# How accurate is YTM as a reference measure of non-holding-to-maturity return?

We need to offer an estimated return of a non-hold-to-maturity strategy. Essentially, we borrow money from the market and buy a bond. Instead of holding the bond to mauturity and locking in a return equal to YTM, we will sell the bond before maturity. How should we provide an estimated return/quote for such strategy?

So far, we are using YTM - borrowing cost, where YTM is known at the time we start the trade, and the borrowing cost, for simplicity, is a fixed annulized rate (for example, if we borrow 100 to purchase the bond, the cost is 5 given a 5% borrowing cost per year. Usually we swap to such fixed rate from 3-month LIBOR using IRS).

My concern is if we use above measure, naturally we will look for bonds with higher YTM (for now, let's assume the borrowing cost is not related to YTM, bond ratings, hence possible lower cost from repo) and filter out bonds with lower YTM. However, since we are not holding the asset to maturity, is it possible that we might filter out some eligbile bonds with low YTM that offer high coupon and whose price won't change much in the future? In short, does YTM - borrowing cost serve as a reasonable estimate of holding-period return/quote for our clients?

(p.s., we tend to avoid rocket science predictive model if it's not siginificantly better than an easy measure like YTM. A ballpark measure that is relatively simple to explain and get is the best)

• Loosely speaking, the "best" estimate of returns would be the historical returns for the strategy. – user42108 Mar 16 at 13:00
• You need to calculate what's called a Holding Period Return (HPR). Crucial assumptions include the reinvestment rate for the coupons that you receive over the holding period, cash flow assumptions (if they are non-deterministic) and the yield/spread assumption that you use at the end of the holding period. (You may also want to perturb the financing rate). General practice is to show a grid of projected returns for the different assumptions versus a single estimated return. Of course, the spacing of the grid itself is tied to the (expected) volatility of these assumptions. – Sharad Mar 16 at 21:22
• Appreciated, sounds like a simple but reasonable idea! Just to confirm, if I'm entering a 3-month trade say from March to May, then you're suggesting back calculating return for trade from Jan to March, from last year December to Febuary and so on and use perhaps the average, is that right?@user42108 – Nicholas Mar 17 at 2:40
• @Nicholas - my experience has been that investors put the greatest weight on actual results. If you've been running the strategy for some time, presenting the historical results would be preferable to a forecast or a backtest. Ideally you want results over a period that spans different interest rate and credit regimes. – user42108 Mar 17 at 13:03
• @Nicholas -- The grid of yield/spread assumptions for the end of the holding period is typically centered at the starting yield/spread but you could use a lower horizon yield if the bond has significant roll-down. The perturbations can correspond to simple shifts like +/- 10bps, 20bps etc or reflect historical +/- 1 sd, 2 sd etc moves over the horizon period. Even this simple minded-view can be helpful in revealing the risk-reward profile of different bond choices. – Sharad Mar 17 at 19:21