Curious on the paths people took to end up with Pralana software. What other products you try first? What issues with other products led you to try Pralana and to stay with it? Do you use others simultaneously. What do you think are it's "signature" features not found anywhere else? What are its limitations for you in your overall approach?
First I wrote my own spreadsheet. It had some things that were kind of cool, like 150 tabs, each representing a year since 1871. My solution to the end point problem (where historical traces only use the starting and ending years once, but the middle years over and over) was to wrap the data around. So after the final year of data, the next year in the sequence was 1871 again. I learned a few things from that
- When I put the results into bins, I found that the distribution of final estate value vs. historical frequency of occurrence was pretty linear, except for the top 25-30% of historical results and those great sequences all occurred after serious, long term, bear markets. Since we haven't had one of those recently, I tune my assumed returns to be about the 25%-30% of historical sequences as I know we are not living in the kind of bad times that set up the great future return sequences.
- I found that the US tax code is very, very complex and I was missing things so I shouldn't develop this any further 😀
I then ran across the bogleheads' Retiree Portfolio Model and it had features that I had missed in my own spreadsheet. I wanted Roth Conversion evaluation and it had more of the tax code, but was very manual. If you wanted to stay below an ACA limit or an LTCG phase-in, you had to do side calculations to determine how much to convert to hit those. It also had just one cell for adding expenses and one for windfalls, so I ended up writing very big formulas in those cells to accommodate the lumpiness of our plan. It had nothing to help with evaluating the effect on heirs' taxes from inheriting an IRA, or the effect of making gifts to the kids, so I made copies, one for each of our kids, filled in some guesses about their future and had formulas to move data between them. I made the mistake of trying to put in the iterative math needed to balance cash and that meant I played with the structure of the RPM program, so when the next year's version came out, I was stranded.
I ran across I-ORP and because of its LP engine, it could find ideas that you might not think of on your own, like using just a bit of Roth money to bridge cash flow gaps. However, when you started to constrain it, by, say, wanting an ACA subsidy, it might fail to converge. Also, the programmer had left out big chunks of that tax code and IRMAA was charged for the current year, not two years prior.
Then I happened to see Pralana advertised on the "Can I Retire Yet" blog and gave it a try. Wow! While Pralana has come a long way since then, even back then it was a breath of fresh air in its power and thoroughness.
I didn't try any of the others, but carefully read threads at bogleheads where folks compare them and it's clear that Pralana is far and away the leader in fidelity to the tax code and flexibility. Some of the others are reported to have better graphics that help in visualizing the results, so it's great to see Pralana adding some better graphs.
I am a statistics and spreadsheet person, and I had various spreadsheets going. (Still do.) I started a "big" one similar to Pralana, but found out pretty soon it was too easy to mess up on taxation and other details. Bogleheads RPM is widely admired, but I find it very counter-intuitive in its basic set up, too hard to figure out its formulas, and limiting in some things I wanted to do. In short I didn't feel it was worth the learning curve. At the same time, since I am a 'safety first', "lifecycle model" / "actuarial", I was using Ken Steiner's excel sheet to figure out my funded ratio and actuarial withdrawal.
I also fooled around with ERN (Kirstin Jeske's) Early Retirement Now SWR google sheet. I still use it, it is a work of art in doing historical analysis.
I-ORP was an eye-opener (pun intended), @Ricke, and it awaked my eyes to (1) the idea of optimization and linear programming, across the whole life-cycle, (2) the idea that when modeling, simplification of key elements is better than capturing every little details. (I think of I-orp a lot with some of the wish list requests I see for Pralana.)
I can't remember where I found Pralana, but what appealed to me was that it was excel and seemed very nuanced and nerdy. It only takes one "essential" feature to be a deal-breaker, and so far Pralana hasn't had any! It was also very reasonably priced. I didn't like having the formulas hidden from me. But I learned that Stuart (and now Charlie too) are very thorough and reliable in their methodology. They have very good instincts on what needs to be fixed or put off or ignored, and test new ideas for veracity. To be honest, I am stunned how few mistakes they have. And I really admire the way they vet new features. They also were very open and transparent. So in other words, even with formulas hidden, I could trust the results (i.e. trust them). That's priceless, and I can't say that for other software.
