Thursday, June 27, 2013

Asset Valuations and Safe Portfolio Withdrawal Rates

In the current issue of the Journal of Financial Planning, I co-authored an article called, "The 4 Percent Rule Is Not Safe in a Low-Yield World " with Michael Finke of Texas Tech University and David Blanchett of Morningstar.

The three of us have now followed up with a new article which we have just posted online called, "Asset Valuations and Safe Portfolio Withdrawal Rates."
 

This new article is on the same general theme about how current market conditions can impact sustainable withdrawal rates, but it uses a different methodology and also incorporates stock market valuation levels.

The issue is that bond yields today are well below and stock market valuations are well above their historical average. Importantly, there are no historical periods in the United States where comparable low bond yields and high equity valuations have occurred simultaneously. It is only in the last few years that we've simultaneously seen 10-year government bond yields fall below 2.5% while the cyclically-adjusted price earnings ratio is also above 20.

Both current bond yields and stock values have been shown to have predictive power for near-term returns. As well, portfolio returns in the first decade of retirement have an outsized impact on the final outcomes for retirement income strategies, which diminishes the importance of any notions that the markets may normalize toward their historical averages at a later date. Retirements today are occurring in a historically unprecedented environment in the United States and concepts like the 4% rule may not be reliable.

Traditional Monte Carlo simulation approaches generally do not incorporate valuations for either stocks or bonds into their analysis. This is because they assume that stock and bond returns are independent over time, fluctuating around some unvarying average return. In this article, we move away from those assumptions so that market returns are connected to what happens in the past. We simulate evolving bond yields and stock market valuation levels (specifically the cyclically adjusted price-earnings ratio sometimes called CAPE, PE10, or Shiller's PE), and then link those to bond and stock returns.

Our simulations indicate that the safety of a given withdrawal strategy is significantly affected by the initial bond yield and CAPE value at retirement, and that the relative impact varies based
on the portfolio equity allocation. Using valuation measures current as of April 15, 2013, which is a bond yield of 2.0% and a CAPE of 22, we find the probability of success for a 40% equity allocation with a 4% initial withdrawal rate over a 30 year period is approximately 48%. The sustainable withdrawal rate for a 30-year retirement, 40% stock allocation, and a 10% accepted failure rate is 2.8%.

To close out this post, here is a 12-minute interview on these topics conducted with Michael Finke and myself in late April for the American College's Wealth Channel:




32 comments:

  1. That's a shocker, Wade. 10% failure with withdrawal down at 2.8%. Should reinvigorate that long last-two-days SWR debate on Twitter.

    Can you summarize briefly how you relate years-ahead stock arithmean to starting PE10?

    Isn't it true that linking future stock returns to current PE10 amounts to assuming reversion toward mean? If so, are you confident of that assumption for the future?

    Is SWR @ 10%fail sensitive to allocation? If so, what allocation is best?

    Dick Purcell

    ReplyDelete
  2. Dear Dr.Pfau:
    I have found your studies on the 4% swr very interesting. I have come to the conclusion (although I may change my mind in 10 minutes) that I will withdraw anywhere from 1 to 4% without an inflation adjustment at the beginning of each year. I don't forsee any problems of this lasting 35 years (especially where I don't expect to last more than 10 to 15 years). Would you say that without inflation adjustments that even in this recent, unique low rate environment that a variable swr of 1 to 4% is viable for 30 to 35 years?

    ReplyDelete
    Replies
    1. Hi,

      It sounds like you are talking about withdrawing a percentage of remaining account balance each year, and that percentage will vary between 1 and 4% (depending on what?). It's not something I've specifically tested, but it sounds reasonable. Using a constant percentage strategy provides some flexibility since it allows spending to decrease after a market drop, so there is less sequence of returns risk. Exploring this aspect more is something I will try to in the future so that I can be more specific with my explanation.

      Delete
    2. Hi Wade, what you discuss in your response here is precisely what Frank, Mitchell and Blanchett (FMB) methodology does. Actually the percentage of withdrawal starts just under 4% (depends on age) and can go up from there. Depending on what? In FMB papers it depends on the age(s) and time remaining from Period Life Tables. Therefore, less time remaining (older retiree) has a higher withdrawal rate compared to more time remaining (younger retiree). This work has been done ... as you've said to me, it is on your reading pile since you've been busy moving lately ... and you'll get familiar with it on the new project we've been discussing offline with Dr Mitchell. SWR method does not measure either end of the distribution time periods ... i.e., age it starts, and age it ends. Result: paradoxes Kitces talks about.

