I'm most interested understanding how assets perform in a portfolio context, and that's the focus of this post.jalbert wrote: ↑Wed Jun 06, 2018 2:00 pmThe problem with long duration bonds for investment portfolios is that the institutional investors who drive the market for them are insurance companies and pension funds and other players who need to cover long duration liabilities. Because they are matching duration to liabilities, they are not taking term risk, so they do not require to be compensated for term exposure. That is why the yield curve slope flattens so much as you move out on the term structure (even when the yield curve has average or above average slope by historical standards).
You are definitely not alone in preferring shorter duration bond funds in investor portfolios, but I think this stems from thinking about the bond fund in isolation rather than in the way it interacts with the portfolio as a whole. I think that's one advantage of the developments over the past ten years or so in looking at risk optimization and risk budgeting (developments which, I might add, seem to have been spurred by real investors finding out in 2008-09 that portfolio's they THOUGHT were diversified really weren't).
Let's focus on Treasuries for a minute, in part because we have a pretty good series of index-based ETFs that have a track record of more than 10 years.
- iShares Short Treasury Bond ETF (SHV)
- iShares 1-3 Year Treasury Bond ETF (SHY)
- iShares 3-7 Year Treasury Bond ETF (IEI)
- iShares 7-10 Year Treasury Bond ETF (IEF)
- iShares 10-20 Year Treasury Bond ETF (TLH)
- iShares 20+ Year Treasury Bond ETF (TLT)
The blue bars represent the ETFs alone, and you can see your claim has some support: the ETF with the highest risk-adjusted return is iShares 1-3 Year Treasury Bond ETF (SHY) at 0.90.
But when you look at the ETF as part of a portfolio instead of in isolation (red bars) you get a different story: the best risk-adjusted PORTFOLIO in this example is 60% VTSMX & 40% TLT. That combination DRAMATICALLY outperformed
Still, this post is really about risk management NOT about chasing better returns (risk-adjusted or otherwise). So look at how dramatically differently the duration impacts the ETF volatility versus a portfolio containing the ETF.
The portfolio containing iShares 10-20 Year Treasury Bond ETF (TLH) had LESS volatility than the portfolio containing iShares 1-3 Year Treasury Bond ETF (SHY). It was a much better diversifier in that it balanced the risk of VTSMX better. That the TLH portfolio had a CAGR that was 180 bps higher than the SHY portfolio was a happy bonus.
Which leads to the next point:
I think this is an important observation. Because most risk-optimization procedures do NOT, by design, include any information about expected returns it is necessary for the person constructing the portfolio to have some mechanism for ensuring that the final portfolio meets their needs in terms of both expected reward and risk. This can be done through asset selection (so that you are optimizing risk starting from a basket of assets that all have acceptable returns), weight constraints (limiting low return assets, like short duration bonds, to XX%% of the portfolio no matter what the optimization regime might prefer, etc.), factor exposure targets, etc. Diversification ratio algorithms will at least pare down the assets in the portfolio as part of the process: risk parity and inverse volatility approaches will include every asset you feed them. Attilio Meucci has done some interesting work on using PCA and other techniques to interrogate the effective number of bets a portfolio contains, though the math is little daunting.