Ron said:
However, simple mathematics shows that if you take the 11.96% (actually
11.9552927760608%), divide it by 24 to get 0.498137199002532% and apply that
interest rate sequentially to the cash flows posted by the OP, you obtain the
final result that he posted. To me that means I am earning an effective rate
of 11.96% annually, or .498% per period.
And I could say that if I take my 12.6658%, take the 24th root [1] to
get 0.4981%, and apply that interest rate "sequentially" to the cash
flows posted by the OP, I obtain the "final" result that he posted [2].
To me that means we are earning an effective rate of 12.6658%
annually, or 0.4981% per period.
The point is: we both compute the same periodic rate (0.4981%) first,
then apply some transformation (multiply or compound by a factor 24) to
get our respective annualized rates. So of course, when we each apply
an appropriate inverse transformation (divide by 24 or take the 24th
root), we get back to the same periodic rate that we both started with.
And of course that periodic rate works, since we both computed it
using essentially the same IRR() computation [3]. It proves nothing
for either of us.
I believe the real proof is as follows ....
We can compute the PV of each cash flow using PV(0.4981%,n,0,CFn),
where "n" is the period number and "CFn" is the cash flow for the
period [4]. The sum of the PVs is zero, which is the definition of the
IRR.
Conversely, we can compute the FV of the PV of each cash flow using
FV(0.4981%,n,0,PVn), where "n" is the period number and "PVn" is the PV
of the cash flow [4]. Of course, each FV equals the corresponding
periodic cash flow. That is not surprising: again, I have simply
applied inverse transformations.
But the point is: as we know, FV() assumes that "interest" compounds
periodically. So by the same token, we can take a PV of -$1 and
compute its FV a year later over 24 semimonthly periods using
FV(0.4981%,24,0,-1). The result is $1.126658. Thus, we earn $0.126658
in a year for each $1, which is an annual rate of 12.6658%.
To me, that means that 12.6658% is indeed the annualized IRR of the
semimonthly IRR.
Be that as it may, the fact remains that many people do indeed multiply
or divide to convert between annual and lesser periodic IRRs. Of
course, that is a simpler computation. The dollar error is typically
relatively small -- 5.61% in this case.
HTH.
Footnotes
--------------
[1] I really mean: (1+12.6658%)^(1/24) -1 .
[2] The correct procedure is to use the periodic rate to compute the PV
of the cash flows and demonstrate that the sum of the PVs is zero,
which is the definition of the IRR.
[3] IRR(L11:L25, 0.2235%), although you might have used a different
"guess".
[4] Table of PV and FV of the cash flows. Forgive me if the columns do
not line up. Some newsreaders do strange things, notably Google Groups
:-(, and different font spacings do not help. Hopefully, the only
confusing line might be the last one: the zero is the sum of the PVn
column.
n CFn PVn FVn
0 -600000 600000 -600000
1 -150000 149256 -150000
2 0 0 0
3 0 0 0
4 -5000 4902 -5000
5 -150000 146319 -150000
6 -200000 194125 -200000
7 -400000 386326 -400000
8 -150000 144154 -150000
9 -310000 296442 -310000
10 -150000 142729 -150000
11 -350000 331383 -350000
12 -300000 282634 -300000
13 -415000 389040 -415000
14 3288288 -3067310 3288288
SUM(PV0
V14) 0