Monday, November 2, 2015

CD 5-Year Report Card: Part 3

In the first two posts in this series, I compared the 5-year return of a 5-year direct CD to a 5-year Treasury security, and provided a fairly detailed explanation of how to evaluate risk and return of fixed-income investments, such as CDs, bonds (including Treasuries), and bond funds. In this post, I conclude the series by presenting the 5-year returns of various Vanguard bond funds, and by comparing these results to the CD and Treasury in terms of risk as well as return.

The comparisons in this post are for the 5-year period ending September 30, 2015 for the bond funds, and September 15, 2015 for the 5-year CD and 5-year Treasury note. The bond fund average annual returns were conveniently obtained directly from the Vanguard web site during October 2015, when the returns ending September 30, 2015 were displayed (as I write this, the returns shown are for the period ending October 31, 2015). Although the holding period for the bond funds doesn't exactly match that for the CD and Treasury, I think that they're close enough for a reasonable comparison.

The bond fund returns are for the Admiral shares class, with a minimum investment of $50,000 for all funds compared here, except the Total Bond Market Index fund, which has a minimum investment of $10,000.

The comparison table shown later in the post has quite a few columns, but if you read Part 2 of this series, you should be able to understand the table contents with a little additional explanation. The table lists the securities (CD and Treasury) and funds, the average annual returns, several risk-related measures, and the difference between the return of each security or fund and that of the CD.

From Part 2 of this series, you'll recognize the terms Credit risk and Term risk, the headers of two of the columns in the comparison table.

Note that on the Vanguard web site, if you use the left navigation bar to select Bond as the asset class, you can filter the bond funds by credit risk (Credit quality) and term risk (Maturity).

The credit risk value shown in my comparison table for each fund is a numerical approximation of the average credit quality rating provided by Morningstar (M*), but with some adjustments applied. Other than the adjustments, explained below, I use a scheme based on the credit rating tiers shown in a Wikipedia article on bond credit rating to translate the letter-based credit ratings to numeric values, as follows:

M* average credit qualityNumeric rating
AAA0
AA1
A2
B5

So the larger the number, the higher the credit risk. Note that there's a larger numerical gap between A and B than between AAA and AA or between AA and A. This is because a credit quality rating of B actually is quite a bit lower than a rating of A, which you can see in the Wikipedia article on bond credit rating.

M* assigns an average credit quality rating of AA to the Vanguard Treasury funds, I assume because one of the credit rating agencies (S&P) downgraded the US from AAA (outstanding) to AA+ (excellent) in 2011. My sense is that most investors still consider US debt obligations to be essentially risk free, so I assigned a value of 0 to all federally-backed securities (Treasury note and CD) and funds that include only Treasuries (all Treasury funds).

M* also assigns an average credit quality rating of AA to Vanguard Total Bond Market Index fund, but this fund holds about 35% of its portfolio in corporate bonds, so it has more credit risk than a fund that holds only Treasuries. So I rated this a 1 for credit risk.

M* assigns an average credit quality rating of A to all of the Vanguard investment-grade bond funds, so these are all rated a 2 using my scheme.

Finally, M* gives an average credit quality rating of B to Vanguard High-Yield Corporate Bond fund, so the numerical rating is 5.

I'll show the table of results now, and will explain the other risk columns below.


Security or fundCredit riskTerm risk (duration)Overall riskReturn / RiskAvg annual returnReturn vs. CD
5-year direct CD00.50.56.082.74%
5-year Treasury Note02.02.00.731.46%-1.28%
High-Yield Corp. fund (VWEAX)54.69.60.666.32%3.58%
Int-Term Invest-Grade fund (VFIDX)25.57.50.574.25%1.51%
Int-Term Treasury fund (VFIUX)05.15.10.542.73%-0.01%
Short-Term Invest-Gr fund (VFSUX)22.64.60.472.18%-0.56%
Total Bond Mkt Index fund (VBTLX)15.76.70.442.98%0.24%
Long-Term Invest-Gr fund (VWETX)213.015.00.446.54%3.80%
Short-Term Treasury fund (VFIRX)02.22.20.420.92%-1.82%
Long-Term Treasury fund (VUSUX)016.116.10.386.11%3.37%

The securities and funds are sorted by decreasing values in the Return / Risk column, which I'll explain after explaining the other risk columns to the left of it.

Update: an astute Boglehead pointed out that I mistakenly assigned a credit-risk rating of 2 to the Long-Term Treasury fund in the comparison table, which also resulted in a slightly higher value for overall risk, and a slightly lower value for Return / Risk. I have modified the credit-risk rating to the correct value of 0, and updated the other values accordingly. It does not change the rank order of funds in the table though.

