Non-Parametric Econometric Analyses of Algerian Inflation: An Application of KRLS and QQKRLS Methods

Abstract

This article reexamines the determinants of inflation in Algeria between 2002 and 2025 by mobilizing two advanced non-parametric econometric methods: Kernel Regularized Least Squares (KRLS) and Quantile-on-Quantile KRLS regression (QQKRLS). Unlike standard linear models, these approaches make it possible to capture nonlinearity, distributional heterogeneity, and complex dependencies between oil prices, exchange rates, and inflation. Preliminary specification tests (Ramsey RESET, Breusch Pagan, Jarque Bera, BDS) all reject the hypothesis of linearity and independence, fully justifying the use of these techniques. KRLS estimations reveal a positive and significant average marginal effect of oil prices on inflation, but with marked nonlinearity: the impact weakens for very high oil prices. In contrast, the average effect of the exchange rate is not statistically significant. The main contribution lies in the QQKRLS analysis, which highlights strong heterogeneity of effects across the quantiles of inflation and explanatory variables. For low inflation, the effect of oil prices is paradoxically negative, whereas under moderate to high inflation regimes, oil shocks exert an amplified effect. The exchange rate shows a strong negative effect when inflation is low (disinflationary appreciation of the dinar), but a positive and destabilizing effect during periods of high inflation. These results suggest that “one-size-fits-all” monetary policies are inappropriate: the response to external shocks must be conditioned by the prevailing inflation regime. The study contributes to the literature on the economics of rentier states by providing robust evidence of nonlinear and asymmetric dependencies, and proposes concrete implications for stabilization policy in Algeria.