SCHOOL OF ECONOMICS, FINANCE and MARKETING ECON1195 - FINANCIAL ECONOMETRICS DUE DATE: 25th September 2016 TUTORIAL 3 CONTRIBUTION: 5 % ___________________________________________________________________ QUESTION 1 Dynamic Models This question again investigates the relationship between government revenue (REV) and expenditure (EXP) for Australia. The data is on the sheet Economic in the file called Tutorial Data Sem 2 2016.xlsx, and this consists of monthly data in millions of Australian dollars. You will need information from Modules 5 and 6 to answer this question. If you decide to use logs in that question, continue to do so now. (a) Carry out a Granger causality test to find the relationship between government expenditure and revenue. Comment on what this tells you about government policy. Does the Australian Government seem to decide how much to spend based on what it receives, or does it seem to decide how much to spend, then endeavour to raise revenue accordingly. (b) I will assume that the government does not just look at the revenue received this month when making decisions, but uses some average of past months as well. This suggests a distributed lag model would be suitable. Find the optimal lag length k for the Distributed Lag Model of EXP. For this and all models in this question, use the data from June 2006 until December 2015 for estimating the model. EXPt = 0 + i REVt-i + ut using the Akaike Information Criterion. Check whether you obtain the same value of k when you use the Hanna-Quinn and Schwarz criteria. Do not use more than 12 lags. (c) Comment briefly on the results obtained in part (b) using the Statistical output and the graph obtained by clicking View Actual, Fitted, Residual You should also check to see whether collinearity is a problem in the model. (d) Using the optimum lag length found in part (b), decide whether an Almon PDL model is appropriate here. If it is not suitable, explain why. If it is suitable fit the model. (e) Using the results from part (b) and (d) if available, to obtain the Static Forecasts of EXP from January 2016 until the end of the data set and compare these to the actual values. Comment briefly on how useful these forecasts are to anyone who wishes to forecast EXP.   QUESTION 2 Box-Jenkins Models Use the data for revenue on the sheet Economic in the file called Tutorial Data Sem 2 2016.xlsx for this question. You will be able to do this question after Modules 7 and 8. Make sure the range on your work file for this question ends in June 2016, otherwise you will not be able to do the forecast in part (e). (a) Plot the line graphs of revenue and log(revenue). Decide which of these you will use for the rest of this question giving brief reasons for your answer. (b) Does the series contain a unit root, using a level of significance of  = 0.05. (c) Using data from June 2006 up until December 2015 obtain a suitable Box-Jenkins model for forecasting revenue. If you are using MA terms in the model, you may have to change your estimation period to start later. (d) Obtain static forecasts over the remaining period and compare these with the actual values. Do you think these forecasts would be useful if you were a public servant helping the treasurer prepare the budget? (e) Re-estimate your model using the same variables until the end of the data period. Have the coefficients changed much? What does this tell you about the model? Using this model produce dynamic forecasts from January until June 2016.