To quantify term risk for the funds, I used the duration values from Vanguard's web site. Duration is an approximate measure of the percentage decrease (increase) in fund value (or share price) for each percentage point increase (decrease) in the yields of all bonds in the fund. For example, if the yields of all bonds in the Total Bond Market Index fund increased by one percentage point, the value of the fund would fall by about 5.7%.

The duration of the 5-year Treasury note is calculated with a spreadsheet duration formula, using a term to maturity of 2.5 years, since that's the average maturity over the 5-year life of the bond. The duration is as high as 4.8 at time of purchase, but declines to 1.0 after four years, and approaches 0 as the bond nears maturity. So the bond is much riskier early in the holding period, but the average duration of 2.0 over the entire 5-year holding period is a better value for comparing with the funds (for which duration is roughly constant over time).

Coming up with a duration value for the direct CD is kind of a finger-in-the-wind exercise, because the liquidation value of the CD does not change as interest rates change--it always equals the value of the principal and earned interest minus the early withdrawal penalty (EWP) until the day of maturity, at which time the EWP no longer applies. The number shown is approximately the value of the EWP in percent, which would be a reasonable number for comparison with the funds assuming a one percentage point change in yields. For smaller changes in yield, the CD would lose more than indicated by the duration value shown, and for larger changes it would lose less--potentially much less.

This particular CD had an EWP of about two months of interest, so the percentage loss would be about 2 / 12 * 2.74% = 0.46%. This is rounded and expressed in percent for an estimated duration of 0.5.

To derive an overall risk measure, I simply add the credit risk and term risk numeric values. This is a very rough risk measure, and I'm sure that someone else could justify something different. As always, I welcome comments on this and everything else I post.

Note that the very high term risk for the long-term bond funds causes these funds to end up with the highest overall risk ratings--even higher than the High-Yield Corporate Bond fund, which some might think would be higher risk. However, another common risk measure is the standard deviation (SD) of returns, and over the five year period being evaluated, the SDs of the long-term Treasury (12.03%), long-term investment-grade (8.35%), and high-yield corporate (5.35%) funds do correlate fairly well with my overall risk numbers. I use Portfolio Visualizer to determine the SD values; Portfolio Visualizer is a very nice, free, web-based tool with which you can do many kinds of historical investment returns analysis.

The Return / Risk values are calculated as average annual return divided by overall risk. This is very similar in concept to Sharpe ratio, a measure that is widely used to evaluate the risk-adjusted return of investments and portfolios.

I think it's very important to think in terms of risk-adjusted return. Although some of the bond funds had higher average annual returns than the CD, these higher returns were obtained only by taking considerably more risk. 

Sometimes risk pays off, and sometimes it "shows up". It so happens that over the 5-year period examined here, both term risk and credit risk generally were rewarded within the bond fund universe. This can be seen by examining the relative returns of the funds for a given level of credit risk, or for a given level of term risk, or simply by observing that higher overall risk generally resulted in higher average annual return.

To see a recent example of when bond risk showed up, take a look at the bond fund returns for 2013, when interest rates increased enough to make a significant difference in returns. For example, the total return for the fund with the highest overall risk of the funds evaluated here, the Long-Term Treasury fund, was -12.94% (that's minus 12.94%) in 2013.

One of the most striking results in the comparisons shown in the table above is that the risk-adjusted return (return / risk) of the direct CD is an order of magnitude larger than for the Treasury note or any of the bond funds. This is why I favor direct CDs so much, and why they now comprise about 75% of my fixed income.

On the other hand, since taking more risk sometimes pays off, I've hedged my bets by keeping some of my fixed income in bond funds, including some of those shown in the comparison table. I now have about 25% of my fixed income in bond funds.

An interesting exercise is to compute the metrics I've used here for a combination of 75% in the CD and 25% in the Intermediate-Term Investment-Grade Bond fund. This is a fund I'd recommend, especially in a tax-advantaged account (IRA or 401k/403b), and preferably in conjunction with direct CDs or other safe, high risk-adjusted return fixed-income investments (like the TSP G fund, or a good stable value fund). Computing the metrics for this combination is a simple arithmetic calculation, but since this post already is so long, I'll leave this to the interested reader, or perhaps will show these results in another post.

10 comments:

  1. Thanks Julieta! Always appreciate feedback.

    ReplyDelete
  2. Your chart is quite invaluable. Once suggestion; you can give different weights to (credit risk) vs. term risk. This would probably weigh even more heavily in the favor of CD's.
    The adage about more concern about return of my money than on my money.
    Still; this is a great chart to start with.
    Many thanks.

    Rob5TCP

    ReplyDelete
    Replies
    1. Thanks for the comment Rob.

      Are you suggesting increasing the values for credit risk, so for example, B rated bond funds might have a credit risk weighting in the range of 10-15 instead of the value of 5 that I assigned?

