I've been working with spreadsheets on and off for 30 years, literally. In 1987 I started to use Lotus 123 for the first time in my first year of grad school. I even had one of the first retail-friendly portables about the size of a decent sized woman's purse. It had those big giant floppy disks. (usually around this time I'll mention that I had to use a slide rule in what they now call middle school; I have to go to google images to prove to my kids there was such a thing). That spreadsheet skill plus a little finance knowledge carried me pretty far through however many careers I've had so far.
When I started looking into retirement fiance both before and during my blogging phase I leaned heavily on spreadsheet modeling to look at various aspects of the retirement problem. For the most part that was ok and it got me where I needed to go. But I kept bumping up against boundaries in spreadsheeting, things I couldn't do, places I couldn't get to. Either the data was too big or the solutions were too awkward or integrated sheets were too brittle or the packaged statistical features were lacking, and so forth. I was limited in what I was able to do for myself in unlocking some personal understanding of retirement finance processes. Here is one simple early example. Back in 2010 when my banker offered to re-run with his "team" what was essentially a MC sim for a $4000.00 fee (keep in mind I was already paying more than a point for limited services at that time) I said no thanks (to myself in my head I used more robust language) and I thought "I wonder if I could do that myself if all I know is excel." I stuck that in the back of my head and about 3 or 4 weeks later when I was folding laundry it popped back into the front of my head that I knew how I could do it, except that I realized I'd have to learn a little VBA. Took me about a day to hack it out. The problem was that it ran for about an hour or two and locked down Excel if not the computer. This was a limit.
Then I was reading something by Dirk Cotton about some simulation he'd done in R. He's about my age so I said to myself "what's an R? And if he can figure it out at this age so can I." So I did. It took a couple of days to figure out the platform and basic language terms. With that and the ability to google: "how do I do ____ in R?" it took me about a day or two to re-write the simulator and another week to add about a zillion more features I couldn't have done in excel. The run time for 10 or 20,000 cycles was now down to about 2 minutes. It'd be even a little faster without the naive amateur coding. That, however, opened the floodgates for me with what seemed like an infinite capability and there were about 10 other problems I had that were instantly solvable with the ability to code it out: e.g., backward induction with SDP, perfect withdrawal rate sims, geometric intermediate term expected return sims with spending, and flexible ruin rate estimation that satisfies the kolmogorov PDE among other things. That last gets me the same result as my simulator except the process time is now down to about 3 seconds. Between the flexibility, the prepack statistics and finance modules, and the data visualization I'd say that learning R is one of the more explosive productivity jumps I've ever had over my career(s). A youthful expert might mock the code but either way I don't feel very limited in what is possible these days.
That brings me to the math. I do not have formal training in calculus or statistics. My last calc class was in 1977 and I took some basic stats in my MBA program. That was limiting because I couldn't read half the academic papers I try to read (many of which are probably a waste of time even if I could). I also was having a hard time understanding and taking the "good stuff" and implementing something in either Excel or R. I don't like boundaries so back to school for me. Online MOOCs while imperfect are at least available and got me part of the way in brushing up. Books helped. Asking people took me another part of the way. Google always helps a bit. The rest was just blind stubborn perseverance at trying to figure stuff out. For example I read Huang and Milevsky's "Ruined Moments in Your Life: How Good Are the Approximations?" about 20 times before I got the "ah, now I see it" moment amidst all the dense calculus. Same thing with the withdrawal rate math in the technical appendix at earlyretirementnow.com except there without the calc. That site took me a few days or weeks before I got the feel for what he was doing at more than a surface level. Why the effort? I just needed to know how this stuff works and I wanted to be able to see it in my own head, not intermediated by someone else. It also allows me an easier path to synoptic reading because the same principles, but in highly varied notation, keeps showing up in paper after paper. That kind of contextualizing is useful and if one is naturally skeptical in life, it's helpful to know what people are saying and why they are saying it and where you have seen it before. It's maybe a little like when I tried to learn French again at age 40. While I can't speak it, getting myself to the point where I could read (at least juvie lit, anyway) without thinking in English I found gratifying on its own terms and also helpful in reading through some of the harder french that is sprinkled around the English lit I try to keep up with.
So what's my hindsight point here? In order for me to try to come to terms with the retirement finance that I felt like I needed to know -- something Bill Sharpe has called "the hardest problem I've ever looked at" -- I had a choice. I could walk away and trust others to tell me what it all means or I could embrace the math and the challenge and the technology and see where it takes me. I chose the latter, and while it's not everyone's cup of tea, I'm glad (so far anyway) that I did....in hindsight. I probably should have started earlier, though.
No comments:
Post a Comment