Asset allocation puzzle

[noppenbd]

I was doing some reading on asset allocation recently and this post really surprised me:

The Personal Financier: The Relation between Age and Portfolio Risk – Counter Intuitive Results

The idea he is putting forward is that starting with the riskiest asset allocation when young (more stocks than bonds) and gradually reducing to a conservative AA when nearing retirement (more bonds than stocks) is not necessarily the best strategy.

I put this idea to the test with the Flexible Retirement Planner ( www.flexibleretirementplanner.com ), which does Monte Carlo simulations to determine the probability of success. I was blown away by the results. First I started by assuming a 25 year old retiring at 65, with life expectancy 95. I left the inflation, tax, and spending numbers at their defaults. I used zero for the portfolio starting value and assumed $11K a year savings in tax deferred accounts. For the first run I assumed the asset allocation got incrementally more conservative every 10 years:

Aggressive from age 25-35
Above Avg Risk from age 35-45
Moderate Risk from age 45-55
Below Average Risk from age 55-95

Chance of success? 68.9%, with a Median Portfolio Value at Retirement of $1.2 million.

Then I inverted the asset allocation, making the AA go from conservative to risky as the age increased. This seems counterintuitive at first. So the AA went:

Below Average Risk from age 25-35
Moderate Risk from age 35-45
Above Avg Risk from age 45-55
Aggressive from age 55-95

Guess what? The chance of success increased, pretty dramatically, to 77.2%, with a Median Portfolio Value at Retirement of $1.5 million.

This seems really strange until you factor in the portfolio size. By keeping the AA conservative when the portfolio is small you give it a chance to grow at a regular rate. As the portfolio grows (and retirement nears) you increase the return by ratcheting up the risk level. Apparently the larger portfolio is able to absorb the regular withdrawals of retirement more easily even when there are down years.

This seems to go against all the common knowledge about asset allocation. I would like to play around with it some more, but what do you think about this? Does it have some validity?

[feh]

I think it only has validity if you're willing to see your portfolio drop in value by 30% the year before you retire.

I haven't used the tool you reference, but I imagine the distribution graph of portfolio value at retirement age for the second case would be interesting/scary.

[simpletron]

if you have a vanguard account, check out this calculator

https://retirementplans.vanguard.com/VGApp/pe/education/calculators/RetirementNestEggCalc.jsf

it is a monte carlo simulation on "will your nest egg last during retirement". if you play around with it for a while it shows that you will want 70+% stocks with 80% stocks being a winner most of the time. 80% stocks corresponds with roughly the above average risk on that calculator.

your results make sense because you always need growth. I have never understood why you would want to go conservative(less than 60% stock) when in retirement because you will still need money in 10, 20, 30 years. if you withdraw 4% and inflation is 3-4%, then you need 7-8% growth just not to lose purchasing power and to get that growth you need stocks.

try running the calculator at just aggressive. 1.7 million at retirement with 82% chance of not running out of money.

My asset allocation plan has always been go 100% stock until about 10 to 15 years out, then slowly convert to 80% stock, 16% bonds, 4% cash. use the cash for funding current year, bonds replace cash every year and stocks replace bonds only on good years. the 4 years worth of bonds acts as a buffer against stock market down turns.

note: the second link doesn't have a ) at the end.

[simpletron]

I think it only has validity if you're willing to see your portfolio drop in value by 30% the year before you retire.

I haven't used the tool you reference, but I imagine the distribution graph of portfolio value at retirement age for the second case would be interesting/scary.

even if your portfolio drops by 30% the year before you planned to retire you can push off retirement a year or two which greatly increase your chances of not running out of money.(less years + more savings). also with the conservative to aggressive case you start out with 20% more on average according to the calculator so you can absorb that kind of blow and have more growth potential to rebound.

[noppenbd]

I tried simpletron's suggestion, and the truth is that you get the best survival results when you stay Aggressive/Above Average Risk as long as possible.

I guess the question this raises is in regards to all of those Target Retirement funds that are so popular. The current allocation for a retired individual would be 68% bonds and 32% stocks (for Vanguard). Maybe those TR funds are not such a good idea after all?

[feh]

I tried simpletron's suggestion, and the truth is that you get the best survival results when you stay Aggressive/Above Average Risk as long as possible.

I generally agree more aggressive is better. I am (hopefully) about 10 years from retirement, and I am still 100% stocks.

I guess the question this raises is in regards to all of those Target Retirement funds that are so popular. The current allocation for a retired individual would be 68% bonds and 32% stocks (for Vanguard). Maybe those TR funds are not such a good idea after all?

What I think some people might be missing is the volatility such an approach could bring. I don't doubt that on average, you'll have more money staying aggressive. However, the distribution of possible portfolio values will be much wider. In other words, there's a significant chance you could end up with less money going the aggressive route than the conservative route, even if the average value is higher for the aggressive plan.