      Assuming this is the case, then of course, there are an infinite variety of numeric assignments that could be used. Assigning numeric values to credit risk and term risk, then adding them to get a value for overall risk already is a pretty rough approach, so I'm not sure that increasing the values for credit risk would necessarily be better.

      Also, note the observation that the long-term Treasury fund, with no credit risk but large term risk, has a higher standard deviation of annual returns than does the high-yield corporate bond fund, which has significant credit risk but much less term risk. So the rough scheme I've presented aligns fairly well with the common measure of risk, standard deviation of returns.

      With respect to return of money vs. return on money, we should remember that what matters is return of money in real terms, not nominal terms. Long-term nominal bonds, like those in the long-term Treasury fund, have significant risk of returning much less in real terms than in nominal terms, due to the risk of unexpected inflation. This basically is incorporated into the term risk.

      Kevin

      Delete
  3. Kevin, this series is excellent. I have been struggling with rationalizing the bond funds in my portfolio. I had previously lowered my asset allocation by 5% in bonds due to concerns of a bond bubble. Your analysis has led me to move a portion of my total bond fund into direct CDs in an IRA from a CU. And I'll be going back to my original asset allocation. Just wanted to thank you, much appreciated.

    Jim

    ReplyDelete
    Replies
    1. Jim, glad you found it useful!

      Kevin

      Delete
    2. Thanks Kevin. I learned a lot from your posts!

      Delete
  4. It's been a long time since I took a statistics class, but it appears that you are doing arithmetic operations on ordinal data.

    ReplyDelete
  5. Given it’s now June 2018, I guess I’m a little late to see this, but find it really interesting!
    One question if you’re still taking them. The argument for bond funds is typically that they inversely correlate to stock returns.
    So while on a “stand-alone” basis, they lose the horse race, is this still true if “paired” with stocks in say a balanced portfolio?

    ReplyDelete
    Replies
    1. Good question, Marybeth. Sorry for the late reply. I'm just now going through my blog comments (mostly to remove the spam that has accumulated).

      It's true that Treasuries (bills/notes/bonds) often, but not always, have inverse correlation to stocks, but the long-term correlation is about 0 (about like cash). We saw good negative correlation in late 2008 when stocks tanked and Treasuries increased in price (yields decreased); this was a classic flight to safety scenario.

      Also important to point out that it was only nominal Treasuries for which this was true. Corporate, investment-grade, municipal, and other types of non-Treasury bonds, even TIPS, declined in value in late 2008--of course for the most part not nearly as much as stocks.

      To really take advantage of this, and not just have it show up as lower portfolio volatility on paper, you had to sell Treasuries and rebalance into stocks during a fairly short timeframe of a few months, as Treasuries fell in value in early 2009.

      A high enough yield premium on CDs can more than make up for the rebalancing bonus you might get by rebalancing from Treasuries into stocks when stocks tank. Even Larry Swedroe, who is one of the biggest proponents of looking at the whole portfolio, has generally preferred CDs to Treasuries when the yield premiums were large enough.

      Finally, it doesn't have to be all or none. You aren't going to need all of your fixed income for rebalancing into stocks when they tank. You could take advantage of the yield premiums of CDs for a portion of your fixed income, but also hold some Treasuries for the potential rebalancing bonus in a flight to safety scenario.

      Incidentally, the 5-year report card would favor CDs even more now than it did when I wrote this blog post, with the increasing yields and declining prices of bonds and bond funds since about mid-2017. The 5-year annualized return on the Vanguard Intermediate-Term Treasury fund is about 1%, while the CDs I have maturing in the near future have returned about 3%.

      On the other hand, the levels of CD yield premiums we were seeing five years ago generally are not available now, although there may be an exception now and then.

      I generally have been using proceeds of maturing CDs in taxable accounts to buy Treasuries from 6-month to 2-year maturities, as they actually provide higher taxable-equivalent yields (TEYs) than CDs at my tax rates, due to the state income tax exemption for Treasuries. The yield curve flattens out too much after two years for me to feel adequately compensated for taking the extra term risk. Someone who wanted a simpler, fund-based solution might use the Vanguard short-term Treasury index fund, which holds Treasuries in the 1-year to 3-year range, with average maturity of about two years.

      Treasury yields are higher than CDs even without the state tax exemption at maturities of less than one year, and they are about tied at 1-year maturity, so I would buy Treasuries out to 1-year even in an IRA, where the tax-exemption is irrelevant. CDs start to take the yield at 2-year maturity, and widen the gap as you extend further out. I would buy brokered CDs in an IRA from 2-year to about 3-year maturity at this point. Again, the CD yield curve is not steep enough beyond that for me to be comfortable taking the extra term risk.

      Note that with this approach I own both Treasuries and CDs, so I will have some potential rebalancing benefit if stocks tank and Treasuries increase in price, while still taking advantage of the CD yield premiums in my IRAs (although much smaller than in past years).

      Kevin

      Delete