Jim Richmond, developer of the Flexible Retirement Planner, discusses his tool.

Hosted by the Chicago Bogleheads chapter. Recorded on March 24, 2021.

Chapter meetings are listed in the Bogleheads blog calendar. You can add this calendar to your Google account. Notifications are also sent from an email subscription list. See the blog for more information.

**Transcript**

**Bogleheads® Chapter Series – Flexible Retirement Planner**

Welcome to episode number four the Bogleheads Life Stages Podcast. Bogleheads are investors who follow John Bogle’’s investing philosophy for attaining financial independence. Today’s episode features Jim Richmond, designer of the Flexible Retirement Planner. This recording was made on March 24, 2021. Nothing in this video should be construed as personalized investment advice.

**Jim Richmond**: Well what about retirement. What about how much do I need to have saved? And I started looking around and I wanted something that had implemented variable withdrawal. I was looking for a calculator online that had some sort of decision rule. I had read some recent paper, I think Guyton, Jonathan Guyton had written about how you could get a slightly higher withdrawal rate with a good probability of success if you have some decision rule in the withdrawal process that adjusts the withdrawal based on how well the plan is going.

So if you have a bad sequence of returns you could maybe cut back spending a little bit and if you allow that in the model, you end up with a higher withdrawal rate supported for a given probability of success, or given success rate for the retirement plan.

So anyhow I couldn’t find that calculator. So I decided to put this one together. And I initially just did it on a whim, but I posted it to the web and did get some positive feedback about it. And after six months or a year I decided to turn it into a business, and have done it.

Okay. So right now you should be seeing, or within a few minutes, seconds, you’ll be seeing the main page for the Flexible Retirement Planner. This is what you get if you hit “new plan” up in the top left. It resets all the inputs to their defaults.

Let me see if I can–and the inputs of the plan should be relatively straightforward: current age, retirement age, life expectancy. For this demo, I’m going to start out, we’ll assume the retiree is 40 years old. They are the person using the planner and the retirement age 65, and the life expectancy 95. We’ll say 3% for inflation.

We’ll leave it at the default, these income tax and investment tax rates should include state tax and federal tax and any local income tax if it’s paid.

And then for portfolio value we’re going to say that there’s currently a $300,000, should I get that $300,000 tax deferred portfolio. By the way this “min IRA 401-k withdrawal age” trips people up sometimes. And actually the next release is going to move that into settings just because it gets in the way. Basically that’s the minimum penalty free IRA, 401-k withdrawal age. So most people never need to touch that. And that’s part of why I’m going to move it to settings, just so you don’t get tripped up on that.

And then I’m going to say that for this plan, this person, this 40 year old, is going to put $15,000 away every year in their tax deferred annual savings. And then the other inputs, we have investing styles, and I’ll just leave it at the default.

And these are sort of canned portfolios that have given this investment mix. I strongly suggest folks take a look at the Simba’s backtesting spreadsheet, that’s a great resource for figuring out your own portfolio.

What’s a good return and standard deviation to use, but you could use these canned ones at least to get started. By the way, I guess I should do the standard disclaimer. I’m not a financial planner and I’m not an investment expert so nothing I say should be interpreted as investing advice or financial planning advice.

I do know about the tool and I’m happy to answer questions about, of course, how to model things that you’d like to model in your financial plan, but I can’t provide any kind of help or advice on the financial planning side.

Next we come down to retirement income and in this case we’ll say that there’s going to be a $25,000 annual retirement income. Let’s say it’s a a couple and they’re both the same age and they’re going to start at age 67 earning that, withdrawing from that Social Security, or taking that Social Security income. Retirement spending is $50,000 and the spending policy is again, that decision rule that I talked about. And I’ll get into the details of this later. But we’ll leave it at the default for now.

And then we’ll hit run. And so what we see with this set of inputs,. we have a fairly robust plan.

On the left you see the graph, in which the purplish bars are the tax deferred portfolio, the blue bars are the taxable portfolio. And the reason that we’re seeing this changeover from tax deferred to taxable is that RMDs [Required Minimum Distributions] are kicking in automatically at age 72. And so that’s why we’re seeing the tax deferred portfolio decreasing.

On the right hand side we can see the probability of success is 99%. We run 10,000 passes through the retirement, so basically that’s going to draw 10,000 different unique random sequences of return. And based on those sequences of return that we draw, one for each year of the retirement plan, we’re going to have a varying result. And so in this case we ran 10,000 paths through the plan and we had probably between 50 and 100 failures to get a 99% probability of success.

