Econometrics analysis

Option 2: Forecasting UK Inflation

“Since Central Banks have abandoned the control of monetary aggregates and have implemented inflation targeting rules directly or indirectly, by means of aggressive Taylor rules, forecasting inflation rates has become crucial for both policy makers and private agents who try understand and react to Central Banks behaviour. Several methods have been proposed to estimate and forecast the dynamics of inflation rates but the overall performance has been, at best, mixed: the information contained in the dynamics of past inflation appear suffice and very few other variables add marginal predictive content to univariate specifications.” European Central Bank, working paper 151, Fabio Canova, 2002.

“Long-term nominal commitments such as labor contracts, mortgages and other debt, and price stickiness are widespread features of modern economies. In such a world, forecasting how the general price level will evolve over the life of a commitment is an essential part of private sector decision-making.” Jon Faust and Jonathan Wright, ‘Forecasting inflation’, Handbook of Economic Forecasting, 2013.

In this coursework, you will be fitting a univariate model (ARMA) to UK CPI inflation and use this model for a forecasting experiment. The out-of-sample forecasting performance of the model is to be compared with a benchmark model.

You will be using a monthly time series dataset which spans the period 1994m1 to 2021m8. The data is taken from the Office for National Statistics (CPI INDEX 00: ALL ITEMS 2015=100), and seasonally adjusted after download. The inflation variable, i_t, in the file “UK inflation monthly.dat”, is an annualised inflation rate in percent. It had been computed according to: i_t=1200*ln⁡(p_t/p_(t-1) ).

The assignment instructions are as follows:

  • Comment on descriptive statistics and on graphics
  • Fit an ARMA model to the data for the sample 1994m1 – 2016m8
  • Test the residuals of the model for serial correlation
  • Perform pseudo out-of-sample forecasting experiments for the period 2016m9 – 2021m8
  • Perform dynamic forecasts with STATA (The one-step forecast is used to compute the two-step forecast, and so on.).
  • Compute the mean square prediction errors (MSPE) and compare the model performance with an average-of-last-12-months model.
  • Perform a one-step prediction experiment, by forecasting one step ahead at each point in time.
  • Compute the mean square prediction errors and compare the model performance with a one-step average-of-last-12-months model.