I am just another user but can offer my (possibly incorrect) explanation of the three spending scenarios.

First understand that each asset type that makes up your portfolio has both a specified real return rate and a relative risk. Most portfolios you set up will have multiple asset types. Let's assume you have an asset with an assumed real return rate of 10% and a relative risk of 50%. (I don't recall exactly but I have read here on the forum or in the help file that the relative risk is the asset's published beta squared.)

Aggressive Spending will run the Monte Carlo simulation assuming a real return of 10% adjusted for inflation and vary it randomly +/- 50% during the many simulations.

Cautious Spending will run the Monte Carlo simulation assuming a real return of ~5% adjusted for inflation and vary it randomly +/- 50% during the many simulations.

Conservative Spending will run the Monte Carlo simulation assuming a real return of 0% adjusted for inflation and vary it randomly +/- 50% during the many simulations.

If you have assumed an inflation rate of 3%, then these three cases will simulate varying your asset's return rate around roughly 13%, 8%, and 3%. (Exact values will be given by (1 + nominal interest rate) / (1 + inflation) = (1 + real interest rate).

Dick or Lowell, please correct my understanding. I'm pretty sure I haven't gotten it exactly correct but hopefully this is close.

John

Hi, There is a confusion here between spending behavior and investment returns. We're drawing from the true distribution with the true mean of the portfolio being held in the year in question. But when you specify the spending behavior, you are telling the program how risk averse or not you want to be with respect to spending at any point in time out of your assets and other resources knowing that in the future you could draw lots of bad returns. So the spending behavior is purely about spending. Someone whose sole resources are assets invested in stocks would surely want to spend more cautiously than if he had all his assets invested in TIPS. Call if this is unclear -- 617 834-2148. best, Larry

Hi, To see the difference in actual discretionary spending based on the three spending options, just run the Monte Carlo three times with each option chosen and compare the recommended discretionary spending in the annual suggestions for the current year. You'll see that they are different.

Your spending behavior is not about your beliefs about returns on your portfolios, but about the extent to which you are risk averse. If you are risk averse, you'll want to specify cautious or conservative spending behavior and the current year discretionary spending recommendations will be more cautious or conservative.

Larry,

I'm afraid I'm confused now. (I haven't checked in here for a few weeks.)

I have done exactly as you suggest multiple times in the past. I just did it again with my current profile.

I can hand calculate an average real return of the investments in my retirement asset portfolio. It is 7.51% real rate.

When I run with Spending Behavior set to Aggressive and export to Excel, I can easily see that my portfolio is growing at 7.445% real rate. That's close enough for me.

Now, I run with Spending Behavior set to Cautious and export to Excel. I see the portfolio growing at 3.723% real. This is exactly half the "aggressive behavior rate."

Finally, I run with Spending Behavior set to Conservative and export to Excel. The reports show no (0.0%) growth in the portfolio. This corresponds in my mind to results just equaling inflation.

So now my confusion:

Are your explanations of this being about spending behavior and not return rates a reflection of intent within ESPlanner or implementation within the program? I can wrap my mind about the reasoning for the differing spending behaviors and how risk averse I may be as well as it's affect on Discretionary Spending. However, it sure seems that the implementation is being done by adjusting the portfolio return rates in the calculations.

Can you speak to why we see portfolio growth rate changes when you speak about this not being about returns on our portfolios? Is this simply a means to an end to show the effect of spending behavior on discretionary spending? I'm OK with that but get confused when I see differing portfolio returns when you say there should be none.

John

I see the Monte Carlo spending behavior settings as serving two main functions. First, MC shows a range of potential outcomes for each spending behavior and gives us a feel for probabilities at each level along with many details. While "aggressive" is the most realistic setting, it assumes that future returns will, in some sense, follow historical trends for risk and return. This may be true, but the downside risk is highest for the aggressive setting.

Second, while every MC portfolio will be different, there will almost always be some probability that you will run out of assets before end of life and have reduced consumption. By setting the spending behavior to cautious or conservative, you are stress testing your porfolio to see what would happen if future returns are worse (or much worse) than you expect to occur.

