Consumption Smoothing Procedure
Thanks for the webinar yesterday – hope to see more of these in the future.
The Social Security example that was discussed is the kind of scenario I frequently have: someone at or near retirement needing to juggle IRA balances, Social Security timing and usually some other issues.
I’d like to review my basic methodology for handling these situations and get comments.
Typically, delaying Social Security requires (if consumption smoothing is desired) requires that I withdraw from my IRAs or related accounts more aggressively in the early years – say ages 62-69. Then the withdrawal rate can be reduced starting at age 70.
But ES planner does IRA withdrawals at a fixed rate. My goal is to figure out how much I need to take out each year from IRAs to achieve the smoothed standard of living. In order to implement with ES Planner, I follow this procedure:
**1. Take guess at how much IRA money will have to be withdrawn in each of the years 62-69. Call these amounts X62,…X69.
**2. I need to reduce my IRA balance by these funds, and show them as taxable special receipts. However, the IRA balances are growing at an assumed rate (say, 3% real). Therefore I can’t simply reduce the IRA account by the sum of X62 to X69. Instead I need to do a side calculation on a spreadsheet, discounting each of the cash flows by 3%, yielding Y62,…Y69.
**3. Enter X62,…,X69 as taxable special receipts. Deduct the sum of Y62,…,Y69 from my original IRA balance.
**4. Run ES Planner
**5. The results may indicate that my original guesses were off one way or the other (resulting, for example, in recommended savings. I shouldn’t be withdrawing from IRAs to add to savings, so this would mean I guessed too high and need to adjust one or more of the X’s downward). So I fine tune the Xs & Ys and repeat the process.
This type of thinking applies to lots of scenarios --- not just the social security timing issue. Retirees may have other significant expenses (health insurance during pre medicare years) or receipts (special payouts from former employers) that take place after retiring.
I think this procedure works OK (although it does render any Monte Carlo sims less meaningful) but does anyone have a better process? It can be somewhat cumbersome.
RSS
Hi David:
Yes, we'll do more webinars--so let us know what you'd like to see in that regard.
Here's my thoughts on your question:
The trick is to somehow create some "regular assets" so that ESPlanner can smooth consumption. If they have regular assets, the program will use those to smooth things. If they are a few years away from retirement, ESPlanner will try to get them on a smooth saving plan in anticipation of the need.
Rather than trying to figure it out per year, could you just lower the amount you have entered in their retirement accounts and add that amount to regular assets (say in a money market fund) (representing a one-time withdraw) and create an expenditure for the taxes? This would then show you the dissaving pattern in the Saving column.
Perhaps I haven't thought that all the way through, but does that give you any other ideas?
Dan
Dan, I would have loved to join the webinar but was never notified. How do I find out about the next one? Was the webinar archived for later viewing?
Thanks,
James Mavrogenis
Interesting idea. It does solve the smoothing issue. However now I'd have to calculate the taxes each year -- and without knowing what the "dissaving" would be until after I run ES. So off hand, it seems like it would be a more difficult approach.
Might try experimenting with it, though.
thanks
I'm going to have Kotlikoff weigh in on this--he's out of town at the moment--but if you actually do take out a chunk from the retirement account and put it in a saving account of some sort--are you really any worse off from the client's point of view, i.e., having moved a chunk of money out of retirement fund and into a money fund? Then the taxes would be accurate if you get the first year right. I'm not a planner, but does it matter much where that money resides? (retirement fund vs money fund) And perhaps they would even pay less taxes since there's no SS coming in?
And too, you can push up the date of last withdrawal on the retirement funds--but I don't know how folks feel about that. But it does compress the money and make the withdrawals bigger--but perhaps not big enough.
In my puzzle solving mode, I just think: how can we get them some regular assets?
Or, perhaps . . . this . . . take a chunk of the money (trial and error to get the right amount?) and put it in another account in the Retirement Accounts panel--one of the options that has the same tax law. Then, put the latest withdrawal on that one just four years out, so that it takes that money out to smooth consumption. This way the taxes would be correct.
Or, can you tell ESPlanner that one spouse has a large chunk and the other spouse has a much smaller chunk. Then, with the smaller chunk, put the last date of withdrawal just four years out or whatever.
Would this work? Larry?
Dan
Well the infrastructure for doing arbitrary withdrawals from retirement accounts is in place in the CE. The problem we're having is figuring out a rational user interface to represent such things (it pretty rapidly gets hard to deal with if you really go after a year by year thing). Anybody having any thoughts on how such an interface might look (consistent with the extant UI for retirement accounts), feel free to chime in. I don't think we'll make the May update cycle with this but miracles do happen from time to time.
Best,
Dick Munroe
Dan & Dick:
Your solutions are all thought provoking. The idea of using alternative Retirement Accounts makes sense. But I think it is about as much work as my original solution of figuring out the annual Taxable Special Receipts.
The essence of the problem is that ES tells us how much of the regular assets to use each year based on consumption smoothing, but we have to tell ES the rate at which to take out IRA money. In my practice I am usually faced with large amounts of Retirement assets, and small amounts of regular assets. I need to use the IRA money to achieve a smoothed standard of living.
Ideally I'd like ES to tell me the consumption smoothing solution for both asset types combined (i.e, ES also tells me how much to take out of the Retirement Account each year). Or, if I had to pick one, I'd have more interest in the optimal Retirement Account withdrawal pattern.
Steve Waas
Right. Yes, this is causing us to think of ways to make that possible in the user interface.
