Prediction on the volume ofnon-performingloans in Turkey using multivariate adaptive regression splines approach


Kılıç Depren S., Kartal M. T.

INTERNATIONAL JOURNAL OF FINANCE & ECONOMICS, 2020 (Peer-Reviewed Journal) identifier identifier

  • Publication Type: Article / Article
  • Volume:
  • Publication Date: 2020
  • Doi Number: 10.1002/ijfe.2126
  • Journal Name: INTERNATIONAL JOURNAL OF FINANCE & ECONOMICS
  • Journal Indexes: Social Sciences Citation Index, Scopus, ABI/INFORM, Aerospace Database, Business Source Elite, Business Source Premier, Communication Abstracts, EconLit, Geobase, Metadex, Civil Engineering Abstracts
  • Keywords: banking sector, macroeconomic determinants, multivariate adaptive regression splines, non-performing loans, Turkey, BANK-SPECIFIC DETERMINANTS, LOANS EVIDENCE, CREDIT RISK, GREECE, MODEL

Abstract

As the volume of non-performing loans (NPL) plays an important role in the capacity of providing loans, increasing NPL problem is perceived as one of the major challenges in a bank-based financial system. In recent years, Turkey has been confronted with this problem. In order to reduce NPL volume, necessary precautions that can contribute to government administrators should be taken. Therefore, the most important predictors on NPL volume were determined by using predictive analytics such as multivariate adaptive regression splines that is found out the optimal parameter effects from high-dimensional data. The dataset, which is obtained from the Central Bank of the Republic of Turkey and Banking and Regulation Supervision Agency, consists of 14 independent variables from 2005 and 2019. As a result of this study, it is indicated that credits, USD/TL exchange rate, and the unemployment rate are the most statistically significant factors in defining the level of NPL volume.