In effect, cautious or conservative spending behavior applies a "safety factor" to MC. So ... if you are considering spending at a reduced level than "aggressive" (meaning using a safety factor), the cautious or conservative spending behavior shows a presumably higher probability of maintaining (or raising) your level of consumption.

Larry has written about spending cautiously if you invest aggessively because the downside risk of poor returns is higher if future returns are lower than expected. He also has written about the flip side where if you invest conservatively, then you can spend (somewhat) more aggresively since you have a built in safety factor.

Best regards,
Brian Vezza

John and Pleonasm,

I hope this isn't further muddying the waters given that it's already a bit confusing. We're probably saying very close to the same thing with some different wording.

To Pleonasm's specific question, I agree with you that the same assumption about "future returns" is made for all three MC spending options. Where we may differ slightly is that even though the risk of the portfolio is identical, the risk of your overall ESPlanner profile is higher if you use "aggressive" than if you use "cautious" or "conservative" spending behavior. My guess is that we don't even differ here. If you view your overall risk from the profile level, instead of at the MC portfolio level, your aversion to risk (demonstrated through your spending behavior / safety factor) IMHO leads to Larry's statement "Your spending behavior is not about your beliefs about returns on your portfolios, but about the extent to which you are risk averse." Clearly this is just my view of things and clarification is definitely appreciated.

John - I think there are two main places to look at MC rates of return being "managed" per your comments in #7 above. First, in the main report under annual consumption. As you describe, under cautious spending behavior, this consumption figure appears to be derived from a real rate of return half way between the mean return and zero. Second, in the "Monte Carlo Report". In the MCR, for example, take a look at your regular assets under cautious spending behavior. Under the Variability of Reg Assets tab, I see much higher Regular assets than in the main (not MC) report. I think this is because ESPlanner is using the full MC rate of return to calculate asset growth while spending (consumption) is at the reduced rate shown in the main report. This causes assets to grow and increases your probability of a higher living standard / reduces your downside risk. In my mind this is ESPlanner's way of implementing a "safety factor" for MC scenarios. I think I get the how's and why's of this, but am not 100% certain so my apologies if I have this wrong somehow.

I've been thinking a lot about "safety factors", how to implement this (first through Excel with manual calculations, then through ESPP), and how to address Pleonasm's items above. The "cautious" MC approach does a decent job for me, but you might want to consider an approach I described at http://www.esplanner.com/forum/new-version-planning (about half way down) using adjustments in living standard index. After playing around with it, I am ~90% happy with how it worked out and have also run it in combination with the "cautious" MC approach. I've customized this Safety Factor approach for my profile and think you might find it helpful.

Best regards,
Brian Vezza

Brian,

Your comments:
"Second, while every MC portfolio will be different, there will almost always be some probability that you will run out of assets before end of life and have reduced consumption. By setting the spending behavior to cautious or conservative, you are stress testing your porfolio to see what would happen if future returns are worse (or much worse) than you expect to occur."

This I understand and is the reason I base my spending plans on the "Cautious" spending levels output by ESPP. In the big picture, this gives me more confidence that my plan can remain valid under historically "normal" conditions. Thus I let my spending behavior be influenced by the recommendations of ESPP under a cautious spending plan. That is the intent of the "Spending Behavior" setting as I understand it. My confusion has arisen not from the intent of the plan but rather the implementation. Clearly rates of return appear to be managed to achieve the end even though it is said that the rates are not changed by changing the spending behavior setting -- or at least I think that is what Larry and Dick have said.

Ideally, one could base a plan on the "Aggressive" spending level and on average expect it to work out. However, as Larry has pointed out here and elsewhere a string of down-market years near the beginning of the plan can leave you short of your plan's goals even if the average over all those succeeding years is above average. The problem is that a few bad years at the beginning of your withdrawal period can reduce the portfolio by an amount that cannot be made up with a string of above average years later on even if the average over the whole period is indeed "average."

