Invariably when this concept comes up there is a comment or two about margin call risk so it's important to agree first off that this strategy only really makes sense if there is a reasonable buffer between the amount of money you will need in retirement and your total assets.

It's not really the margin call risk that gives me pause, but rather that if you plan on doing this any % of fixed income in your portfolio no longer makes any sense at all. It is not sensible to borrow money at a (p+x)% rate while you have fixed income investments paying out at a (p-x)% rate. So this got me thinking about whether fixed income is really that important on the longer time scales, and I have more or less concluded it doesn't make much sense at the 10+ year time frame. The rationale is as follows.

Consider an investment into VTI, and a reasonable risk model for this based on historical trends, the asset should grow an exponential rate (e.g. 9.85% annually) and will have error bars around that number, say at the p95 quantile (i.e. the 5% worst case). A key observations is that even under the most conservative assumptions it's generally assumed that the error

**does not grow exponentially**over time, and certainly not at a faster rate than the mean.

All I'm trying to observe here is that the

**bottom of the error bar**is going up steadily if we plot this over time. With a formula that looks like this: 1.0985^t - c*t^2. For some constant c, which depends on VTI holdings etc. And that c^t functions grow a hell of a lot faster than t^2 functions.

The upshot here is that at some point in the future the

**worst case**for VTI is better than the

**best case**for fixed income. We would have to squabble about precisely when that is, as we debate about how much past variance can predict future variance. But when I sit down and make what I consider to be reasonable assumptions cross over point seems to be 10+ years.

Another way of saying this is that risk only really matters in the short term, when your planning to actually draw down from the portfolio. But I see so many investment channels, forums posts, and so on about using the 80/20 on the 15, 20, or 30 year time frame and I just don't see how that makes any sense. Also, when you throw in rebalancing on deposit into this it's even worse as you end up contributing almost everything to the fixed income channel (Rob Berger has a good video on this).

Appreciate any feedback on these thoughts.