I'm not doing a good job explaining this - a proper explanation would be steeped in statistical terms, which is probably not worthwhile to most folks. I wish I could draw a couple pictures - that would help.

It all depends on how much risk you're willing to endure.

[jIM_Ohio]

When in the withdraw phase of a portfolio, the risk is volatilty (not a given return). So maybe 80% stocks shows something positive because in a given 30 year time period that 80% trended upwards.

Your biggest risk when retiring is a down year or two or three when you start retirement- if you have to draw down a portfolio which is going down, retirement might be short lived.

So the conservative allocation is designed to reduce volatility, not increase returns.

I also would not be retiring on 68% or 77% success rates. Shooting for above 80 and probably above 85. I don't have time to look at the calculator now, but in general I have not found one calculator which covers both accumulation and withdraw well wrapped into one. Most calculators I have seen can help predict one or the other, but none I have seen do a good job of telling me how much to accumulate, then also accounting for how I withdraw.

I like Firecalc and raddr as the two spend down calculators of choice. Again looking for 80-85% success rates for me.

[jIM_Ohio]

even if your portfolio drops by 30% the year before you planned to retire you can push off retirement a year or two which greatly increase your chances of not running out of money.(less years + more savings). also with the conservative to aggressive case you start out with 20% more on average according to the calculator so you can absorb that kind of blow and have more growth potential to rebound.

If I have $1 M to generate a 40k income stream and it drops to $700k, I would need a 42% one year return to recover from the 30% drop. Getting that 42% back even in 3 years would be tough.

My advice would be to go conservative closer to retirement to reduce the volatility risk.

[noppenbd]

However, the distribution of possible portfolio values will be much wider. In other words, there's a significant chance you could end up with less money going the aggressive route than the conservative route, even if the average value is higher for the aggressive plan.

I'm not doing a good job explaining this - a proper explanation would be steeped in statistical terms, which is probably not worthwhile to most folks. I wish I could draw a couple pictures - that would help.

It all depends on how much risk you're willing to endure.

I'm not sure this is true. The simulator calculates your success rate by running 100s (1000s?) of possible scenarios given your return and standard deviations. The success rate is defined as the number of scenarios where your money did not run out divided by the total number of scenarios run.

So a higher success rate means there were fewer numbers of times you ran out of money.

[noppenbd]

I also would not be retiring on 68% or 77% success rates. Shooting for above 80 and probably above 85.

Agreed. This was just a hypothetical example, trying to look at different asset allocations relative to each other.

[jIM_Ohio]

I'm not sure this is true. The simulator calculates your success rate by running 100s (1000s?) of possible scenarios given your return and standard deviations. The success rate is defined as the number of scenarios where your money did not run out divided by the total number of scenarios run.

So a higher success rate means there were fewer numbers of times you ran out of money.

True- but running out of money 30 times or more is too much for me. it is 100's of scenarios- a monte carlo used past market performance in the sequence they happened to back test a portfolio (using past performance to give a probability of success).

Success means the withdraws had money in the account to withdraw. Success has little to do with standard deviations (standard deviation of returns can be calculated/back tested, but to best of my knowledge this is a calculation and can only be controlled from the asset allocation).

[jIM_Ohio]

Agreed. This was just a hypothetical example, trying to look at different asset allocations relative to each other.

My point was a 76% success rate is not really success.

[jIM_Ohio]

I have used this calculator before.

I ran some general runs through this calculator:

ages 25/65. Inflation 3.5% tax rate 20% $11000 annual tax deferred savings. 50k retirement spending. Moderate risk (41% stocks, 45% bonds, 14% foreign).

Success 71.3% (site suggests this is not a good enough percentage to user). $1.2 M accumulated (SWR=50/1200=4.2%).

To get success higher than 80% I reduced income need to 43k and site even gave a caution. Site did not give "go" until spending was dropped with 36k (90% success was when web site gave it a go). 36k is a 3% SWR.

went back to original inputs with a above average risk (62-17-21). Kicked out an 80% success (caution). Had to reduce spending to 37k again to get 90% success. Portfolio at retirement was 1.6 M. 37k is a 2.3% SWR.

key conclusion- if two asset classes give same result/percentage, go with the least risky portfolio.

same original inputs, but an aggressive investing style (70-7-23). 80% success with caution. Had to reduce spending to 37k again to get 90% success and the go. 1.7 M is portfolio value and 37k represents a 2% SWR.

The trinity study (which used withdraw rates to calculate simulations like this) suggests a 4-5% starting withdraw rate. If you are seeing 2% withdraw rates with less than 100% success you are taking on way too much risk.

The article is flawed in this regard. Because what a person will do is use the spending to decide when to retire, and shift to an asset allocation which produces that spending level.

I could not use the calculator to run this type of simulation. Aggressive until age 65 (balance should be 1.7 M) then spending at 68k level (4% SWR) with a moderate risk (55-45 portfolio). My guess is the interest on the 55-45 portfolio could generate the income needed for about 10 years before inflation kicked in.

