Trends and Forecasting of Total Credit to the Economy in Albania: A Time Series Modelling Approach

Abstract

Bank lending channels funds from surplus to deficit economic units and supports economic activity, making aggregate credit a relevant indicator for financial monitoring and medium-term planning. This study models and forecasts the monthly evolution of total credit to the economy in Albania using 192 observations covering January 2010 to December 2025. Following the Box–Jenkins framework, non-seasonal ARIMA and seasonal SARIMA specifications were evaluated using stationarity diagnostics, information criteria, forecast-error measures, residual tests, and chronological out-of-sample validation. The logarithmically transformed series required two non-seasonal differences (d = 2) and no seasonal differencing (D = 0), indicating strong trend persistence but no stable annual seasonal pattern. During the training-stage analysis, ARIMA(5,2,0) was retained as a parsimonious baseline because its residuals were consistent with white noise; a subsequent full-sample log-scale grid search provided an ex post robustness check and ranked ARIMA(5,2,0) first by AICc. A 2024–2025 holdout evaluation of the training-sample ARIMA(5,2,0) order on the original credit scale produced a MAPE of 4.11%, while an expanding-window robustness exercise across five 12-month forecast origins yielded an average MAPE of 1.61%. Formal structural-break tests also identified instability around the COVID-19 period, which qualifies the interpretation of long-horizon projections. After validation, ARIMA(5,2,0) was re-estimated on the full logarithmically transformed sample to generate ex-ante forecasts for 2026–2028. Back-transformed forecasts imply year-on-year credit growth of approximately 13.8% in 2026, 14.6% in 2027, and 14.6% in 2028. The results provide a statistically supported reference path for aggregate credit monitoring, but the projections remain conditional on the continuation of historical dynamics and should not be interpreted as causal forecasts. The analysis therefore contributes an empirical forecasting benchmark that can support, but not replace, broader macroeconomic and financial-stability assessment.