Forecasting Lithuanian Inflation

The paper presents a short-term Lithuanian inflation forecasting model for predicting monthly inflation of 5 main HICP subgroups. We model inflation employing a set of univariate equations, which are mainly based on firms’ mark-up pricing. We make use of disaggregate HICP data, consisting of 92 price series, which naturally evokes discussion of potential pros and cons of forecasting disaggregate series vs. forecasting an aggregate. Besides exploring potential gains of using disaggregate data, we are also interested in the international commodity prices transmission mechanism, which we implement employing a distributed lag model. To examine the performance of model’s forecasts, we employ a recursive pseudo real-time out-of–sample forecasting exercise, generating inflation forecasts up to 15 months ahead. We find that our suggested set of univariate equations produce more accurate forecasts than the competing factor model, VARX model and various benchmark models for all 5 HICP subgroups.