Thanks for all of the thoughtful discussion, everyone!
I was actually thinking about this a week ago. I feel like analysis that I do is somewhat flaws because it is not including those years, some of which are most likely going to be failures in my plans (2000,2001,2002,2007). This is how all calculators work, not just yours, but it would be nice to have an idea on those. One idea that I was thinking was a box that would fill in historical data for returns and inflation for the years that we do not have data for. For example, on a 30 year time horizon and that boxed checked. The year 2000 scenario would use 2000-2020 actual data and then from 2021-2030 it would use the historical average (or even a user input number) for returns and inflation. Thoughts?
I've been thinking about this, too. I think it's OK to use averages. In fact – and you may already know this – William Bengen (who invented this style of retirement planning analysis afaik) did this in his original study for some of the retirement durations (he describes this in "Appendix: Assumptions of Computations of Portfolio Values" in "Determining Withdrawal Rates Using Historical Data" - William P. Bengen, 1994).
I agree with
@swashbucklinstache that incorporating this steps the calculator in a direction toward becoming a Monte Carlo calculator. You suggested adding a full Monte Carlo option earlier, and I think that is a great idea, so I'm not sure if it makes sense to add this partial option. Does it provide much more value over a full Monte Carlo option? Would it make the UI overly complicated? I'm undecided right now, but I absolutely think it's worth considering further. Right now I'd say I'm leaning in the direction of having all 3 options:
- real data only
- real data + averages to fill in the gaps
- full Monte Carlo
A thing to consider though, is that you might want a way to tell the user / weigh results based on e.g. "95% of this run's results are using averages, not actuals" because as we all know that makes a huge difference.
Completely agree. If the results ever included average data, then that would need to be clearly stated.
I think my preferred way to do this is, though I haven't thought too much, done the arithmetic, nor had any coffee today, is a hybrid approach:
I
really appreciate you taking the time to write out this algorithm in such detail! Thank you! I'm going to sit down in a bit and walk through it really make sure I fully understand it before giving a reply. I'll let you know if I have any questions.
Would love to see real estate investment options added. A bunch of folks have between a few and a whole bunch of properties to fund retirement.
Great idea,
@Blindsquirrel ! I'll add this to the todo list. Thank you!
1) incorporating other features of popular calculators like graphing make a lot of sense as you get there =).
What kinds of graphs would you like to see? Other than the graphs that currently exist (portfolio over time and withdrawals over time), I was thinking of adding one to show the stock market performance over the period. i.e.; a chart showing how $1,000 that is 100% invested in the S&P 500 changes over the retirement, with no withdrawals/additions to it (kind of like Vanguard's performance charts for its funds).
At the moment I'm not considering adding in a graph that shows the portfolio values of the simulations on top of one another. I'm not really sure what information those graphs provide, but I'm open to hearing how others use that particular graph if anyone feels strongly that it should be a part of FI Calc!
2) A down the road thing might be allowing users to customize the color coding - yellow for you might mean "money left at the end, but not much" and for me I might want to model it with yellow to mean "you've dropped below 75% of your inflation-adjusted stash" a.k.a. adjust the definition of failure. This gets away from the trinity study mindset but drifts towards real-world use of the calculator. Read through ERN to get a better sense for this (and for why things like "i'll just get a part time job if my stash drops below X%" aren't as great as they sound). It also blends the line between calculator and something more than that, e.g. the portfolio visualizer website.
Yeah! I love this idea, and it's been the long-term vision for the color-coded years. It's definitely a down-the-road thing, but I'd love for users to be able to write their own little "algorithms" to color code years according to the things that they're interested in. Although I'm skeptical that I could ever cover all of the custom analysis folks do in spreadsheet apps, I think I might be able to get a good chunk of those use cases in the app? It's a cool feature to think about.
3) This one depends on the core intention of the calculator.
This is an interesting idea. It sounds kind of like Monte Carlo, but on the spending side rather than the market performance side? Is that right? I'll need to think more about it. If you have more thoughts on it over time ya gotta keep me updated! I'm definitely curious to hear more.
Love it!!! The only add-on thing I recommend is being able to adjust for taxes, so we can adjust for higher federal/state/city tax rates as appropriate (especially given all the deficit spending going on now).
Absolutely, this is a great idea
@tk2356 . My main concern is getting the tax code for all 50 states + the federal government for as many years as possible. If I spend a long time reading the tax docs that they publish, I can usually figure them out, but doing it 51 times across 100+ years (or however many years are available)...that might take me ages, and I'm worried I would make some mistakes. I need to do more research into the feasibility of implementing this feature, but it is a really good idea. Maybe it's not even as complicated as I'm thinking. Either way, I'll add it to the list!
Other things I'm working on:
- Alternative withdrawal schedules. Right now, it is set to 1 year, but I am working on rewriting the algorithm to support monthly, quarterly, and twice-yearly.
- Customizing the CSV export columns. i.e.; you can add/remove columns based on what you find useful. These settings would be saved, so once you configure it once you don't have to change them again (unless you want to). In the distant future you might even be able to write calculations for custom columns (using a similar system to the custom color-coded years described above).
- an optional summary CSV export. Rather than year-by-year results, you would just see 1 row for each year.
I'm also likely to add a "Guides" section that goes into extreme detail how all of the algorithms in the calculator work.
One example of a guide could describe what happens in a simulation year. Bengen's analysis worked like this:
1. Year starts with a given portfolio value
2. Adjust portfolio for gains/losses
3. Withdrawal occurs based on initial portfolio value
4. Portfolio is rebalanced
FI Calc swaps 2 and 3, but there's no way for anyone to know that unless they're reading this thread. So I want to write that in a guide somewhere. Also, I think it might be interesting in an advanced setting to allow users to customize this order of things.