Weekend Musing: getting super right

This is a guest post from long time reader Jackson, who I have been in correspondence with since I started my articles on superannuation. The following analysis was done completely independently of my own research:

There has been a lot of recent chatter about the asset allocation of superannuation, with an emerging contrarian view that the majority of Australian superannuation accounts are overweight in equities, and underweight in fixed interest components. Whilst it’s pretty obvious that there is no ideal allocation that you can “set and forget”, the majority of accounts end up this way.

As an MB tragic, what drives me nuts is the absolute lack of quantitative analysis out there. What does the data actually say? Can we make any quantitative sense of it?

I thought it might be useful to present some results of my own analysis in this area. My objective was pretty simple – what is the probable outcome for my superannuation, if I use a pretty dumb set-and-forget strategy? Of course I’m not going to do that, but framing the problem this way should give a conservative set of outcomes, and remember most accounts end up this way.

The approach was as follows:

  1. take the data from annual returns posted by The Prince here,
  2. build a statistical distribution from the data,
  3. create the ability to choose an asset allocation that varies in time, and
  4. run a Monte Carlo simulation to determine a distribution of possible outcomes from that choice.

To simplify the problem, I chose a 30 year time horizon, an initial account balance of $100K, and annual contributions of $10K. All calculations are in today’s dollars and fees have not been accounted for. There are only two choices for asset allocation – fixed interest and equities. I used both the Australian fixed interest and equities data, as they are correlated well with their international equivalents but the positive returns are higher and the negative returns are less negative.

The asset allocation can be changed once annually at any time over the 30 year time horizon. The analysis of course assumes that the statistical distribution of past performance is an indicator of the statistical distribution of future performance, which may well not be the case but is the same information all advisers and commentators make prognostications upon.

I report three numbers from each simulation – the 25th, 50th and 75th percentile. You have a 25% chance of ending up with less than the 25th percentile (ie 1-in-4 of getting less than this value), a 50% chance of ending up with less than the 50th percentile, etc.
Let’s start with the dumbest of the dumb – a single asset allocation for 30 years, either cash, fixed interest or equities.

 

An interesting result – whilst the 50th percentile (the median) gets larger with the more risk you are prepared to take on, the 25th percentile gets smaller. In other words, for cash only you are 25% likely to end up with less than $1.1M in the example above, but for 100% equities this number is $850K – a worse result.

The distribution is far more spread for equities, both to the high and low end. If you’re an optimist, it’s all roses in equities. If you’re a pessimist, or just plain unlucky, then maybe you want to bolster your portfolio with something a little less volatile.

With those building blocks in mind, the issue is really one of limiting the downside risk whilst preserving the potential upside – the absolute return approach. The premier question is what combination of fixed interest and equities will do just that? What gives you a pretty solid 25th percentile outcome (ie a 3 in 4 chance of doing better than that), with a rosy picture if you get lucky?

I consider four cases:

  1. a 75% fixed interest, 25% equities split for the entire 30 years (75/25);
  2. a 25% fixed interest, 75% equities split for the entire 30 years (25/75);
  3. the 75/25 split for the first 20 years, moving to 100% fixed interest for the last 10 years (75/25 =>100 FI); and
  4. the 25/75 split for the first 20 years, also moving to 100% fixed interest for the last 10 years (25/75 =>100 FI).

These last two are meant to replicate the oft-quoted approach of moving to less volatile asset allocation as you approach retirement.

Given the first set of results already shown above, there should be little surprise that the 75/25 fixed interest/equities split gives a higher 25% percentile, but lower 50% percentile, than the 25/75 split. So are you a glass half-empty, or glass half-full person? Do you want to take a 1-in-2 chance on your retirement balance, or a 3-in-4?

For the final two examples, those that move to fixed interest in the last 10 years actually do worse than if you’d left it alone – all percentile values are lower. This seems to suggest you’ve chopped off the upside opportunity in the last decade of your working life, and the decision to move into a low volatility asset allocation is more likely to be about risk aversion than a statistical assessment of probable outcomes.

There are an infinite number of potential combinations, but I’ve found the ones described above the most instructive. Much to my surprise, the outcomes of this are pretty much in line with the barbell approach The Prince was discussing late last year (in a nutshell, be overweight fixed interest against a small allocation to shares). This analysis was done completely independently.

I’m more than happy to provide the spreadsheet used to create this freely to anyone, so long as you accept it’s assumptions, failings, projections and outcomes do not construe any kind of advice on my part. Constructive criticism welcome!

Comments

  1. Thanks Jackson, a top piece of work, and good to see the quality of analysis/thought driving discussion on MB.

    I occasionally wonder about the ability to time the market when it comes to superannuation. If I watch my super obsessively in the final decade or 15 years of my working life, and change to fixed interest when the amount is at a defined point (eg equivalent to 8% compound) could I achieve a better outcome? The 10 or 15 year window gives a decent amount of time, and if I go in with a pre-defined target, then this might maximise my super. I know you wrote that those who switched to fixed interest in the last decade were worse off, but my question is slightly different.

