Help Sheet for Project 2
The goal here is to show how to extract the appropriate data from the Bloomberg screen to create the data set you need. Remember that the code for this was FWCV (see screen shots below). We are using data from Treasury securities. Hint: read instructions through before starting.
This will be illustrated with the extraction of a single observation. For a sample to be statistically valid you should have 20-30 such observations. If you have less than 30 you should use the t-distribution rather than the standard normal distribution for your cutoff values. If you have forgotten how to do all this, the notes Topic 2: Statistics and Calculus under Classnotesis a good review.
Example using 6 month and 3 month securities
Date / Actual 6 mon yield / Actual 3 mon yield / 3 mon implied forward rate / Actual 3 mon yield 3 months hence / u from yield curve / w from naive13 Jun 2011 / entry / entry / entry / entry / entry / entry
13 Sep 2011 / entry
13 Dec 2011 / entry / entry / entry / entry / entry / entry
13 Mar 2012 / entry
13 Jun 2012 / entry / entry / entry / entry / entry / entry
13 Sep 2012 / entry
13 Dec 2012 / .000974 / .000523 / .001420 / .000927 / .000493 / -.000404
13 Mar 2013 / .000927
The data above was collected from the Bloomberg screens below. Notice that the actual 6 month rate on Dec 13, 2012 is circled in red and colored in red here (and so on). The actual 3 month rate on Dec 13, 2012 is circled in blue. From these two, a projection of the 3 month rate 3 months hence can be obtained as follows, using the pure expectations theory of the yield curve (which ignores the pesky term premium):
This can be solved to get the projected 3 month rate, i, that will occur in 3 months (that is, on March 13, 2013):
Bloomberg calculates this for you and on the Dec 13, 2012 screen it is circled in green, but you should understand the principle behind the calculation and include one sample calculation in your paper per instructions.
Now you have one observation. Collect 20-30 observations in the same way. You switch the dates on your screens by changing the appropriate orange box. Space out your collection so that the maturity (i.e., tenor) of your long security (6 month in this case) matches the frequency of your observations (6 months apart). For your visualization these are boldfaced dates in the example illustrated in the table above. If this is not a tedious process then you are probably doing the project incorrectly. As instructed, start with your birthday approximately 15 years ago.
Pure Expectations Yield Curve Model Forecast
Now compare the forward rate implied by the pure expectations model of the yield curve (circled in green) to the actual 3 month rate that really does occur onMar 13, 2013 where it is circled in yellow. Calculate u as
Random Walk Forecast
See if a naïve individual who simply believes that the actual 3 month rate on Dec 13, 2012 circled in blue is a good prediction of the actual 3 month rate on Mar 13, 2013 circled in yellow. Compute the random walk forecast error as
Perform the statistical tests for each forecast error series, u and w, described in the project sheet and see which approach wins.