work with the Survey of Professional Forecasters (SPF) from the Philadelphia Federal Reserve. Background information about the survey can be found at this link:
https://www.philadelphiafed.org/research-and-data/real-time-center/survey-of-professional-forecasters
Download the individual responses of the Survey of Professional Forecasters (HOUSING) from the following link
https://www.philadelphiafed.org/surveys-and-data/housing
and download the realtime hstarts data from the following link
https://www.philadelphiafed.org/surveys-and-data/real-time-data-research/hstarts
(Each group should use different data period. I will inform you about this with the group formation)
Task 1 [5 marks]
Convert the actual housing data into quarterly data of the right type and merge it with the forecasts. Graphically examine the forecasts together with the actual housing data and explain your analysis and findings. (Max 300 words of analysis and explanation)
Task 2 [5 marks]
Evaluate the forecast accuracy over time graphically and explain your analysis. You can define the forecast error by:
𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡𝐴𝑐𝑡𝑢𝑎𝑙 𝑟𝑒𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 −1 where zero means perfectly accurate forecast.
a. Is there a difference in forecast standard deviation among forecasters from different industries? Compute the average standard deviation of the error by industry. (Max 100 words)
b. Compute the average across the forecasts per quarter. Then calculate the quarterly changes and regress the mean forecast change on the change in realized number of housing starts. Explain what this analysis shows. Focus on the coefficient of the forecast variable. (Max 200 words)
Task 3 [5 marks]
Can we use the %-changes in the housing start forecast to predict %-changes of the S&P500? Use a regression to test whether the forecasted change in housing starts for a quarter has a statistically significant relationship to the return of the S&P500 for that same quarter. Motivation of your analysis is as important as the code. (Max 100 words)
Task 4 [5 marks]
Perform individual level analysis on a forecaster basis. Are some forecasters better than others? Use your own analysis to show whether this is the case or not. Use graphs as well as statistical tests to motivate the results. (Max 100 words)
