Tuesday, October 1, 2013

Stocks for the Long Run?


David Blanchett, Michael Finke, and I have finished a research article called, "Optimal Portfolios for the Long Run." This post is the second is a new weekly tradition of a video blog post made in cooperation with The Wealth Channel at the American College. Based on viewer feedback, the background music has been removed when I'm talking. I summarize the article in the video:




For email readers, the videos never show up in the email, but you can see the video by clicking here.

13 comments:

  1. An excellent post Wade. Your comments at the end on Monte Carlo simulations brings to mind a couple of points we've talked about via email on our paper we're currently working on together with Dr Mitchell.

    First, the methodology using Monte Carlo should string together a series of simulations. This would more closely simulate annual reviews and adjustments, one adjustment of which may be changing the allocation. So, the first thought is programming Monte Carlos to be able to string simulations together, and secondly, to be able to adjust the allocation within the simulations. Presumably, this adjustment would be towards less equity since the time frame is less (less longevity remaining as the simulation ages the retiree). However, this may not be the case if a significant portion of the retiree's income was already bond like (pension and/or Social Security); in this case, a rising equity glide path as you suggest in another post may be appropriate, especially if that fixed income is adjusted for cost of living.

    The second thought relates to our email exchange about returns trends. Under a serially connected series of MC simulations, the returns data may also be adjusted.

    As we've seen, serially connecting MC results provides deeper insight into cash flow and portfolio balances.

    This is how research keeps moving forward by building on the thoughts and observations of other good research.

    ReplyDelete
    Replies
    1. Larry,

      Thanks. You are getting at an important issue related to how asset allocation should change over retirement, and how those decisions would also be impacted by serial correlation in asset returns. One "benefit" of the usually i.i.d. assumption is that leaving out mean reversion is more conservative than assuming markets will tend to go back up after heading down. Mean reversion reduces both the downside and the upside.

      Delete
  2. This comment has been removed by the author.

    ReplyDelete
  3. Please excuse my novice-level question, but after reading the research article, I cannot determine which type of stocks you and the other presenters are describing. By ‘Stocks’, do you mean pure purchases of individual companies with dividends reinvested or do you mean all equity type investments, such as Stock ETFs, Stock-Heavy Mutual Funds, and Individual Company Stocks? You may have addressed this purchase choice theory in an earlier article, so I apologize if this seems like a simple question.

    Spreading the equities portion of a retirement nest egg over 10-20 individual company stocks and holding them for long run after only a few initial purchase costs is way less expensive to own than constantly paying annual management fees on mutual funds ($$) and ETFs ($) for the same long run. For example, if the portion of an investor’s nest egg invested in equities was $500,000, then the investor would pay on average $5,000 (average 1%) annual management fees for the luxury of having a fund manager buy and sell stocks for us. One single year’s worth of management fees could cover 250 buy/sell/rebalance transactions (average $20), which could cover the total investing period for those 10-20 individual company stocks. The investor would also be able to choose their income tax implications with separate sell transactions at their discretion.

    Thanks for another great article.

    ReplyDelete
    Replies
    1. Paul, when talking about stocks, I'm assuming investment in a low-cost index fund. The implicit assumption is usually that the index fund is for the S&P 500.

      Thank you, Wade

      Delete
  4. Wade --

    This is another TERRIFIC report from you.

    For me it raises two big questions:

    1. It suggests that in Monte Carlo simulation, reversion toward mean should be built in. Do you have suggestions as to how this can best be done?

    2. Do you have concerns that this makes the simulation even more dependent on the assumption for return-rate arithmetic mean, which for the future we do not know?

    Dick Purcell

    ReplyDelete
    Replies
    1. Thanks Dick. Those are still the great mysteries!

      Delete
  5. Wade,

    Thank you for sharing your research.

    Ever since I read "Winning The Loser's Game" I have had this nagging feeling that my equity allocation is too low. Bill Bernstein's recent ebook "Deep Risk" was like throwing fuel on a fire. Now your research is like adding liquid oxygen to the fire.

    I am a very low withdrawal rate person who really is investing for his heirs. I'm starting to think I should act and gradually begin increasing the amount of equities I hold, especially foreign equities. I just hope I'm not succumbing to recency bias.

    ReplyDelete
    Replies
    1. Thanks for the feedback.

      If your withdrawal rate is low, you might think about how much it would cost to build a bond ladder / use income annuities to ensure that your lifestyle can be met. This would free you up to worry less about what happens with the rest of your portfolio... i.e. you could focus more on growth without risking your lifestyle. This is what Bill Bernstein means about stop playing the game if you've already won.

      Delete
  6. Would it be possible for you to post a transcript of these video sessions? Thanks Matt

    ReplyDelete
    Replies
    1. I'm sorry, I recorded these based on some handwritten notes and haven't taken the time to transcribe them.

      Delete
  7. A few months ago I took the S&P 500 and grabbed the data from Yahoo and calculated Standard Deviations for different time periods. The SD showed an interesting function of time. If you use in a monte carlo the monthly SD you would get wildly different results than if you used an annual SD. There appears to be a short term variation that disappears as you hold a stock to 6 to 10 years.

    The second thing, Montecarlo or any other algorithm that predicts the future based on the past has a significant flaw. I'm sure you know this and this is why your working so hard to improve your simulation. The flaw is, to make it obvious, is the past will never model the future. But it is the best we can do. What we need is a "Safe Plan". That is how do I monitor my portfolio on a monthly, quarterly or yearly basis and ensure that by boat is not filling up with water so to speak.? Then, how do I pump the water out by managing my spending rate? The failure of the 4% rule is a 1/30 to 2/30 failure rate. Maybe there should be less concern about the 4% rule's failure rate and more advice on how to know when it is failing and how to catch the failure. I wrote my own montecarlo using annualized growth rates and Standard Deviations. I found that if I do minor corrections when the market dips I can eliminate totally running out of money. What Algorithm I used was to calculate the P99% expense (probability of success to 99%) at the beginning of each cycle and if it was lower than my planned expense, I lowered my spend rate.

    I have a problem with the 4% rule in general is that I have 44 years to 95, I only have one stream of income coming in now and 4 more to come in later. I need to figure out how to fill the gaps between and before the other streams of income come in.

    ReplyDelete
    Replies
    1. Ralph,

      Thanks for sharing. Do you mean, when you simulated based on monthly returns and then annualized the results, you got different outcomes than directly using annual? Was the distribution wider?

      About making spending adjustments, what you are describing sounds related to the "mortality-updating constant probability of failure" approach. I saw you made a comment at a post which also links to that. Indeed, reducing spending after market declines can make a big difference.

      About your last point about dealing with different spending needs at different points in time, that is one of the big shortcomings still in the research. I understand that David Blanchett is currently working on the problem.

      Delete