      A Separate point: this is great insight from your paper. However, making a dramatic adjustment all of a sudden up front has a huge effect on retiree spending capability (or inversely how much pre-retirees need to save). It seems that making these adjustments over time allows the retiree to incorporate the effect of lower rates over time on their portfolio values - i.e., when they actually happen.

      At the end of the day, it is the CURRENT portfolio value that matters for current spending for the year. Projections and expectations are simply predictions that may or may occur. But, a review and adjustment each year keeps the spending rooted in reality as it is then known.

      Philosophically, at no time in the past could I accurately predict my future income for the rest of my life. I could have some expectations about the near future (next one or two years), but beyond that predictions always get fuzzy. Thus, having an expectation of predicting retirement income for the rest of one's life is not realistic either. Research should look at measuring prudent spending, so better decisions may be made such that a retiree doesn't inadvertently spend more than they should in any given year.

      Delete
    3. Larry,
      Thank you. You are helping to clarify the case for the acceptance of the idea that spending can't be carefully planned for far into the future. Thanks for the clarification about your article as well. There are certainly strategies that allow for an increasing withdrawal rate as the time horizon shrinks (a less sophisticated version of what is in your paper is to simply use the withdrawal rates implied by the IRS RMD rules). But this comment about switching around between 1% and 4% didn't seem to be based on increasing with age. We'll have to wait for further clarification. Thanks.

      Delete
    4. Hi again,
      Clarification? I am winging it. My wife is still working and will probably retire in two years. I just started collecting social security (my doctor gave me around ten years to go; I prefer fifteen). We will have my social security and my wife's pension as a base which should cover all necessary costs. The 1 to 4% will be as needed; that is, when we need a car or new roof, we may have to spend more. We will have a two-comma pot to raid. I would like to build up our savings to pass on to our kids and grandkids, but it is not necessary. By the way, we have had a very rewarding life, living in several countries. Arrigato.Sayonara.

      Delete
    5. Yes Wade, those RMD withdrawals do work at first. However, what we (Frank, Mitchell, Blanchett or FMB)found that happens in both our work, and also would apply to the IRS RMD approach, is the "exponential growth" of the withdrawal rate, and thus "exponential decay" of the portfolio value. An adjustment needs to be made to the withdrawal rate based on shorter distribution periods to mute the exponential nature (JFP Paper by FMB Dec 2012).

      The adjustment turns out to be quite simple: where n = number of years remaining in distribution period, 1/n = adjustment to withdrawal rate (WR% * (1-1/n)).

      If using the RMD tables, in order to preserve assets into very old ages this can also applied by first calculating the IRS required withdrawal and then applying 1/n adjustment to determine how much the retiree should NOT spend, but keep. In other words, the RMD is what has to be withdrawn for taxation. But it doesn't all have to be spent! If this is not done, the retiree may outlive their remaining balance.

      At young ages the difference between the two approaches is small because the 1/n adjustment is also small. However, at older retirement ages, the difference grows due to the exponential nature.

      Delete
  3. Hi

    I have a question on your paper. It is probably just due to the way your graphs are drawn but the data in Table 2 doesn't look like it is consistent with Figure 8. For example, in Figure 8, at a 0% tax rate, the 40% equity portfolio (with bond yield of 2% and CAPE of 22) looks like it has a probability of success of around say 57-58%. But Table 2 would lead me to believe this portfolio has a probability of success of around 47%. Could you please clarify for me? thanks

    ReplyDelete
    Replies

    1. Thank you Angela. You are right. Figure 8 is not consistent with the other tables and figures. We will check that again.

      Delete
    2. Angela, I've got it now. I forgot a basic point. For all of the earlier figures and tables, the baseline assumption is a 0.5% fee, not a 0% fee. So in Figure 8 you need to look at where the 0.5% fee is, and then things match up.

      Delete
  4. Dr. Pfau, Thanks for posting this paper - I enjoyed reading it. Although it paints a pretty grim picture for retirees, that is not surprising given what the US Federal Reserve has been doing.

    A significant point made was that the 'static' approach to SWR (i.e. 4% then adjusted for inflation) is flawed, but the paper then goes on to show that the static withdrawal rate of 4% is unsustainable under current valuations. If the static premise is flawed, is not the conclusion as well? I do think the study findings have meaning for dynamic withdrawal strategies, but I am not sure how they would be applied. I expect you would just adjust your real return assumptions for the dynamic model downward, but I did not find specifics on that point. Could you share the real return assumptions underlying this model?

    I assume CAPE applies only to US stocks. As I understand it, the CAPE is significantly lower outside the US. Also, it appears the model did not take into account use of inflation-indexed bonds in portfolios, such as TIPS and I Bonds. Would this materially affect the results? Thanks again to you and your partners for this work.

    ReplyDelete
    Replies
    1. Hi, thanks.

