Doing some yield curve forecasting and unsure whether should be working with yield or change of yield.
-
$\begingroup$ Typically it is considered stationary and mean reverting. Yield curve analysis can be done on either level or returns since, for example, correlations on levels are explainable. $\endgroup$ – Kch Apr 19 '18 at 17:42
-
$\begingroup$ There was actually a good answer on here saying that it is non-stationary because its variance is changing. If empirically it has been indeed observed that variance (or mean) of say Gaussian change through time -- then indeed it would be a non-stationary process. Having thought about it more, there are interest rate models. I think all of them assume nonstationarity (just like geometric brownian motion for equities). So price and yield both seem to be non-stationary processes, hence requiring looking at relative change (which at least resembles more of a stationary process). $\endgroup$ – A.L. Verminburger Apr 19 '18 at 21:48
Following Meucci (Risk and Asset Allocation book, page 112-113) you should use "change of yield to maturity" (simple change, not percentage) since they represent Fixed Income´s invariant. Change of yield to maturity would be the equivalent to change in price (in ln terms) for equities.
-
$\begingroup$ Don't see how relative change and percentage are any different, but do appreciate the reference. $\endgroup$ – A.L. Verminburger Apr 20 '18 at 15:20