And the average spending shortfall is 16%. And that’s really just the metric–I think it was Marshall who had some research that was basically wanting to try to quantify when the plans fail in a monte carlo simulation. How bad is it? Is there a way to tell plans that fail earlier versus plans that fail later?

And spending shortfall tries to do that So that if you had two different retirement plans with two different sets of inputs and they were both showing a 90% probability success, one still may be more robust than another if it turns out that the failures happen earlier in the plan for one compared to the other. And that’s what’s average spending shortfall says. Just looking only at the failures, how much of the spending that was intended to happen in that plan, got a chance to happen before the portfolio ran out of money. and so with this 16%, that means 85%, on average, even with the failed plans, they made it 85% of the way through the retirement before the portfolio ran out of money.

And again, we’re only talking about 50 or so failures here, given the 99% probability of success. But for those 50 failures, even they made it about 85% of the way through retirement. That’s what the spending shortfall is telling you.

The other thing I want to talk about here, the initial portfolio value. The other values are median values. So portfolio value at retirement is 1.5 million dollars, median ending portfolio value 2.5 million dollars. And so the way those medians are pulled, we do 10,000 simulation paths and we sort the results.

So we have basically 10,000 ending portfolio values to deal with. How do we display that most effectively? You know, sort of an easy to understand approach is to show the median. And that would be to sort all 10,000 results in a table and then choose the result at row 5,000. That’s how we would pull a median out of that.

So again, ten thousand results, sort them in ascending order and then pull the middle one, the five thousandth one. and that’s what we see here. Just to give a little further on that, if we click this checkbox, that shows the portfolio value bands. Notice that we see the lines here are showing the lower and upper band. And that’s the 10 and 90 percentile bands and we can see. That looks like about $800,000 is the bottom ten percentile value, and about $7,000,000 is the top 10 value, or the 90th percentile value.

So that the bottom value is actually, again if you think of that sorted table, with ten thousand ending portfolio values, the median is the five thousandth, the bottom 10% is actually the one thousandth row. And then the top 10% is the nine thousandth row in that table, again sorted in ascending order. So that gives you an idea of the variation that we have.

The simulation is not producing one result and says, hey here’s your answer, we’re done. It’s actually producing a large amount of output and the job is to try to consolidate it into something that’s understandable.

And the other thing I wanted to draw your attention to is this percent of expenses funded. We’ve got the flexible spending policy selected. That’s the one with the decision rule. But because this is such a robust plan, we’re actually funding between 97% and 113% of expenses. So we’re not having to cut back our spending much more than just a few percent here on average in the retirement plan.

The other thing that I want to bring to your attention is the detailed view table. This is a year by year table that allows you to see the result– oh, by the way, the results both in the summary view and in the detailed view table are all shown in present value dollars, so the planner takes inflation into account, and all of these amounts are adjusted for inflation as we go through the plan, but they’re shown in present value, non-inflated dollars.

So this, the ending portfolio value, will be a much greater amount in nominal dollars but it will have spending power of 2.5 million in today’s dollars. And the detailed view is the same way. So are all of the results in the table.

And the simplest way to grasp this, I think, for many people is that the new investments and the retirement spending are actually adjusting for inflation. So they will increase every year in nominal amounts. But notice in the table they show as being flat. And that’s because they’re exactly keeping up with inflation.

And because we’re showing this in present value dollars those amounts stay stable, stay steady from year to year because they’re exactly keeping up with inflation. If an amount is growing greater than the rate of inflation it will increase from year to year in this table. And if the amount is growing less than the value of inflation it will decrease.

So we can see here at the retirement year at age 65, we see the cash flow situation starts to change. We go from putting $15,000 a year into our deferred account to now at age 65 we start planning to take $50,000 out of–we need $50,000 of spending. And so you’ll notice here the total withdrawal is actually $62,500, which is because the plan is grossing that up in order to account for income taxes that are going to be due on that tax deferred withdrawal.

The other thing I want to bring your attention to is that on the right hand, the rightmost column, is the probability of success and the number of failures in parentheses. So the probability of success is a cumulative number as you go from row to row or year to year. And then the number of failures are the number of times out of 10,000 that the simulation. ran out of money in that particular year.

So you’ll see we start to pick up some failures starting at age 80 even if we have one, and so it still shows 100% percent because out of ten thousand it’s just a rounding error when you only have one failure out of ten thousand.

But you’ll notice as go down the list we get more and more failures until we finally get enough, we probably need 50 to get from 100% down to 99%. I assume it’s rounding down there. And so we’re seeing the failures here from year to year, just as an indication of how things are going, what’s happening.