So, I sort of think I get it with regard to Spending Behavior. The intent is to influence a spending behavior that gives your plan the best chance for success. The implementation appears to work by adjusting return rates to ensure that your plan does in fact succeed and get you to your goal. The comments here get me confused when I cannot separate the discussions of intent from those of implementation.

John

I’ve avoided entering this conversation so far, but maybe another perspective will help. If not, please ignore what I say.

One key thing to recognize is that traditional financial planning will simply spend a certain amount, and have some left over at the end, or else not make it to the end. ESP doesn’t work this way. Your std of living will vary for the rest of your life, depending on market returns.

A second key thing to recognize is that you will NOT have the std of living that ESP recommends. Your std of living will be better, or worse, based on market rtns, but the probability that it will be identical, to the penny, are negligible. So your choice is really what do you want the probability to be that your std of living will be better than ESP recommends, and what to you want the probability to be that your std of living will be worse? Do you want a 50/50 split? 75/25? 90/10? 95/5? The more skewed the split you chose, the less you can spend now, and the more drastic the probable upside, which may not be the best choice. Furthermore, your investment choices affect both how much your spending must be reduced to achieve the split you want, as well as how much upside and downside potential there are on either side of the split.

Setting your spending behavior to Aggressive will simulate spending as if you expect to get mean returns. The Projected Trajectory in MC (which is also what will be used for the Main Report) will match this rtn assumption. This will give you a 50/50 probability of having a greater/less std of living than ESP is planning for you. Check out the Monte Carlo rpts to see this, as well as to see the potential upside/downside for your particular asset choices.

Setting your spending behavior to Cautious will simulate spending as if you expect to get one half of the mean returns. The Projected Trajectory in MC (which is also what will be used for the Main Report) will match this rtn assumption. This will give you a greater than 50% probability (check out the Monte Carlo rpts) of having a higher std of living than the Projected Trajectory.

Setting your spending behavior to Conservative will simulate spending as if you expect to get no returns. The Projected Trajectory in MC (which is also what will be used for the Main Report) will match this rtn assumption. This will give you a greater than 50% probability (check out the Monte Carlo rpts) of having a higher std of living than the Projected Trajectory.

I have found two different profiles that give me almost identical recommended consumption and a 5% probability of a 10% decline in my std of living.

One profile uses Aggressive spending with a reasonably conservative portfolio (in my case, starting with about 30/70 stocks/bonds split, moving to more bonds as I age).

The other portfolio uses Cautious spending with a very aggressive portfolio for my age (starting at 90/10, but ramping down quickly).

Before the Spending Behavior functionality was introduced into ESP, I tried to put my own spending “safety margin” in by raising my std of living by a certain rate (e.g., 2-3% per year). My rationale was that if I spent less than ESP recommended, but then actually got the mean returns that ESP was using in its calculations, then every year I would have a little more left in the bank than ESP expected (since I had spent less than it expected), and would be able to raise my std of living a little bit (because my assets were greater than ESP had planned for). This trick forced ESP to lower my recommended consumption, and twisted the Monte Carlo “cones” upward. The percentage I chose to raise my std of living was based on trying to get the 5% MC line to have a reasonbly acceptible slope - a 5% risk of a 10% reduction in std of living.

This was EXTREMELY tedious, and required a lot of iterative runs. I was a very happy guy when the spending behavior was introduced.

Pleonasm,

I'll try to answer #1.

You need to differentiate between the "main" report and the MC reports. If you select conservative spending, the main report will show (e.g. in the annual consumption figures) consumption with a 0% real return.

However, in the MC reports, your assets grow at your full real portfolio return %. At the same time, in the MC reports, your consumption level starts as if you are only getting a 0% real return. The difference causes your assets to rise more under this scenario. My impression is that the MC code increases your "Living Standard" (consumption) over time in response to these rising assets and ESPlanner attempts to spend down the increased assets to end up with $0 at maximum age of life (unless you have a bequest).