    Is there any reason this analysis and portfolio structure cannot be applied outside of super to personal savings?

    • Thanks Monkey. To your first question, what I present assumes no intelligence at all – that you don’t learn anything as you go, and that you can’t identify if you’re in a secular bull/bear market, etc. Applying some intelligence should result in a better outcome, but could result in a worse outcome.

      To your second question, I don’t see why not, and I’m guessing (but don’t know) that it underpins a lot of how traders operate.

      Fundamentally the above boils down to the fact that if you make a 5% loss, you need to make more than a 5% gain to get back to where you started from. So, do what you can to not make losses, or at least limit them (traders call these ‘stop losses’). The way to limit them in the above is to have most of the funds in non-volatile classes, so even in a bad year for equities the weighting stops the portfolio from returning a negative result.

  2. Interesting thought process Jackson. I have thought that the diversification principles often sprouted from “knowledgeable sources” doesn’t take into account diversification of risk. Therefore if you were to be placing your investments on a daily basis how much would you allocate to each sector. I think to manage risk you would be looking at investing more in the fixed interest sector.
    Further the larger your portfolio the more use of futures would be necessary to manage the downside of the equity sector of a portfolio, by this i mean that if you lose 10% of your capital on a equity position then you need to either take on more risk to bring it back up or ?? well its easier to build wealth by not losing capital ?? (this really is the holy grail that hedge funds attempt)

    • A fair bit of your question is way outside my expertise Jack. “it’s easier to build wealth by not losing capital” – yes, see my reply above.

  3. “The analysis of course assumes that the statistical distribution of past performance is an indicator of the statistical distribution of future performance”

    This is the bit I’d love to see more discussion on. We’ve gone through a boom period over the last 20 years or so, some of which was driven by deregulation of financial markets. With the frequency of crises these days (asian 97/98, tech 01, gfc 07) could the markets be said to be overly deregulated?

    Furthermore, given you can’t re-apply the same deregulation twice, and that people are in debt up to their eyeballs, wouldn’t that suggest that the next 30 years are unlikely to be as ‘boom-like’ as the previous 30 and the statistical distribution may be completely different?

    Thoughts from a layperson, so perhaps misguided?

    • Not misguided at all, Sidamo. I don’t have any clue what it will be, but nor do any of the professional advisors – they use the same data as I did, and end up in a different place.

      The reason for this, I’m guessing, is they are transfixed on the “average”, ie the “median (50th percentile)” above. If you want to maximise the 50th percentile, the above shows you go overweight equities, ie the traditional advice. My fundamental beef with that is that there is a range of possible outcomes, and there’s a 50% chance you’re not going to get to the average. If you told Joe Punter there is 1-in-2 chance they won’t get to the average, they’d probably be concerned.

  4. Hi Jackson, interesting work – well done.

    One approach to portfolio allocation I’ve long thought of is Ben Graham’s recommended approach in the first edition of the Intelligent Investor.

    The idea is to maintain a 50/50 split of fixed income to equities, re-balancing every year or two, or whenever the split deviates by more than 5-10%. The idea would be that if equities have had a good year, or fixed-income a bad year, you are allocating capital away from the asset class that has gone up (ie getting more expensive), and buying the one that has gone down (become cheaper). It would naturally result in buying fixed interest when yields increase, or investing in equities when the market has dropped. And on the sell side, it would be taking money off the table after a bull year in stocks, or selling fixed interest when yields drop.

    It would be interesting to see how this approach performs using your model. I guess the one problem that I can see is that to an extent, it relies on the probability distribution of future returns being dependent on the returns of previous years, (for instance, as a bull market runs, the chances of a correction increases with time since the last correction) – although I am not sure whether this assumption is actually born out in the numbers, or just a gut instinct.

    If you think it could be modeled without too much pain, I’d be more than happy to play with the spreadsheet.

    • Really you’re talking about introducing some intelligence into the mix, probably a useful next step. Not a straightforward thing to include, but let me have a go at it.

    • Actually, thinking about it some more, this is actually what the above does, but at the 75/25 or 25/75 level, once per year.

      I’m pretty confident the 50/50 split would fall between the two cases presented in the second figure.

  5. Minus Zero-Sum Game

    One other potential tweak to the equity component of superannuation portfolios is to apply a pre-set moving average crossover rule (e.g. 30/150-day EMA combination). A free interactive charting program such as Yahoo Finance is sufficient for this purpose. The main advantage with longer-term moving averages is that they minimise equity value drawdown. By using this simple trend-following MA crossover system, an investor would have received an exit signal around Jan 2008 (All Ords). The next entry signal was indicated circa June, 2009. Hence, in this example, equity drawdown was significantly reduced during the GFC share market decline compared to the buy and hold method.

