Predictability of Government Expenditures: A Machine Learning Application Kamu Harcamalarının Öngörülebilirliği: Bir Makine Öğrenmesi Uygulaması

AKDEMİR T., KAPTAN S.

Sosyoekonomi, vol.34, no.68, pp.273-292, 2026 (ESCI, Scopus, TRDizin) identifier

  • Publication Type: Article / Article
  • Volume: 34 Issue: 68
  • Publication Date: 2026
  • Doi Number: 10.17233/sosyoekonomi.2026.02.11
  • Journal Name: Sosyoekonomi
  • Journal Indexes: Emerging Sources Citation Index (ESCI), Scopus, TR DİZİN (ULAKBİM)
  • Page Numbers: pp.273-292
  • Keywords: ARIMA, Budgeting, LSTM, Prediction, Public Expenditures
  • Open Archive Collection: AVESIS Open Access Collection
  • Ankara Yıldırım Beyazıt University Affiliated: Yes

Abstract

Budget predictions are important for public financial management officials, as one of their primary responsibilities is to control expenditures and revenues in line with the government's macroeconomic goals. In this respect, we empirically analysed the predictability of Türkiye's public expenditures. The results from both the LSTM and ARIMA models showed that public expenditures can be largely predicted from their past values. The LSTM model demonstrated greater predictive capability for public expenditures than the ARIMA model. The contribution of macroeconomic and financial variables to predicting public expenditures was found to be infinitesimal.