So, the asset allocation is still in effect to calculate the growth of your assets as they grow at the full real return % and you spend at a reduced rate.

For #2, I've never tried this, but it would be interesting to test a portfolio with 100% cash (which shows a negative real return) and see what happens. My guess is that you are right. FYI - I just tried this and it worked like you thought it would.

Best regards,
Brian Vezza

I'm just an ESP user, but I'll still take a stab at your questions.

If you've selected "conservative spending", then the asset allocation is irrelevant w.r.t. the Projected Trajectory (which is also used for the Main Report) - but it is still very relevant w.r.t. Monte Carlo. As Monte Carlo generates the 500 trajectories, it will use your spending assumptions to determine your consumption, but will use the asset allocations to determine the "random" investment returns. So your spending behavior as well as your asset allocation will both affect the percintile distributions and probabilities that are shown in the MonteCarlo_report. The whole point of the Monte Carlo is to show you the realistic alternatives to what you are planning for. So if you're probably not getting what you're planning for (i.e., if you're planning for NO returns, but you probably get some return), your percentile distributions and probabilities will reflect that.

I think you are correct that if you had a portfolio that was 100% cash, and chose "conservative spending", you would end up with a Projected Trajectory that was higher than if you chose "aggressive spending". It would be an interesting experiment.

Hi All,

When you run the Monte Carlo with aggressive spending, the projected trajectory shows you what you would spend each year if you were to always receive the mean real returns on the portfolio you are holding in each year.

When you run the Monte Carlo with cautious spending, the projected trajectory shows you what you would spend each year if you were to always receive half the mean real return on the portfolio you are holding in each year.

When you run the Monte Carlo with conservative spending, the projected trajectory shows you what you would be able to spend each year if you were to always receive a zero mean real return on the portfolio you are holding in each year.

Regardless of the spending behavior you specify, when we run the Monte Carol we are always drawing from the true distribution (with the actual historical mean real return). But if you specify cautious spending rather than aggressive spending, you are telling the program to spend less in any given circumstance than if you specify aggressive spending. And if you specify conservative spending, you are telling the program to spend even less than in the cautious spending case in any given circumstance.

The less you spend at any given date, the more assets you'll have to invest and the more resources out of which you can spend in the future, for any given draws of returns you might experience. This lowers the initial level of spending and tilts the living standard percentile distribution curves upward. The intuition is that by spending less in the present, you'll be able to spend more in the future for any given set of returns you earn because you'll have more assets out of which to spend due to the fact that you haven't eaten them up today.

best, Larry

PS, if there is still any confusion, my cell is 617 834-2148.

Pleonasm,

I'm just a user, but IMHO when I read the most recent comments from Larry (the authoritative source), Jfshelton, and myself I see the comments basically saying the same thing. MC is confusing and while I'm still not 100% sure I've got it right, I think this is accurate.

#1. As Larry has described, all ESPlanner MC portfolio attributes matter.

#2. When you run MC reports your asset's (regular, spousal, yours) rate of return varies according the the MC portfolio specified in #1. This is completely independent of your setting for spending behavior.

#3. Your spending behavior determines the level of safety factor MC uses. If "aggressive", there is no safety factor for your consumption. If "cautious", your consumption is midway between mean and zero real return. If "conservative", your consumption is at the level of zero real return.

#4. Your specific ESPlanner profile determines how your specific MC results are shown in the reports per ESPlanner programming. I can't comment on your specific results and reports because I don't know anything about them. However, in any profile, ESPlanner's consumption smoothing will try to find the solution to your specific profile.

#5. The "Main" report (not MC), "Annual" tab, "Consumption" column shows your consumption level as if you manually calculated your MC portfolio's real rate of return and then, combined with your spending behavior setting, input the full mean real return, 0% real return, or the midpoint of those two options (converted to nominal return). You can experiment with this by manually changing your rates of return with MC turned off to see their impact.