    I’d like to eventually see some MB articles on using longer-term moving averages when used with superannuation.

  6. Nice analysis Jackson. But as we known there are two phases to superannuation. Accumulation and pension phase. When investment markets can last 10, 20 even 30 year bull runs. You need to consider what is best when it comes to drawn down phase. Sure shares and cash are fairly liquid but if you have a really bad couple of years and don’t have time for the market to recover, cash looks a lot better, when it comes time to drawn down.

    It annoys me a lot when I see clients come in wanting to buy geared property in super funds when they are at the very end of their working lives, won’t be contributing (or be able to contribute) for much longer. Hoping that property will keep going up on the capital gains side.

  7. Nice piece of work, it seems.

    You used illustrated distributions of data before processing them – so you gets hugs from me!

    More of the same, please – there is not enough done, anywhere.

  8. Hey Jackson interesting work. How do I get a copy of your spreadsheet and working as mentioned?

    • Already included Freddy, the original data used for Australian Equities is the S&P/ASX 300 Accumulation Index, referenced in Prince’s original post.

      • Well then that is where your comparison is flawed. Most funds will only deal with ASX200 stocks. In fact many attempt to allocate shares so that their funds closely approximate the ASX 200 index.

        Over that time period the ASX 200 Accumulation Index will have you closer to $3 million gross. About $2million after taxes.

        Also, doesn’t underestimate fees. 2% over 30 years works out to be half of the money. That means your choice of fund manager is just as important as the asset class.

        • Freddy, I’ve re-run the numbers using your data. Reporting 25th percentiles

          Original results above
          75/25: $1.1M
          25/75: $1.05M

          Using your data:
          75/25: $1.0M
          25/75: $0.8M

          The conclusions still hold.

          I completely agree with you on the fees, moving to a low cost fund is actually the most important decision a person can make.

  9. “with an emerging contrarian view that the majority of Australian superannuation accounts are overweight in equities, and underweight in fixed interest components”

    this is not a contrarian veiw. this is the new consensus veiw. this would have been a a contrarian veiw when the share market peaked in 2007 but is not a contrarian veiw now, going underweight equities after the market has crashed. even the governemnt is weighing in on this debate and shares this very wrong veiw. a goverment that encouraged investors to put a mill $ into equity based super in 2007, right at the top. a governemnt that encouraged first home buyers to get into property, right at the the top. now they want a higher allocation to fixed interest? the smart money is doing the exact opposite and moving from already underweight equities to overweight equities. what any smart investor does after a market crashes, they go overweight, not underweight.

    • GB, the intention was to take a long term view, irrespective of the current situation. Active, intelligent management should do much better than what I present.

      “Emerging contrarian” actually doesn’t make much sense on reflection – it’s either contrarian or it’s not.

  10. Nice start on the asset side of things Jackson; the drawdown side Roscs003 refers to is just as interesting. And it highlights that timing is brutally important because for most retirees who consume capital to fund some of their regular drawdown for living costs, capital losses amplify the proportion of that drawdown against your capital and hence your participation in the next rally is much less than for a younger accumulator.
    The issue is much less pronounced if you have enough capital to live off the dividends only – not that many retirees really.

    I did some simulations because I was thinking about my parents – for example … if you stop working at the top of the market in 1966, when Ming and growth were still going strong, by the mid 1970s the capital losses on equities and fixed income were so great that if you are relying even on a modest level of drawdowns your retirement funds are near exhaustion by the mid 1970s … unless you were in cash, but inflation was eating you alive there too.
    So … this is why a seemingly sub-optional outcomes from the asset perspective makes a little more sense from the drawdown perspective. It also demonstrates the importance of active management if you can do it right. Of course you might be more focussed on the grandkids by then … and your health.

    As an aside … I agree with GB, consensus views that equities are “too risky” tend to coincide with equity market bottoms. And consensus calls to buy an asset that has historically performed tends to coincide with a top. A well known Warren Buffett quote on consensus seems appropriate.

    • Interesting comments Curious, thanks. All of this is sub-optimal of course, you’re never going to get the perfect result consistently. The asset/drawdown phases also assume the binary work/retire paradigm continues, which I’m not sure it will for much longer.

      I don’t disagree with GB’s comments, they need to be taken into account if you’re going to get into active management. Unfortunately most people don’t even get the order 0 stuff right (ie a single account for starters…).

  11. What correlation assumptions have you made between asset classes, or have you modeled the distributions jointly?

    • The data shows no correlation between them at all, r values of way less than 0.05.

      Based on this, the Monte Carlo chooses two random numbers, one for each class, as they’re not correlated.