jiHymas wrote:The IRR calculation assumes symettry between borrowing and investing, which is not the case.
When you withdraw cash from your account, you are not forgoing the returns that you achieved while the cash was there. You are not paying 428% annual interest for the loan, as in this example. Your rate of return on the money withdrawn is zero.
Perhaps you're expecting too close of a parallel between the investment account what I called the symmetric LoC. The latter is a simple conceptual device which selects (hopefully!, more below) a rate that when applied to the same cash flows produces the same final result at the end of the same time period. The two "accounts" get there in entirely different ways, that's true.
jiHymas wrote: This is how people think (and should think), and it explains why the answers given by IRR look so wrong.
I'd say one "can think" about it the IRR way, amoung others. It is what it is.
I still find your belief that "the answers given by IRR look so wrong" very puzzling. Let's revisit briefly the Joe Trader example and look at it several ways, some of them naive:
(a) The inherent/intrinsic return, and TWR, on the market was 100%. This is just fine, but it doesn't tell us whether Joe's timing in this index was helpful or not. If we want a feel for this, we need to look at something else, so let's do that:
(b) The average amount of his own capital Joe had invested was $125. He ended up with $600. Naively the return was (600-125)/125 = 380%.
(c) Let's internalize the cash flows into the portfolio, with an interest rate of 0% (just to keep it simple). He started out with a $100 investment and cash of $0, and ended up with a $600 investment and debt of $100, net $500. Naive return: (500-100)/100 = 400%.
Compared to both of these naive gauges, the IRR of 328% (not 428% which is 1+r) looks rather modest, if anything
, but certainly in the right kind of ballpark.
So I will ask, how would you prefer to get a handle on Joe's return taking account of his market timing, but not using IRR? We know IRR is "wrong", but what is your idea of the "right" answer to this question? (TWR doesn't apply here since it removes the market timing.)
ghariton wrote:Another problem is that using IRR can lead you to prefer a very small but very profitable investment to another, much larger project with a lower IRR but a higher absolute return -- that's a problem when the two projects are mutually exclusive.
I do agree with that. IRR isn't a panacea that should be used for everything. It really focusses on a given portfolio which could be quite narrow (e.g., a single bond). In a wider picture the cash flows may be important, especially if they stay within your control, and so what is done with them and the returns achieved there can matter.
IRR applied to a given portfolio is sensitive to the cash flows. That's what it can tell you. It doesn't tell you about everything outside the portfolio. That can be a benefit (a clear simple focus) or detriment (ignores wider effects of a project). That's the crux of it to me.
ghariton wrote:And of course, if there are many inflows and outflows, finding the IRR can require you to solve a high-order polynomial, with many different solutions: Which one should you be using for your analysis?
This is a great point. It is possible for IRR to have multiple solutions when there are both + and - cash flows (a necessary but not sufficient condition). At the very least this is something that would be hard to explain to "Granny Oakum" when she asks why there's no return (or several!) given on her quarterly statement
. Theoretically, it's also rather unsatisfactory.
OTOH, as probably reflected in the wide use of IRR, it does seem that the actual occurrence of this problem in practice is quite rare. Pragmatically, I look at it as the price to be paid for what IRR can tell you. Does anyone know of an alternative, which does reflect market timings, and is similarly tractable but without drawbacks?