Thursday, November 7, 2013

Stocks as a Long-Term Hedge for Inflation

This morning I've been playing around with the Dimson-Staunton-Marsh Global Returns dataset for 20 developed market countries (1900-2012) to try and look more at how stocks and bonds have performed historically with respect to inflation over long periods of time. This discussion is a follow-up to my earlier discussion of William Bernstein's e-book Deep Risk.

Actually, I've tried a lot of different things, but I'm not coming up with any really good ways to present the data. What is clear is that both stocks and bonds are risky, even over long periods of time. However, when there is high inflation, stocks do look to be a bit less risky. Essentially, bond investors can be wiped out by inflation.

What follows are two tables I've made which combine the financial market data for the 20 countries.

First for 10-year intervals and then for 20-year intervals, I separate the data by the compounded inflation rate for all the rolling periods of that length in the dataset. Then I provide information about the cumulative inflation-adjusted gains (or losses) for stocks and for bonds categorized by the different inflation levels. 

What do you think?  Is there a case for saying that (at list in historical data) stocks are less risky than bonds over 10- and/or 20-year periods?


15 comments:

  1. Sounds like you're referring to long bonds. I think long bonds are always too risky for their return. I like TIPs for inflation protection. Stocks compensate for inflation, but not necessarily when the inflation is occurring (like the seventies). As Bernstein says, short bonds are an inflation hedge. I'm sticking with short-intermediate TIPs for inflation protection. Even if they have a small real loss in the short term, that's better than a large real loss.

    ReplyDelete
    Replies
    1. Dirk, you are right about the long bonds. The other choice in the dataset is bills. No short-term bonds and no inflation-adjusted bonds.

      Delete
    2. Dirk,
      I ran the tables for bills instead of bonds. They get hammered even harder than long-term bonds.
      Of course you are right about TIPS and inflation protection. Actually, just last Friday I purchased my first I-bonds. They have an attractive feature that after a year they can be sold without any capital loss no matter what happens with interest rates. Seems like a great place to protect assets and have them available to buy stocks on the cheap next time there is a financial crisis.

      Delete
  2. It would be useful to see how a globally diversified portfolio would have performed with the database as that is currently the investment norm. Averaging over the 20 countries would work but a cap weighted average would be preferable, even if only approximate. You also might want to integrate short term bond yields over 10 and 20 years as Dirk suggests.

    ReplyDelete
    Replies
    1. I could calculate a GDP-weighted world index without too much trouble, at least for post-1950. Or, I might see if I can also get this data series when the new numbers including 2013 become available. Thanks.

      Delete
    2. I forgot that I had Angus Madisson's dataset with GDP data for all of these countries going back to well before 1900. I'm working on this now.

      Delete
  3. Hi Wade—

    You essentially have correlational results and it’s important not to imply causality between inflation and asset returns with this kind of analysis; there are many other potential factors at work. It would be perhaps more useful to employ multiple predictors of the cumulative gains – including but not limited to inflation rate, GDP growth rate, etc. – in a multiple regression analysis and then partial out the variance that is uniquely attributable to inflation and examine its magnitude and statistical significance.

    It would also help to use continuous rather than interval inflation data, especially since about 2/3’s of all of the rolling periods analyzed fall in the inflation range of 0-6% annually (i.e., only three intervals), with about 1/3 of all the data concentrated into only one interval. In addition, your results seem to reflect the returns of long bond FUNDS; it’s very hard for total returns of these to do decently over 10 or 20-year periods with markedly increasing interest rates, even with dividends reinvested at the increased rates. I’m reasonably confident that the bond data would look much more favorable if they reflected the performance of a (rolling) ladder of long bonds, all held to maturity, with maturing bonds reinvested in higher yielding issues during rising rate environments.

    ReplyDelete
    Replies
    1. Thanks for the comments. I'm not trying to imply any causation. I'm only trying to look at the question of which asset class tends to hold up better against the ravages of inflation over longer periods of time.

      And yes, continuous data would be nice. I was also trying to plot this data in a meaningful way, but the plots are a mess. There are a massive clump of points in the low inflation area, and some extreme outliers. It's hard to plot this data in a way that shows something meaningful.

      About building a rolling bond ladder, I don't think that is possible with international data. This dataset only has total returns information. No yields.

      Delete
  4. Hi Wade - So enjoy your work, thank you for sharing it via your blog.

    Could you add a supplement to this post, presenting this data on a compound annual growth basis (rather than cumulative?).

    For individual investors who tend to conceptualize returns in terms of cagr/annualized results (and for me personally :) thought that could be a very powerful way to think about the relationship on a practical basis (ie in ABC interest rate environment, the 10-year or 20-year CAGR on stocks or bonds has been XYZ). Therefore if I, Jane or Joe investor, expect inflation to go up/down, I may want to make these {fill in the blank} adjustments to my asset allocation.

    Thanks for considering <-- and if my logic is off and it's nonsensical to think about these figured on an annualized basis would love to learn that too!

    ReplyDelete
    Replies
    1. Thanks,

      I could have gone either way with the table, but I thought that the cumulative returns (or losses) would be most meaningful. Especially, when you compare 10 years to 20 years, 20 years could have a smaller average compounded loss but a bigger cumulative loss. I will keep both measures in mind when moving forward with this.

      Delete
  5. What a pity there aren't, say, bonds whose principal and interest are automatically adjusted for inflation. Real return bonds, or inflation-protected securities, or index-linked gilts, so to speak. I know what you're going to say: "Dreamer! Get your head out of the clouds!" Well, OK, but a fellow can wish, can't he?

    ReplyDelete
    Replies
    1. I guess you are being sarcastic, and certainly one of the great advances in the recent past is that we do now have inflation-adjusted bonds. This certainly changes the dynamics of the problem. Just last Friday I made my first personal purchase of I-Bonds and I've owned VIPSX and VAIPX for a long time.

      Delete
  6. I've just read several of you articles and the one about the true impact of immediate annuties triggered some thought since I like your work on SPIAs. (1) Since fixed SPIAs are less costly than inflation adjusted ones, it seems I could allocate more to stocks using fixed SPIAs and hedge the inflation risk that way. (2) In general, I could buy SPIAs over time and geared to certain expenses (e.g., joint SPIAs covering joint expenses such as housing costs and individual SPIAs covering individual costs such as food and medicine). (3) In line with your thoughts about having higher stock allocations, if I consider social security with SPIAs, I may justify a larger stock allocations than 50/50 split.

    ReplyDelete
    Replies
    1. Thanks.

      About your points:

      (1) Yes, I agree.
      (2) Interesting. This is something I've never seen discussed before. Sounds reasonable.
      (3) Yes, I agree. Increased safe income allows for a more aggressive allocation for remaining financial assets.

      Delete
    2. Thanks for the response. Adding to item #2, some things become apparent when you look at actually implementing an SPIA. States only insure a certain amount, typically around $300k to $500k. So you'll probably want to buy a few SPIA's anyway. You can disperse them among different insurers, buy them over time which increases payout, and design them to fit certain spending patterns. I've even looked at buying an SPIA to cover a remainder of a mortgage, although I think that may not be a good use unless you have a low mortgage rate, and SPIAs better serve you as longevity insurance. And there's a decent chance that you can have a greater terminal value by putting less proportion in SPIAs than you would a bond ladder (each covering the same income need) and so having more in stocks with a chance to earn considerable more than bonds over several years. And as I understand it, you can buy an SPIA within an IRA or 401k and avoid tax consequences, although I don't think you can buy an SPIA for a spouse within your IRA/401k.

      Delete