If a 3% withdraw rate is being used on an aggressive portfolio, the success rate should be 100% if the investment is in dividend paying stocks.

Again, my conclusion is the blog which was linked to the first post is misguided information- I would like to see how they simulated the returns for the various risk levels of the portfolio for a given age, because even with the additional inputs option, I could not see those same results- maybe the std deviation and returns input were inaccurate?

[noppenbd]

True- but running out of money 30 times or more is too much for me. it is 100's of scenarios- a monte carlo used past market performance in the sequence they happened to back test a portfolio (using past performance to give a probability of success).

Success means the withdraws had money in the account to withdraw. Success has little to do with standard deviations (standard deviation of returns can be calculated/back tested, but to best of my knowledge this is a calculation and can only be controlled from the asset allocation).

Depends on the calculator inputs. FRP and firecalc will both let you put in a return along with a standard deviation. FRP then generates a string of returns based on the combination of the two (not based on historical returns). Firecalc has the same option on the Portfolio page. I prefer to use these because it is more transparent & direct than selecting a stock/bond allocation.

30 times running out of money is equivalent to a 85% success rate with 200 iterations, so I don't really see how that is a problem. Any higher than 85% and it is pretty much noise anyway.

My point is that the typical advice tells you that you will have a better chance of success (meaning statistically you will run out of money fewer times) if you get conservative when you near retirement. But in reality you have less chance of running out of money if you stay weighted towards stocks, even with the increased volatility (because the growth outweighs the volatility).

Now whether one has the stomach for it is a different matter. My mother is getting ready to retire and I really couldn't fault her for going ultraconservative at this point.

[jIM_Ohio]

Depends on the calculator inputs. FRP and firecalc will both let you put in a return along with a standard deviation. FRP then generates a string of returns based on the combination of the two (not based on historical returns). Firecalc has the same option on the Portfolio page. I prefer to use these because it is more transparent & direct than selecting a stock/bond allocation.

30 times running out of money is equivalent to a 85% success rate with 200 iterations, so I don't really see how that is a problem. Any higher than 85% and it is pretty much noise anyway.

My point is that the typical advice tells you that you will have a better chance of success (meaning statistically you will run out of money fewer times) if you get conservative when you near retirement. But in reality you have less chance of running out of money if you stay weighted towards stocks, even with the increased volatility (because the growth outweighs the volatility).

Now whether one has the stomach for it is a different matter. My mother is getting ready to retire and I really couldn't fault her for going ultraconservative at this point.

My 30 times was 30 times out of 100, not 30 times out of 200.

I agree that anything above 90% is noise. There are too many times when the great depression gives 5-10 cases of a given allocation not working. In that case own gold, guns and ammo.

It shows "statistically" that growth will "win out". But the key to that whole puzzle is what the market does those first 3-5 years of retirement. If you get by those first 3 to 5 years with a risky portfolio doubling or tripling the needed withdraw (each year), your risk level goes down. The calculator does not take into account how the withdraw is taken.

Examples of withdraw techniques:
1) dividends only
2) cash reserves/ buckets theory
3) spend down principal

the calculator assumes 3 even though 1 and 2 will give much higher degrees of success (based on what others tell me).

Experience and discussing with people which have done this or are close to doing this suggest that a better way to hedge bets is to reduce risk closer to retirement and base conclusions on withdraw rate, not portfolio growth. For example- many people can use buckets of money (spending bucket, income generation bucket and growth bucket). The dividends, interest and returns of the income bucket go to the cash bucket, and that transfer only happens in up years (do not sell into a falling market). The second bucket would be moderate risk (lots of bonds which generate interest). It follows the "more conservative" approach the blog is trying to refute... and I'd be hard pressed to follow a blog using numbers as opposed listening to people which are doing this every year.

Like in your mother's case, she has risk for her, and no one is there to bail HER out, except either SS or her returning to work. So when there is so much at stake, reducing risk to get a higher success rate makes sense. Success rate=withdraw rate, not portfolio returns.

The issue is "what is success"- the calculator mentioned (and the article) base success on returns and portfolio value. No input for withdraw rate, which is generally the pinnacle of pre-retirement planning and withdraw planning.

Again I saw some cases with a 2% withdraw rate used- that should be 100% success (yet the site returned 80%)- I can get a 3.5% return on dividends alone and dividends of the S&P 500 have grown faster than inflation over time, so the portfolio as a whole should just go up-up-up in this regard anyway with a 2% withdraw rate.

I have a fund which is down close to 15% ytd, but it's dividend payout is higher this year than last (payout is dividend paid per share). So the calculator is not designed for specific withdraw scenarios, it just crunches the numbers and returns and makes it's decisions based on the portfolio return, not the dividend or specific spending patterns (only spend dividends, for example).