One thing that really distinguished Pralana for me was the "mode 2" ability to set overall asset allocations, which constrains optimization. I had begged the I-orp creator to do this years ago, and despite detailed email exchanges, he didn't understand why I wanted it. Low and behold, Pralana had already incorporated it! Lack of "mode 2" was a deal-breaker with I-orp and it gave me false results. And a deal-breaker with most other programs!
On a selfish note, I also found that the Pralana programmers were open to my good suggestions (and filtered out my bad ones), and I feel like some major features have been built in to my and everyone's benefit. This would never happen with a large company.
I also like how they explain where numbers are coming from. So I can check up on things to make sure I am entering and understanding. In fact, usually if something doesn't make sense I learn something by following the numbers. ("Oh, when I die my taxable account gets a step up!".)
If I had to find fault, aside from the fact it's not as "shiny" as Portfolio Lab and Boldin, it's the lack of ability to explore total consumption rather than just total savings. There are "expense" graphs but I don't find them useful. In the lifecycle model, consumption (be it discretionary or total) is key, and I've been forced to download pages from pralana and then do this on my own in excel. I also cannot figure out my funded ratio or actuarial calculations with Pralana, so also use TPAW (tpawretirementplanner.com) to figure out my actuarial discretionary withdrawals. (Maxifi is based on the lifecycle model and linear programming, so I'm eager to compare it to Pralana.)
I also find Pralana's monte carlo very primitive, with crude sd, a normal curve that understimates tails, rather than a scaled bootstrap I prefer for my own modeling. So I mostly use deterministic mode, and historical mode (though ern's is more detailed). I tend to use "fixed amount phased" for my discretionary withdrawals and I use consumption smoothing to approximate i-orp.
I retired in 2017 and had a background in decision analysis where we used modelers who would build stochastic models in excel using financial engines add-ins to provide financial output based on a structured decision analysis process.
About 2 years before I retired I had Vanguard and Fidelity do what they called cash flow modeling and retirement readiness. I found it very lacking compared to models we used at work so I started looking for alternatives for retirement planning.
I had difficulty finding anything until I came across Pralana and gave it a try. I used it for a few years and then me and the lead decision analysis guy at work agreed to guide a number of co-workers through what to include and how to make decisions prior to and in retirement.
It became clear that building a model from scratch was not going to be a realistic approach. We also recognized that a generalized model would not be easy to apply to personal circumstances.
I convinced my partner to try Pralana and we both thought it was very good back then-before all the great upgrades. We introduced it to all the co-workers who were interested and then provided training on how to use it and how to interpret the results. We repeated this process for several years as new co-workers were nearing retirement.
During this process I was introduced to Steve Chen when he was in the early stages of developing what is now Bolden. I agreed to use the software and provide feedback.
I did not like how Steve was developing the software such as ROR inputs and optimistic and pessimistic vocabulary so I quickly moved on.
To date, almost all of the decisions I wanted input Pralana on have been made but I will continue to keep my plan up to date because I am sure others will pop up in the future and I think it is important to have clash flow projections. I do not use any of the withdrawal options such as fixed rate or consumption smoothing - just income and portfolio compared to specific spending projections. I played with them and they may be good approaches for others.
I am very impressed with the online version. Because of my comfort with excel I was a little concerned about the transition.
@golich428 Given your background in stochastic decision analysis, do you avoid Pralana’s simplistic optimizations and Monte Carlo, and stick to just the cash flow? How did you do optimization when retired without your former excel tools?