#6. ESPlanner's algorithm, by default, assumes your assets get spent by maximum age of life (e.g. except for home, funeral, bequest, special expenditures, etc.). So, by default, it calculates consumption without a safety factor. However, in MC when you mandate a safety factor through setting spending behavior to cautious or conservative you get an interesting situation. The following is what I think happens.

As ESPlanner performs its immense calculations on a year by year basis, we know (per Larry) that it spends down assets at your spending behavior rate while growing assets at your MC rate (with varying returns per #1). Over time (assuming your assets grow larger over time), ESPlanner makes adjustments in your consumption. It still spends at a reduced rate to ensure a safety factor, but it also, by default, is set up to spend down all assets by end of life. Larry writes "The less you spend at any given date, the more assets you'll have to invest and the more resources out of which you can spend in the future, for any given draws of returns you might experience. This lowers the initial level of spending and tilts the living standard percentile distribution curves upward."

So ... setting the spending behavior to cautious or conservative lowers your consumption (per description), increasing assets over time lead to increasing consumption (although still with a safety factor), and near end of life ESPlanner ramps up consumption over the last decade or so of life to ensure (per your specific profile and programming) that you have plenty of consumption while you end up spending down your assets as described earlier.

While ESPlanner MC doesn't show assets remaining at end of life (per exceptions above), you can always choose not to consume at the level shown and can ballpark estimate how much in assets you would have left.

I hope this is helpful.

Best regards,
Brian

Pleonasm,

I think I see the issue here. Your questions appear to be mostly directed around the "Main" report. However, this report shows very little MC related information. The Monte Carlo reports are the place to see 99% of MC behavior that we are talking about here

For your #1 "How do you know...?" Most importantly because Larry has described this. Secondly, and here is where I think your confusion is, you are looking in the "Main" report when viewing “Regular Assets Income” in the “Total Income” tab. The "Main" report (as far as MC is concerned) really only takes two inputs (I think) from your MC portfolio / inputs. With MC turned on, the program calculates your portfolio's mean real rate of return (there can be more than one depending on how you implement your portfolios). Then, based on your spending behavior setting, the program effectively inputs these data points into the "Nominal rates of return" fields over-riding any manual settings for those fields when MC is turned on. The "Main" report shows the level of consumption as if you had manually done these steps and input them into the "Nominal rates of return" fields yourself.

For #2, you need to look at the MC reports, not the "Main" report to see these results. Due to the complexity of the calculations, the results are non-linear, but if you run through each spending behavior and look at the MC reports, you will see what you are looking for. Try saving the MC report for each spending behavior setting and look at them side by side and that might help.

Good luck!
Brian

Pleonasm,

Sorry if #1 came across negative. I certainly didn't mean it as such. Interpreting the MC results isn't easy and I'm still trying to figure out what lessons to take from it even after reading everything I could find about ESPlanner's implementation. I'll give my interpretation in the next few paragraphs, but first it is possible that if you are using UDAs or user defined assets that the results won't make sense (see a recent post on this) and your inputs will need to be adjusted. My comments below relate to the core MC assets that have been in the tool for a few years and are not user defined.

I'm looking at my "Monte Carlo Report" (MCR), not the "Trajectories" report. Both have value, but I'm focused on MCR. Also, I'm looking at two tabs - "Variability of Income" (VOI) and "Variability of Reg Assets" (VORA). Given your specific situation is different from mine combined with the variability of MC, I'll point out a few key trends.

#1 For the first ~5 years or so, before the safety factor can build up too many excess assets, I see VOI increase from conservative to cautious to aggressive (e.g. 5th, 25th, 50th percentiles) as expected. However, this reverses when I look at the 75th and 95th percentiles with the conservative being highest for income and aggressive lowest for income.

#2 When I look further out (~15-25 years), #1 trend holds for the 5th percentile, but as I move out to the 50% percentile, I see something interesting. My income is actually higher for conservative, in the middle for cautious, and lowest for aggressive. This also holds true (in my implementation) for the 75th and 95th percentiles.

