I want to forecast the US yield curve for the next year, does any one have a clue from where to start, especially for the short term rates (i.e. O/N, LIBOR 3M, LOBOR 6M….)
i don’t think anyone can actually forecast these rates. i think people tend to just plug the forward curve in for forecasts with the knowledge that they are going to be wrong.
There are easy methods you could implement in excel, but anything more complicated you’d want to use something more powerful. Also, I would say that the forward curve is probably the best guess at what it will be. The short-term rates are actually quite a bit easier than the long-term rates. Let’s take the libor rate as an example. You can break this into two components, the OIS and the Libor-OIS spread. OIS can be modeled from the Federal Funds rate and any model or judgment that you have for where the Fed Funds will go over the next year. For Libor-OIS, there are a lot of fancy things you could do, like regime shifts or what have you. Basically, you’ll want to look at changes in the Libor-OIS spread, since it is probably I(1). One approach would be to assume that it is mean-reverting. In this case you would get like the weekly values of the spread, take the difference and regress those against the lagged value of the spread (not the difference). You could use the coefficients to back out the parameters for a Ornstein-Uhlenbeck. You could also incorporate some macro or other variables. Then you would just want to simulate out through your time horizon. Once you get these values for like the earliest part of the curve, you could also model the other parts of the curve as either a premium relative to the early part of the curve or alternately based on the expected value of the lowest part of the curve over the relevant time horizon. These techniques work reasonably well for the early parts of the curve, but as you go further out (like 2yr+) the whole thing is dependent on the premiums or really far out forecasts and the whole thing gets really uncertain. For instance, Macroeconomic Advisers 10 year treasury model is basically their forecast for the short-rate plus a term premium model that incorporates long-term vs. short-term inflation expectations along with some variables reflecting how monetary policy is expected to be (like the change in the unemployment rate and the change in the inflation gap, and the difference between the Fed Funds from the policy rule). This approach is decent if you’re looking far out and accept that the actual rate could very well be quite different than what the model says. Not very accurate in the short-run.