Empirical Finance Spring II 2020
Forecasting Project: Due 4/30/2020
The objective is to predict the May 8th values (closing price) of the SP 500 index and
CBOE VIX using information which was available at the end of April (April 30th, 2020).
• Download daily data of the SP 500 index and CBOE VIX. According to the FRED
(Federal Reserve Bank of St. Louis) website, data are available from March 29th, 2010
and January 2nd, 1990 for the SP 500 index and VIX, respectively. You can also
include other variables for prediction.
• For example, suppose that you were to use information available at the end of March
and predict the April 3rd (Close*) of the SP 500.
You need to provide your choice of
– the prediction model: specify the regression equation
– the data: predictor variables, sample period.
If you are not relying on any model, you need to explain how your forecasts are derived.
You may take log of variables and run regressions. But, the final forecasts should be
reported in the original unit.
For ease of illustration, let Pt = 2584.59 indicate the closing price on Mar 31, 2020
and PT indicate the closing price on April 3, 2020. According to the Efficient Market
Hypothesis, stock prices evolve according to a random-walk process. So you may report
Et[lnPT ] = lnPt as your prediction about the April 3rd value. Make sure you report
exp(Et[lnPT ]) which is in the original unit.
• We learned in class that it is extremely difficult to predict short-horizon price move-
ments. There is no right answer for this project because of its nature. It is totally
possible that just a random guess could turn out to be a better forecast than other
model-implied forecasts. That said, the grading policy would be based on
PT−Et[Pt]
PT
Points
± 1% 10
± 2% 8
± 3% 6
± 4% 4
± 5% 2
For example, if the difference between your forecast Et[PT ] and the actual value PT
is less than 1% of the actual value PT , then you will get 10 points. If it is greater
than 5%, then you will receive 0 point. So far, the compensation will be based on
the final forecasts only. The maximum points are 20 for forecasting the SP 500 and
VIX index. The other 10 points can be rewarded based on the justification of your
approach. You may try with the historical data and provide some evidence as one of
justifications.