Here is what I think is happening. First, the 5th percentile for VOI, by definition, means that there were quite a few poor returns. Since the returns were so low we don't see much of an impact between the three spending behaviors. However, as we move towards the 50th percentile and beyond, the returns are average or better and the impact of spending less early on gets magnified due to the higher returns. Because the safety factor endures over time, the impact of these average to high returns on higher levels of assets becomes a significant factor.

#3 Now looking at VORA tab, the important item to note is that ESPlanner tries to have a constant draw of tax-sheltered (TS) assets in retirement. It adjusts TS withdrawals to meet required minimum withdrawals, but on the whole tries to keep these at a constant level. As such, your "regular" (taxable) assets may vary significantly year by year. Regular assets are withdrawn in different amounts each year to smooth overall consumption.

#4 After the wordy preamble in #3 (sorry if this is all old hat to you), what I see is that for each percentile (5th to 95th), I have higher regular assets for conservative spending and lowest for aggressive spending. I interpret this to mean that for pretty much any set of portfolio returns (poor to great), by under-spending due to the MC safety factor for conservative and cautious spending behavior, we see the impact of applying the MC returns to the safety factor surplus over time. Now, we don't get to see the under the hood calculations or reports for each MC "lifetime" so I can't prove this by pointing to a specific calculation or result, but the results seem reasonable.

Finally for your #2, I think I've addressed this above. "Income" can come from a number of sources and as you have higher or lower asset values (due to MC and spending behavior), it's reasonable to expect different income levels.

Stick with it and keep experimenting. I hope this is helpful.

Best regards,
Brian Vezza

I'm saying that the rate of return used for the Projected Trajectory is your Spending Assumption, which is to say that the Projected Trajectory "projects" what your std of living will be if your assumption is actually correct.

The other 500 trajectories are generated from the historical data (including mean rate of return), and taking into account your spending assumption. So the MC simulates that each year you will spend based on your assumed rate of return, but will, in fact, get some other (historical) rate of return. That last statement is true, by the way, regardless of which of the three spending assumptions you chose.

So, since "Conservative Spending" assumes 0 rate of return regardless of your assets, asset allocation doesn't affect the Projected Trajectory.

I believe that the Main Report will be identical if you run MC with "Conservative Spending" and if you run non-MC specifying a rtn of zero. But the MC reports will show you the potential varience using this conservative spending approach.

Are the annual Monte Carlo results appropriate when considering actual long term stock market performance? In the past four decades, we have had two entire decades (70s and 2000s) with about nil overall gains in the market, and two decades that were fantastic. We did not have 40 years of market behavior where each year seemed to behave as a random event. Is it fair to model the market based on annual averages, when in fact the stock market is influenced by longer term economic downturns and upturns? I realize that some of the Monte Carlo runs will mimic ten years of mediocre to poor (or exceptional) market behavior, but these runs will be treated as outliers rather than the common occurence that they are. I really don't want to go broke while waiting for the law of averages to take care of me, which I think everyone understands. But I also wonder about putting too much confidence in a model that puts decades like the 1970s or 2000s in the bottom 5% or bottom 25% category.

Which brings me to two questions: 1) Have you run the model against long strings of poor performance rather than annual random performance, and if so, has the Conservative model held up well? Is it a good hedge against the case I've described? 2) Am I totally off base. It wouldn’t be the first time.

(1) No, not specifically.
(2) I'll let Larry chime in on this since I'm not in a position to judge.

However, there is another thing coming, Upside Investing, which allows you to see what happens if you loose all assets other than ones you invest totally safety and also the "upside" in case the assets aren't completely lost.

We have this on the ESPlannerBasic site and will be including it in the upcoming release.

This is a notch more conservative than even Conservative spending.

Best,

Dick Munroe

As a side comment to this topic, I suggest renaming the "Aggressive" spending behavior as "Confident". "Aggressive" has negative connotations where "Confident" does not. The latter choice accurately describes the attitude of the user toward the likelihood of achieving the returns forecasted by the system.