      Yes, the 'static' approach is flawed. A number of research articles, including some by myself and others by one of my co-authors, have shown that it is just about the worst possible strategy to use. Nonetheless, research centers on it as an easy way to explore what may be feasible. Any dynamic strategies should have lowered expectations as well. This is the spirit in which we are doing this.

      The real return assumptions depend on bond yields and the CAPE. When each are at their historical averages, the nominal (not real) stock return is 10% and the bond return is 5%. Note that the model starts off with lower returns than this because yields are low and CAPE is high.

      I don't think TIPS would materially affect the results. But it depends on how their returns are related to other variables. TIPS haven't been around for long enough to understand how they might behave at times when inflation is high, etc.

      Delete
  5. Dear Wade,

    Interesting paper. If I understand correctly, CAPE data only exists for the S&P 500 Index, or simulations thereof so equity returns and hence outcomes may vary for someone who holds equities in a Fama-style diversified portfolio. Also, while I understand for purposes of a SWR study it's natural and perhaps essential to assume annual inflation-adjusted withdrawals, I would argue that's the exception rather than the rule for actual retirees. That may indeed be the behavior of the typical retiree and Bogleheads.org poster, but I would point out that particular group is highly skewed towards high income/high net worth individuals who obtained that status via employment primarily in technical/engineering occupations. As such, while an engineer might consider such a withdrawal style to be perfectly natural I would argue its not the way the average retire withdraws funds from their nest egg.

    At any rate your work emphasizes the need for a retiree to remain flexible in what the can expect their portfolios to provide above and beyond their Social Security income. Which by the way, is the primary income source for many retired Americans further down the food chain.

    YKO

    ReplyDelete
  6. The article is interesting. One point that I have not seen mentioned much is how quickly valuations can change and how much that affects projected withdrawal rates. In approximately two months, the 10 year treasury constant maturity rate has gone up by about 80 basis points and the 10 year TIPS constant maturity (real) rate has gone up by about 120 basis points. This has large consequences for models that use initial valuation, especially if they assume a constant risk premium for equities so such changes propagate to equity returns. We should not be surprised to find that it's difficult to project a 30-year spending path; it's even difficult to figure out the starting point if it's two months away!

    ReplyDelete
    Replies
    1. That you for sharing a good point. On the one hand, rates have been rising, which will support a higher withdrawal rate for new retirees from today. On the other hand, as rates have risen, bond prices fall. From late April to late June, VIPSX (Vanguard's TIPS fund) lost about 10%. So a higher withdrawal rate from a lower account value may balance out somewhat. But certainly your point is well taken that all of these projections are made with great uncertainties.

      Delete
  7. I commend you on your reality-based approach to safe withdrawal rates in the current bond low-interest rate environment. A few thoughts:

    (1) With medical expenses in retirement tending to grow significantly faster than the general inflation rate (and taking up more of retirees' yearly budgets over time), does your analysis (or the 4% rule, for that matter) take into account that retirees' overall costs of living are likely to actually accelerate over time, even if the general inflation rate remains constant?

    (2) When you refer to bond returns in your analysis (or in the 4% rule), are you looking at bond *funds* (which can lose principal when interest rates rise) or bonds held outright to maturity (perhaps well laddered over many years)? The latter is a more robust way to protect principal, provide reliable cash flow, and potentially exploit rising interest rates (by reinvesting interest and/or maturing principal in higher yielding bonds).

    (3) Your paper abstract refers to April 15, 2014. I think you meant 2013.

    ReplyDelete
    Replies

    1. Thank you.

      About the spending path over retirement, I've written about it here:

      http://www.advisorperspectives.com/newsletters12/How_Do_Spending_Needs_Evolve_During_Retirement.php

      About bonds, I am referring to bond mutual funds. You are right, and I recently included a blog post on Asset Dedication, which talks about using fixed income to build a front-end income floor.

      Thanks about the typo notification as well.

      Delete
  8. Wade, thanks for another great paper.

    In your research do you consider the PV of a retirees Social Security as part of the bond allocation?

    Baylor U's Prof Bill Reichenstein advocated this approach in a May 2001 JFP contribution. Since the PV of a lifetime SS stream is sizable, it would indicate a higher equity allocation for retirees.

    What's your opinion of this approach?

    ReplyDelete
    Replies
    1. Hi,
      Though the PV of Social Security is not liquid, it does provide a secure income floor which increases one's risk capacity. This can allow one to be more aggressive in terms of spending and asset allocation. In this regard, you could think of it as part of the bond allocation, since it lets you use a higher stock allocation with your financial assets.

