Total Portfolio Allocation and Withdrawal (TPAW)

[Ben Mathew]

Background:
I was using the planner to explore the impact of liquidity constraints on discretionary consumption. To do this, I set up 2 plans.

Plan 1: Unconstrained

• Plan begins at age 65 (retired) and ends at age 100.

• No Legacy

• No changes to default preferences other than to not have risk tolerance decrease with age

• No changes to default expected returns and volatility other than to turn off bond volatility

• Only asset is the current portfolio with a balance of $1,144,324

• Link to dateless plan: https://tpawplanner.com/link?params=IwZ ... QzMJMGzm74

Plan 2: Constrained

• Same as plan 1 except the only asset is a real monthly income stream of $4,000 (the present value of which is $1,144,324)

• Link to dateless plan: https://tpawplanner.com/link?params=CUU ... 4ZnVIWiiKc

I was surprised to see that the current period consumption for both plans is the same at $3,663. The implication of this is that the presence of a liquidity constraint does not affect current consumption [...]

Yes. The basic lifecycle model that we're using assumes no borrowing constraints and so no liquidity problems. So the recommended withdrawal ($3,663) would not adjust for potential future liquidity concerns. Spending here would be the same in both plans because total wealth is the same ($1,144,324).

But the simulations reflect liquidity constraints. It assumes no borrowing, and so withdrawals would drop to zero if the portfolio balance becomes zero. So liquidity issues will show up in the spending graph that displays the simulation results. For example, if we spending too much on the basis of Social Security that won't start early enough, we may risk running out of money before Social Security starts. That will show up in the spending graph, and you would have to adjust the plan to remove the liquidity risk. The model will not automatically do this. The Learn section of the planner has more about this under Future Savings and Retirement Income -> Liquidity:

• But there is a liquidity risk if you have a delayed Social Security or pension. Seeing that you have a lot of bonds in the form of pensions, the "total portfolio" method may place a large fraction of your savings portfolio in stocks to reach the desired stock exposure. If stocks do badly before pensions start, the savings portfolio can run out of money and you will temporarily be without funds till the pension begins. Withdrawals may also be too generous in anticipation of future income, causing you to come up short before the pension begins.
  
  Liquidity is most likely to be a problem when the savings portfolio is small relative to the pension, there is a long gap before the pension starts, and you have chosen a riskier asset allocation.
  
  If there is a liquidity problem, the simulation will show you running out of money before your pension starts. If this is happening, you can resolve it by adding a constant essential expense (e.g. $1,000 per month) after the start of the delayed pension.

Similar to the break down of the current consumption for the unconstrained plan, how can I think about the break down of current consumption, as well as the evolution of consumption over time, for the constrained plan?

In the example you gave, the liquidity constraint is not limiting withdrawals even in the 5th percentile. What is being limiting is the asset allocation. It's forcing the retiree to hold too much bonds and not enough stocks relative to what the unconstrained model would recommend. So in the constrained version, spending outcomes become safer (lower dispersion but also lower average spending) compared to the unconstrained model.

In the constrained setup, in early retirement the retiree is saving a little bit of the income in early retirement. They invest all of it in stocks, but it's still not enough to reach the target asset allocation of 26/74 on the total wealth. For example, in the unconstrained model, at age 70, at the 50th percentile, the retiree would have $1,080,549 in wealth in the savings portfolio and would allocate 26% of that—$280,943—in stocks. Whereas in the constrained model, the retiree would have only $19,060 in the savings portfolio. Even if they place all of it in stocks, it would be much less than the target asset allocaton of 26% of total wealth. So they are forced to be too conservative relative to their preferences, and their wealth and spending diverges from the optimal unconstrained path.

[GoWithTheCashFlow]

Thanks, Ben. That was very clear.

[Sho]

Thank you, Ben, and to the forum member who recommended this.

Sorry, I haven't gone through all the messages, but can someone please answer this.

1: If I put in a ceiling of expenses of 20 K, is it future dollars or current dollars?
2: If I want a legacy of current 2 Million current dollars, should I put 2 million dollars?