If I had to find fault, aside from the fact it's not as "shiny" as Portfolio Lab and Boldin, it's the lack of ability to explore total consumption rather than just total savings. There are "expense" graphs but I don't find them useful. In the lifecycle model, consumption (be it discretionary or total) is key, and I've been forced to download pages from pralana and then do this on my own in excel. I also cannot figure out my funded ratio or actuarial calculations with Pralana, so also use TPAW (tpawretirementplanner.com) to figure out my actuarial discretionary withdrawals. (Maxifi is based on the lifecycle model and linear programming, so I'm eager to compare it to Pralana.)
I also find Pralana's monte carlo very primitive, with crude sd, a normal curve that understimates tails, rather than a scaled bootstrap I prefer for my own modeling. So I mostly use deterministic mode, and historical mode (though ern's is more detailed). I tend to use "fixed amount phased" for my discretionary withdrawals and I use consumption smoothing to approximate i-orp.
I looked at the MaxiFi author's paper (and advertorial) about Roth Conversions and saw that MaxiFi uses a weird definition of discretionary income that makes it unsuitable to me. Perhaps to make it easy for an LP solver or otherwise avoid programming complexity, it put taxes in the discretionary income category. So it would propose doing Roth Conversions in early retirement while you eat rice & beans, so that when you are 90, you can really party. Crazy, but easy to program while including Roth Conversions in the mix.
Pralana's Consumption Smoothing method is much more sensible, with your listed mandatory expenses plus taxes as the underlying spending and then brute force solving for the discretionary spending you can have during your lifetime. The downside of Pralana's approach is doesn't lend itself to integrated solving with Roth Conversions, but I've found it converges if you Consumption Smooth then Roth Optimize and repeat 2-3 three times. Not elegant, but at least the answer is more physically meaningful to me than what MaxiFi is doing.
In reviewing Kotlikoff's paper, I tried duplicating the results in Pralana. Kotlikoff made it sound like no other software would get the answer, but it turned out that the big conversions recommended came from the unrealistic asset allocation (100% TIPS) causing giant tax drag in taxable. Minimizing tax drag is rocket fuel for Roth Conversions so both MaxFi and Pralana found large conversions useful and I could pretty much duplicate the Roth plan in Pralana, with the caveat that there's no reconciling MaxiFi's oddball treatment of taxes as discretionary.
Can you elaborate on what is in TPAW that you would like to see in Pralana? I tried the TPAW planner and it didn't click with me, but Stuart and Charlie are always on the lookout for useful features.
While return distributions may not be perfectly normal at relatively high frequencies (e.g., daily), return distributions become increasingly normal over longer periods (e.g., annually). Finally, financial plans require a significant number of other assumptions that will have a materially greater effect on the outcome (e.g., saving, spending, length of retirement, etc.) than the specific return distribution.
There is more uncertainty on the average ROR for each asset class than the shape of the return distributions. In most if not all planning software with maybe limited exceptions ROR is the only variable that we attempt to provide a range of uncertainty. The uncertainty in the tax code alone probably creates more so called error in future cash flows than the exact shape of realizedrates of return. What about the error in longevity - huge range of probable outcomes.
I am not sure what you mean by simplistic optimizations? Are you referring to social security and/or Roth conversions etc. I run those tools and use that information as a guide but I may have overriding reasons to deviate from the results.
As far as optimizing my plan - again optimizing what value. Is it maximizing end of life wealth portfolio balance, maximizing early retirement cash flow, maximizing tax free legacy, maximizing inflation adjusted income or minimizing income taxes or some combination of the above.
I don’t think it is possible to optimize. We can only create a reasonable plan that can help us make informed choices among the alternatives we have for each decision that is important to us.
For example, if we could have optimized our plan based on the previous tax laws, what would we think of that plan now that we have new laws - they could change again in a few years.
Planning is a process of making the best assumptions based on our current situation and then using the Bayesian approach of updating them as new information becomes available.
The tools I used before retirement were customized to a specific problem with unique variables that typically only applied to that particular investment.
