Local Polynomial Estimator in The Nonparametric Model of Inflation in Indonesia

Inflation is a general and continuous increase in prices of goods and services over a certain period. Nonparametric regression analysis can be used to model inflation data that does not form a particular pattern. This study applies a local polynomial nonparametric method to model the rate of change rate in the inflation over a period considering two factors influencing inflation: the rate of change in the BI interest rate and the rate of change rate in the money supply from the previous period. The bivariate local polynomial method estimates the nonparametric regression function by considering the optimum Gaussian kernel bandwidth and polynomial order using the Taylor series expansion and WLS estimator. The optimal local polynomial nonparametric regression model was obtained based on a minimum GCV value of 0.015108 with two optimum Gaussian kernel bandwidth values of 0.1 and 0.03 in polynomial order of 1. The best model had a MAPE value of 3.45%, showing that all the prediction models were highly accurate. The benefits gained are additional information and consideration for determining monetary policy, especially inflation in Indonesia, by determining the BI interest rate and money supply.