classified as non credible applicants (Type 2 error). The study recommends that adequate analysis should be given to every conceivable factor that may affect loan recovery and the use of more objective loan evaluation techniques should be given priority in credit transactions.