Analyzing inflation by using various econometric models and comparison of their forecast performances

Since inflation is a common problem of all economies around the world, modeling and forecasting inflation carry an important weight in escaping from a large scale of inflation costs. Anticipated inflation has very few or no effects on real economic variables. For this reason many economic models have been developed and also econometric models have been used in forecasting inflation. However as a result of literature reviews it has been seen that studies using both of these two methods are very few and the studies comparing models use limited numbers of models and variables.In this study, firstly it is aimed to reveal the model which gives the most successful inflation forecast of Turkey between 2003 and 2011 and many models in literature are tried in this framework. Besides examining the suitability of theoretical models in Turkey, time series models were also tried. As a result of the study, as well as comparing forecasting models, more models and variables are aimed to be contributed to the literature.Interest rate, exchange rate, money stock and output are defined as inflation dynamics and are used as varibles being in a mutual relationship with inflation. In this study in addition to these variables, inflation expectations and uncertainty variables which are thought to explain inflation process are also used. The data of expectation survey which was published by Central Bank of The Turkish Republic since 2003 are used to reflect expectations of economic agencies about inflation and inflation uncertainty was obtained by ARCH (Autoregressive Conditional Heteroscedasticity) model. Since the uncertainty variable which was obtained by inflation rate gave insignificant results, inflation uncertainty was obtained by using a defined inflation process.In the study, as time series, VAR (Vector Autoregressive Model), ARIMA (Autoregressive Integrated Moving Average ) and ARDL (Autoregressive Distributed Lag) Bound Test approaches were used. The models which also contained inflation expectations and uncertainty variables as inflation dynamics were estimated with both M1 and M2 money stocks definition and the ones that provided successful estimations were preferred. As a result of estimations, VAR and ARIMA models forecasted CPI (Consumer Price Index) annual exchange rate better than CPI monthly exchange rate.Moreover, CPI monthly exchange rate forecasts of ARDL model were more successful than CPI annual exchange rate forecasts. It is understood that after adding inflation expectation and uncertainty variables, more successful VAR and ARDL models are obtained. Another contribution of the study is presenting very close forecast in comparison with the actual rates via obtaining CPI monthly exchange rate by ARDL Bound Test which was not used before in inflation forecasting in Turkey. Thus these findings will light the way for further studies.To forecast inflation, Phillips Curve is used as a theoretical model. In Phillips Curve Analysis unemployment rates and GDP (Gross Domestic Product) values are used. CPI annual and/or monthly exchange rates are taken as the dependent variables in whole analysis. None of the forms which are; Traditional Phillips Curve, The Expectations-Augmented Phillips Curve and The Expectations-Augmented Phillips Curve including the NAIRU (Non-Accelerating Inflation Rate of Unemployment ) are found suitable in forecasting inflation of Turkey between the data interval of the study. However Expectations-Augmented Phillips Curve in which output gap is used, presented significant results for all used index methods. The model which is estimated by CPI quarterly annual data presents as well successful forecast in quarterly and annual inflation forecast.In this study Phillips Cruve Analysis is presented from Traditional Phillips Curve to Expectations-Augmented Phillips Curve in a wide framework. Like the previous studies that monthly data were used for the applied Phillips Curve, the models using monthly data has been insignificant in this study. Therefore the time series models using monthly data and economic models could not be compared. Only Phillips Curve model presented successful result is Expectations-Augmented Phillips Curve that uses output gap. The limitation of the study is that GDP determinant used in this model is a quarterly data.