      Delete
  9. Dr Wade and interesting paper as always. One significant area of disagreement I have with your methodology is your somewhat arbitrary decreasing of the expected returns of the US equity market by 2% to make them more inline with the rest of the world. I recently read William Bersteins Ebook "Skating where the puck was:The correlation game" and I agree that future returns of the US market are unlikely to outperform the rest of the world by as much as the have historically.

    But what I don't understand is why the US returns should revert to the world's average rather than the rest of the world average return increase to the US level.

    Capital markets for the last 30+ years have been using the US as a model and gradually been adopting the best practices of NYSE, NASDAQ, SEC, and US investment banks etc. This has resulted in more transparency, more liquidity and lower transaction costs in international stock markets. One of the reasons that foreign capital markets (beside wars, and revolutions) have underperformed is because foreign capital markets in the past have been hot beds of crony capitalism and corruption. It is simply a lot harder for El Presidente's Son-In-Law to loot the telephone company of tens of millions of dollar in 2013 than it was in 1963. In short the world's capital markets have become more like America, so it seem logical that equity returns should be more like historical US returns going forward.

    In my simple model if I have 50/50 portfolio and my average equity return is 12% instead of 10% than my withdrawal rate should increase by ~1%

    I'd be interested in seeing what happens to withdrawal rate if you assume historical equity returns.

    ReplyDelete
    Replies
    1. Consider that we were an emerging market during much of the 1926 to date data that most studies use.

      Delete
    2. The US was hardly an emerging market in 1926, by 1913. The US economy was larger than the UK and Germany's combined.

      Delete
  10. I am having trouble understanding Table 3 in the paper which shows that as equity percent goes up, the initial withdrawal rate at a given probability of success goes down. Comparing the 90% success rate at 30 years you get 20% equities allowing a 3.0% IWR, 40% equities at 2.8%, and 60% equities at only 2.4%. That seems backwards to me as I thought higher equities would predict that you could use a higher withdrawal rate and not a lower one. What am I missing?

    ReplyDelete
    Replies
    1. It could be because the high CAPE implies lower future stock returns, but I agree that this finding is not consistent with most other studies. I think we need to double check some of the equations in this one.

      Delete
    2. Yes, please check and let us all know if the study's results are correct or not. Thank you very much.

      Delete
    3. Any update? I came to the same observation as 'Anonymous July 4, 2013 7:31am' ... it literally jumped off the page ... Wade, I respect your work and depend on this type of research to help me plan my transition from 'accumulator' to 'income streamer'. If Table 3 is correct, please elaborate further, if it's not correct, please update your paper. Thank you.

      Delete
    4. I've been talking with David Blanchett about this, and also I've been working separately on my own methodology.

      Hopefully by early next year I'll have a Retirement Dashboard ready, which I'll aim to update monthly. For one thing, the withdrawal rates from these models can quickly become out of date as PE10 and interest rates change. A frequently updated report would be the appropriate way to update this.

      I'm still working through how much emphasis to put on the idea that high PE10 will drive lower future returns, as there are many arguments now to justify PE10 staying at a higher than average level. I do agree with both of you that it is odd for higher stock allocations to produce lower withdrawal rates, and this is surely a result of the link in the model (based on the historical data) between high PE10 and low future stock returns.

      Delete
  11. Excellent research Wade.

    Can you provide us a simple suggested withdrawal rate based on the CAPE each year? I guess retirees could check CAPE each December and decide how much is safe to withdraw for the next year. Another idea would be for pre-retirees (accumulators) to look at CAPE each December and decide on their stock-bond split for the coming year.

    Many thanks Wade.

    Al

    ReplyDelete
    Replies

    1. Al, regarding your first question, I think there are probably better ways to decide on the next year's withdrawal rate than using CAPE, because it is a long-term measure and can be quite noisy over short horizons. So, I don't think trying to find an answer to your first question will be all that helpful if you are seeking a dynamic year to year withdrawal rate.

      About your second question, I have written a couple of articles about that subject.

      You can find these at:

      “Long-Term Investors and Valuation-Based Asset Allocation.” Applied Financial Economics (August 2012).
      http://ideas.repec.org/p/pra/mprapa/29448.html

      “Safe Withdrawal Rates, Savings Rates, and Valuation-Based Asset Allocation.” Journal of Financial Planning (April 2012)
      http://www.fpanet.org/journal/WithdrawalRatesSavingsRatesandValuation/

      Delete
  12. I would very much like to withdraw less from my retirement accounts given that I continue to work. In my case, I am 71, and the IRS requires a minimum withdrawal based on their estimate of my life expectancy -- the IRS of course fails to take into consideration that the men in my family life into their 90's. Fortunately they drank and smoked or they would have never died.

    ReplyDelete