Surprisingly, I have been able to find very little literature that addresses the best way to develop a retirement plan. As I have mentioned before, RIIA’s approach is the best I’ve seen but it is still a little too tactical for me.
A good place to start, I think, would be to decide what a good plan should look like in broadest terms.
When answering questions like this, I try to imagine myself at the future endpoint looking backward. For example, when I make a major decision, I imagine that it has turned out badly and that I have asked myself what I would decide if I could have a do-over. Would I ask myself why I had taken such a bad risk and vow never to make that mistake again? Or, would I believe I had made the best bet and just been very unlucky? In the latter case, I’d make the same decision again. I try to avoid the what-was-I-thinking outcomes.
I find this approach especially helpful when I’m considering a bet with a low probability of a really bad outcome. It's so tempting to imagine that low-probability outcomes, especially the awful ones, just won't happen. But, improbable doesn't mean impossible. Since most of retirement finance is probabilistic, a retirement plan is essentially a bet. I try to avoid having to one day think, “That bet was a huge mistake, but it seemed so unlikely that I’d lose it.” Going broke late in life is one such bet.
When I look at the retirement plan bet, I imagine that I am standing at the pearly gates and St. Peter asks, “So, how did that retirement plan work out for you?”
My first thought was that we should evaluate our plan by how well it met our individual household goals (past tense – remember, we’re standing at the gates), but that doesn’t take risk into consideration. A retiree who successfully funds retirement by investing 100% of her savings in penny stocks didn’t have a good plan, she was just extremely lucky. Lots of people successfully fund retirement with no plan, at all. So, retirement plans can’t be measured solely by the eventual extent of their success.
A good retirement plan, viewed in retrospect, would be one that had a high probability of achieving the individual household’s retirement goals. In other words, it was a good bet. Retirement finance is a stochastic (probabilistic) game, so retirees can win with a bad plan and lose with a good one, though we would expect that to be less likely than the reverse. A good plan is a good plan whether it succeeds or fails. A bad plan (or no plan) is a bad plan even when it succeeds.
[Tweet this]A good retirement plan, viewed in retrospect, would be one that was a good bet.
If my plan succeeded, I would want to tell St. Peter that I had a good plan and I got lucky. Were it to fail, I would want to be able to say that I had a good plan but poor luck and that if I had it to do over again for a hundred lifetimes, I would choose the same plan. Ideally, we should have no regrets if the plan fails, because it had a high probability of succeeding.
Now that we know how to define a good plan, how do we actually measure that? Michael Kitces posted a timely piece addressing this question at Advisor Perspectives recently. He came to a similar conclusion:
“Thus, in framing different retirement income strategies – and the trade-offs they entail – it’s important to scrutinize the measuring stick used to evaluate the outcomes. The best retirement income strategy will depend on whether you measure based on wealth, spending, probabilities of success, magnitudes of failure or utility functions that weigh both the upside and downside risks!”Kitces also lists several metrics that are often used to compare plans, including:
- Terminal wealth
- Total cumulative spending
- Probability of ruin
- Magnitudes of failure and adjustments
- Utility functions and risk aversion
Second, Laurence Kotlikoff and other economists prefer maximum smoothed consumption to total consumption because it smooths consumption over time and provides more spending when it is needed. A plan that maximizes total consumption but provides most of the spending when we are very old and desire less isn’t a good plan.
Consistent with Kitces’ assertion, I don’t believe that any of these measurements is “best.” In fact, we might employ two or more in a retirement plan to estimate the probability of achieving the household’s retirement goals.
Of course, the highest probability of success might not be very high, as in the case wherein a retiree’s goals are beyond their means. Even the best plan will have little chance of success in this scenario, but this is a problem with the goals and not the plan. Still, to cover our bases, let’s expand the definition of a good retirement plan to one that will have a high probability of achieving the individual household’s reasonable retirement goals.
Notice I said, "one that will have a high probability" of success. There will likely be multiple plans that have a higher probability of achieving household goals than other plans under consideration. Among the group of better plans, the retiree will need to make the selection based on his or her other preferences, because all of them will be rational choices. You might find a plan with a high probability of meeting your goals with an annuity and another that avoids annuities and prefer one over the other, for example.
Now we know basically how retirement finance works (nearly everything important is unpredictable and probabilistic and changes as retirement progresses) and how to choose a good retirement plan (pick one that has a high probability of achieving your household's individual reasonable retirement goals). We even know the rational way to cope with that uncertainty and change (update the plan periodically as new information presents itself).
Next time I’ll talk about incorporating those individual household goals into a retirement plan and why this should be the first step in developing a plan.