When it asks for SS, I put in real dollars, as I am assume COLA would cover it
Similarly, I use real dollars for my 403/457, as the contribution goes up with inflation.

Do I understand this correctly? Thank you

[Ben Mathew]

Thank you, Ben, and to the forum member who recommended this.

Glad you're finding it helpful.

1: If I put in a ceiling of expenses of 20 K, is it future dollars or current dollars?

The ceiling is in current dollars. All dollars—both input and output—are current (real) dollars except income during retirement or extra spending inputs that you have explicitly stated is nominal (by selecting "This amount is not adjusted for inflation (nominal dollars).")

2: If I want a legacy of current 2 Million current dollars, should I put 2 million dollars?

When it asks for SS, I put in real dollars, as I am assume COLA would cover it
Similarly, I use real dollars for my 403/457, as the contribution goes up with inflation.

Do I understand this correctly? Thank you

Yes. All of the above are correct.

[Sho]

Thank you, Ben, and to the forum member who recommended this.

Glad you're finding it helpful.

1: If I put in a ceiling of expenses of 20 K, is it future dollars or current dollars?

The ceiling is in current dollars. All dollars—both input and output—are current (real) dollars except income during retirement or extra spending inputs that you have explicitly stated is nominal (by selecting "This amount is not adjusted for inflation (nominal dollars).")

2: If I want a legacy of current 2 Million current dollars, should I put 2 million dollars?

When it asks for SS, I put in real dollars, as I am assume COLA would cover it
Similarly, I use real dollars for my 403/457, as the contribution goes up with inflation.

Do I understand this correctly? Thank you

Yes. All of the above are correct.

Thank you !

[Solicitorious]

Hi

just something I'm trying to get my head around, and maybe Ben or someone could clarify?

The Risk Tolerance, as I understand it, is meant to represent your risk tolerance at age 20.

Does it follow, therefore, that I have to ask myself the question "What was my risk tolerance at age 20?"

With Decrease Risk Tolerance with Age set at a positive value, presumably the risk tolerance at age 20 is reduced linearly? from age 20 until final age? But I therefore don't know what my risk tolerance today actually is?

If I set this to 0, presumably my tolerance remains the same as it "was" at age 20 throughout life?

I can't quite understand why this all has to be "anchored" to age 20 instead of being "anchored" to current age today.

With a positive decreasing risk tolerance does that imply I have an increasing CRRA, which is distinct from IRRA?

[Ben Mathew]

The Risk Tolerance, as I understand it, is meant to represent your risk tolerance at age 20.

Does it follow, therefore, that I have to ask myself the question "What was my risk tolerance at age 20?"

With Decrease Risk Tolerance with Age set at a positive value, presumably the risk tolerance at age 20 is reduced linearly? from age 20 until final age? But I therefore don't know what my risk tolerance today actually is?

If I set this to 0, presumably my tolerance remains the same as it "was" at age 20 throughout life?

Yes, risk tolerance is assumed to be linearly decreasing from y1 at age 20 to y2 at max age, where

• y1 is the risk tolerance input
• y2 is y1 minus the input in "Decrease Risk Tolerance With Age."

So if "Decrease Risk Tolerance With Age" is set to zero, the risk tolerance schedule becomes flat at y1 (the risk tolerance input).

You can see your current risk tolerance in the "Risk Tolerance Table" located under Advanced -> Decrease Risk Tolerance with Age.

I can't quite understand why this all has to be "anchored" to age 20 instead of being "anchored" to current age today.

To obtain the schedule from the user as an input, we could ask for risk tolerance at any two ages and draw a line going through those two points. Age 20 and max age are just arbitrary choices we made at two ends of the spectrum. It's easier for us to use a fixed age like age 20 than the current age because current age will change over time and we would have to auto-update the risk tolerance as we move forward through time. But we are planning to revamp the Risk section and move the inputs under Risk -> Advanced into the main Risk Tolerance and Spending Tilt sections. As part of that change, the risk tolerance glidepath will become clearer and will be specified from current age to max age, just like the asset allocation glidepaths in SPAW and SWR are currently specified from now to max age. These changes will also address some of the issues discussed here:

I believe that Time Preference is an input and Spending Tilt (g) is an output that depends on Time Preference. And, Time Preference is intended to be more stable and reflect the user’s nature while Spending Tilt is meant to tweak the calculated g. If so, then I’d suggest swapping the placement of these two settings so Spending Tilt is in the Advanced section and Time Preference is grouped with the main settings.