And I also want to show more details in this table. You can see even more now–this can get overwhelming pretty fast– but it’s a lot of data. And basically one of the things that you can see here in this additional detailed table is that we start to look at age 72. We start pulling RMDs out, and you can track what’s going on with required minimum distributions from the tax deferred account. So we start with–and we’ve got the amount, and the taxes, and then how much available for expenses, it looks like we don’t need to take any withdrawal out of the portfolio because the RMD is greater than what we need. I believe, yeah no additional withdrawal needed.

Now this table can be a little overwhelming, but if that’s not enough data for you, you could right click on any of the column headers and you can say show all columns which will give you even more data. And the idea is that I’m just basically dumping all of the information that I have in the simulation. And if you right click on any cell you can copy it to the clipboard or export the table, and it’ll paste right into Excel–I’m sorry, yeah, right click on this and this will paste into Excel so you can then manipulate it and manage it a little bit easier.

Okay I want to go back to the summary view. And let’s say because this plan is so robust, let’s see what happens if we can lower the retirement age to see if maybe we could retire at 60. Then we run it again and now we notice we get a 93% probability of success. So that’s not bad but the stoplight is yellow and not green. And the reason for that is because it’s just indicating that the percent of expenses funded varied because of the flexible spending policy down as low as 86%. So you might have had to take a 15% reduction in spending in some of the runs of this particular set of inputs in order to get that 93%.

And we can configure that by going into settings to see what the minimum amount or the maximum amount of spending cut we’re willing to take. So the minimum default is 75%. So we’re actually by default, the minimum spending cut could be as big as 25%, and that’s deployed gradually over the simulation. If the portfolio is continuing to decline, basically a really unlucky bad sequence of returns happens, the planner will knock back a few percent off of spending every year until it gets down to 75% of the desired spending and then it won’t knock back anymore after that.

But we can adjust this. And so let’s set this to 90% so that we’re willing to be flexible, but not that flexible, is what we’re saying here with this 90% minimum expenses to fund. So we’re again, we’re using that decision rule, but we’re only willing to cut back 10%. And let’s see what that does. We had a 93% probability of success and then we had the spending getting cut down below 90%.

We run it again. Now we’ve dropped the probability of success because we’re not willing to be as flexible but we see that we’re maintaining that 90%. So let’s see, maybe if we up the retirement age by a year. Does that get us to a 90%– and I’m picking a 90% just as a random amount. You know everyone has to choose what amount of success is adequate for them.

But notice now we can model that. All right, now we’ve got the green light because we’ve got a probability of success of 90% and we also are not having to cut back more than 10% in order to get that 90%.

So that’s just a general overview of the retirement spending model. Now notice if we chose the stable spending and we were completely unwilling and unable to cut expenses. We wanted to exactly fund what we asked for until the portfolio runs out of money. What happens then? And notice we dropped back seven or so percent so that flexibility got us a pretty decent boost in terms of the robustness of the plan. And that’s what we’re trying to model there.

Okay. Next I want to move on to a feature called sensitivity analysis.And this is a feature that lets you take a look at varying the parameters of the plan in a way that saves you from having to make a change. Hit “run”, make another change, hit “run” and try to make sense of all the different data.

For instance, by default this is set up to look at the portfolio return, standard deviation as parameter one, and the annual retirement spending as parameter two. And so we’re going to vary the return, and this is a case where we’re going to have the portfolio return and standard deviation march in lockstep. So we’re going to have 3% return with 2% standard deviation all the way up to 12% return with 18% standard deviation. And then we’re going to have spending vary from $40,000 to $90,000. So then we’ll see what this generates.

So we’re running the simulation actually about three or four hundred times here. These are three or four hundred times we run the simulation. Each of those times we use a different set of parameters. One parameter, two inputs. And you can see those up here. The right hand panel is showing you the results of one of those 300 simulation runs.

And so we’re at this top square, but we could also select another square by clicking on and we see what were the parameter one and parameter two values. We had a return of 10% and a standard deviation of 14%. We had retirement spending of $50,000 and 93% probability of success.

So we can get an idea here of what kind of return we need and what kind of spending we can support. And it looks like with this model we can just about support $56,000 – $57,000 or so, if we want a 90% something probability. Maybe even up to $60,000 or $65,000 if we assume a really high return.