In the model I'm using, I seem to get the right solution by just "figuratively" moving assets out of retirement accounts into regular savings (even if it's not actually done) and looking at how ESPlanner would smooth it if it had that arrangement to work with. But you say that is not accurate to the taxes.
So, instead of moving them directly to savings (money market etc.) they can be entered as a special receipt with the proper tax button clicked in the special receipt area, and then you have the taxes calculated correctly too.
But I guess the taxes would be on the lump sum, so you'd have to distribute that money across three years as a taxable special expenditure.
Hmm.
Dan
Dan,
We've traversed a lot of interesting ground in this thread and I think your most recent approach (removing funds from Retirement accounts and entering them as various yearly taxable special receipts) is exactly where we started with my original post (although I also indicated we need to discount the cash flows).
Thusfar, this seems to me to be the cleanest approach, even if a bit cumbersome.
Which brings me to my basic conclusion: ES has great optimization technology but doesn't let me apply it directly to Retirement assets (although I believe we achieve this indirectly with the approach outlined above).
Steve
Unfortunately there are some technical reasons why we don't use the retirement accounts other than an exogenous variable (once you start withdrawing, they are on autopilot).
In order to do smoothing using "arbitrary" amounts from retirement assets, we have to run the basic code using that draw, then run it again trying others, etc. Basically it ups the computational load quite a bit, introduces another bunch of state variables into the solution and, ultimately, slows everything down to the point where ESPlanner might not be a commercially viable product (people don't have infinite patience).
Essentially you have to try "all" combinations of withdrawal ages, amounts to withdraw "early", etc. and we feel that this is way too cumbersome.
Larry's made a proposal that I feel is a good intermediate solution where you have a choice to withdraw a given amount of money from your retirement accounts [smoothly and in the specified withdrawal order] over a set period of years and smoothly thereafter or just to withdraw smoothly. Personally I think that being able to adjust a draw on a year by year basis is handy for emergency planning (maybe this year's car wants to be purchased out of retirement account assets) but I also feel that interface gets a bit complex.
Anyway, we're converging on giving you something to work with but probably not something as flexible as the special expense/receipt interface.
Best,
Dick Munroe
Dick,
If I understand your "intermediate solution", I think it would be a major improvement.
My understanding of the proposal: the user would be able to specify a specific Retirement Account withdrawal pattern for the first few years, and then for the remaining balance withdraw at a fixed rate just as ES currently works. Even if you are suggesting less flexibility than that -- for example a 2 stage withdrawal from Retirement accounts (first stage= fixed withdrawal specified by user, second stage =fixed withdrawal for rest of life) -- I think it would still be an important advance. Total flexibility is not important.
While its true that the user would have to do some trial & error to come close to the overall smoothed standard of living he seeks, I don't see that as a big problem. The intermediate solution would eliminate difficulties with the Special Receipts approach: e.g., I wouldn't have to recalculate the starting balance of the Retirement accounts based on return assumptions and re-enter all the special receipts each time I do a trial. Also, since the assets remain in the Retirement accounts section of ES (invested in a portfolio), the Monte Carlo sims would be meaningful.
Steve
To clarify, the proposal is, more or less, as follows:
Choice 1: Draw smoothly from beginning of withdrawals to end of life.
Choice 2: Draw x%, e.g., 25%, of your retirement assets smoothly from beginning of withdrawal to a specified year, then the residue smoothly from then on until end of life.
Using choice 2, I have the effect of additional retirement accounts which I then draw on until they are gone, then shift to the retirement accounts which contain the residue.
I'm still more in favor of year by year control over the withdrawals, but I can live with this proposal as well since it has some value and its user interface is simple (the CE takes a substantial hit in complexity but I think I can minimize the computational impact with careful design).
Best,
Dick Munroe
Dick,
I agree. Year by year control would be nice, but Choice 2 (what I called a 2-stage withdrawal pattern from the Retirement accounts) would be very helpful.
Steve
Having a 2 stage withdrawal option would be a huge improvement! Year by year entry might be too much control for most users.
Why not consider a three stage option....one could use that functionality to handle an individual year or two if desired...a nice middle ground. Perhaps 3 stage vs 2 stage crosses the line in terms of UI complexity, but maybe not.
Peter
Well, if three, why not four, five, etc... and you quickly arrive at year by year control (which is not only simple to implement, but easy to understand). The syntactic sugar of the proposal is attractive due to its complete simplicity, so we'll probably stick with the proposal based on a two stage withdrawal initially.
Best,
Dick Munroe
I'm a fairly new user. Another option that seems reasonable from a tax perspective (not sure about the computational complexity) is to withdraw an amount exactly equal to the minimum distribution requirement each year from tax-sheltered assets.
The MDR is already calculated by ESPlanner. Any additional spending would come from taxable accounts. For some people, this would lower their lifetime taxes and raise their consumption.
I do like the options described above. Thanks for a great tool and I appreciate your efforts to make it better.
Best regards,
Brian
For what it's worth, Larry and I continue to argue about this so there's a chance it may make it into the next update cycle. No promises, but there's more hope...
Bet,
Dick Munroe
I would have to look at the code, but I believe that we already extract the MDRs from the taxable accounts and the rest of the money in the order that you've specified in the user interface.
And, having looked at the code, I was right, the MDRs are extracted from the taxable accounts and the residue extracted in the withdrawal order, so if you want to use up your Roth while obeying the MDRs, just put your Roth account first in the withdrawal order and the right thing should happen.
Best,
Dick Munroe