Yes, within the framework of the basic lifecycle model with CRRA preferences, Time Preference should be in the main section with Risk Tolerance. The extra spending tilt is not necessary and can be placed in Advanced.

But if we go beyond CRRA utility, the extra spending tilt would be an important lever to adjust as well. In particular, if you set risk tolerance is zero (making asset allocation 100% bonds), adjusting the time preference won't have any impact on the spending tilt. This makes sense for CRRA utility. That's because for CRRA utility, the desire to smooth spending across states (low risk tolerance) also implies a desire to smooth spending across time (low spending tilt). So infinite risk aversion implies zero spending tilt. To push past that and allow for strong spending tilts even at low risk tolerance, we need the extra spending tilt.

I was worried about confusing users with two sliders next to each other that control the spending tilt in different ways. So I put the Time Preference slider under Advanced with an arguably neutral default assumption of 0%. I agree that this is not ideal. So, as you suggested, I will try moving Time Preference out of Advanced and into the main section. I'll still need to keep the extra spending tilt slider in the main section because that is important as well. Let's see how that goes!

The Decrease Risk Tolerance with Age slider causes the Risk Tolerance setting to move back to age 20 so the user’s current risk tolerance is something different than the Risk Tolerance setting. I think this is surprising to the user (if they even notice it) and I can’t think of anything in TPAW that depends on looking backwards in time, so I’d suggest that the slider only impacts the Risk Tolerance value from the current age (starting with the user’s setting) and increases it out to the max age.

The reason for this choice was to keep the risk aversion slider stable across time. If the Risk Tolerance slider means "Risk Tolerance at current age" instead of "Risk Tolerance at age 20" then the slider will have to move automatically as you age if you've chosen to decrease risk tolerance with age. i.e. You might enter a Risk Tolerance of 15 today, but at some point in the future the slider would have moved down to 14 without you manually adjusting it. To avoid that, we chose to fix the meaning of the slider as "Risk Tolerance at age 20."

But as you said, this is surprising to the user. Ideally we'd communicate the full graph of risk aversion as a function of age so users can see what the slider is controlling. We'll have to figure out a solution to this when we allow risk aversion to be a function of wealth. I'll defer this issue till we get to that feature and can solve both problems together.

With a positive decreasing risk tolerance does that imply I have an increasing CRRA, which is distinct from IRRA?

Yes. Increasing relative risk aversion (IRRA) refers to risk aversion that is increasing as a function of wealth/spending. Here risk aversion is increasing as a function of age, not as a function of wealth/spending.

[bynomial]

Yes, risk tolerance is assumed to be linearly decreasing from y1 at age 20 to y2 at max age, where
y1 is the risk tolerance input
y2 is y1 minus the input in "Decrease Risk Tolerance With Age."

I somehow missed this fact. I just used the slider until I was 'comfortable' with spending. I was actually ok with a range of risks, but settled on 16 (aggressive). I was kinda bothered by the label, so I took some quizzes, thought about it - eventually just accepted it. (I did struggle a bit to refrain from dialing in my 'right' asset allocation)

'Aggressive' at age 20 seems more 'palatable'

Thanks to Ben and Jacob for this great tool.

[Ben Mathew]

I just used the slider until I was 'comfortable' with spending. I was actually ok with a range of risks, but settled on 16 (aggressive). I was kinda bothered by the label, so I took some quizzes, thought about it - eventually just accepted it.

Keep in mind also that the conservative-moderate-aggressive labels apply only to the risk portfolio and does not take into essential extra expenses that you may have scheduled which would be funded by 100% bonds. So "aggressive" on the risk portfolio might be moderate or conservative after you account for essential extra expenses. We'll communicate this better in the planner when we redo the Risk section.