But the idea here is just to let you explore variations in the inputs. Let’s do this again, only this time let’s say we’re wondering about what we should put down for the inflation rate. And we’re just how careful do we have to be about figuring out what inflation to use. And so we’ll put 1% as the minimum and 4% as the maximum. And then for the second parameter, let’s say we’re wondering about income tax. How much sensitivity do we have in our plan to changes in the income tax rate? And because we have so much in tax deferred, this is going to matter a lot because those tax deferred withdrawals are taxed at the income tax rate.

Let’s run this. And now we’re seeing again those three or so hundred runs and getting an idea of if we vary these two parameters what does it look like .You can see on the x-axis we’ve got the inflation rate increasing from 1% to 4%. On the y-axis we’ve got the income tax rate increasing from 10% to 50%. And so we can just get an idea now with one view. And we can see what the results are with these varying inputs to get an idea of how that is going to work out.

Great. And the last thing I want to cover on the planner is another feature that helps with the flexibility is the additional inputs. And so we get to additional inputs by clicking on the additional inputs button here. And in this case what I’m going to do is let’s just say that this plan is actually for a married couple and they’re not the same age. So I’m going to zero out this annual retirement income on the main input page, And instead I’m going to put it in on the additional inputs page here in the bottom. And I’m going to say starting at age 67 the older spouse is going to start collecting their Social Security and they’re going to get $12,500 for their annual Social Security and we’re going to say it’s 100% taxable. And we’ll say this is John’s Social Security. We add a row and now we have John’s spouse Mary. And Mary’s two years younger than John and she also wants to have her Social Security start at age 67.

But because we’re going to key all of the ages in this plan off of the older spouse, so when John turns 69 Mary’s going to start collecting her Social Security. And so now we have Social Security for two. Notice that we’ve got it set up that it’s automatically going to track inflation and we have it at 100% taxable.

And so now we go back to the main planner window. And if we run this again we notice there’s a slight drop in the probability of success because we had one of the spouses taking Social Security two years later. We bumped the retirement age up. We can get it back up to over 90%, and again, I’m just using that–I like to see over 90% but everybody’s got to make their own own rules about what works for them.

The other thing you can model with this is if you want to do one-time spending. You might have miscellaneous income, and let’s say that we want to downsize our house and we’re going to plan to do that at age 80. And so we’ll say we’ll have a one year cash flow. The start age and the end age will be the same at age 80. And then we’ll say that we’re going to net– maybe this will be to sell the house that’s out in the country and move into an in-town apartment or condo–and doing that we’re going to net $350,000.

So that will be the amount of the cash flow. And let’s say that we’re going to have to pay tax on that whole amount at the tax rate. Again we could adjust this and put a net amount in here and say the taxable rate is zero. If that’s easier depending on how complicated the taxes are going to be. Maybe this house sale wouldn’t be taxed at a normal income tax rate, so we want to adjust that and just do the calculation offline and put the net amount in here. Either way could work. And we’ll say sell downsize and we add a row.

And so now we go back to the main planner page and we run again. And now we can see that we’ve got a much higher probability of success. We probably can support that $60,000 a year retirement again. And we can see that this bump in the portfolio right at age 80, when the house sale kicks in, when the downsizing kicks in, so that extra cash flow goes into the taxable portfolio. And you can see it turns into a much more robust plan.

Another thing you might want to try in here–and we’re getting down to the end- Let’s say that starting at age 85 we want to simulate higher expenses because of maybe some medical things. Now maybe we just want to try this out and see what happens. It might not necessarily be what we’re going to plan for real. But we could explore this with let’s say $20,000 expense, and let’s say this is a medical expense.

And we don’t even need to put that in. We go back to the main plan and run it again and notice We now are down a little bit on the probability of success. We’re still increasing but not as quick of a rate, again after the house sale kicks in.

One last thing we could try here if we just want to– again we’re kind of playing around–and we maybe want to explore what happens if we just have really bad luck and we have a stock market crash at age 75. So 10 years in the stock market does really badly.

Now of course, the simulation, because it’s running this random return, random sequences are returned 10,000 times, we are going to get stock market crashes randomly dispersed through the whole simulation. But we’re going to just force a crash at age 85 and we’re going to force it to be -25%, with a standard deviation of zero. So it’s definitely going to happen, and so basically we’re overriding the portfolio return to make sure that we get this crash to happen at age 75. And let’s see what that does to our plan.

Now we see we have the stock market crash at age 75 drops the portfolio down quite a bit. We chug along for a bit, then we have the house downsizing which brings us back up and now we’re doing okay for the last part of the retirement. So we did manage to work through that, it appears, at least based on this model.

And so that’s just to give you an overview of the kind of things that you can do with the retirement planner.