Because it just does:
The three IRV countries:
Ireland (mandated in their 1937 constitution),
Australia
and Malta (and more recently Fiji
for a brief period of IRV democracy before its coup)
all are 2-party dominated (in IRV seats)
– despite having many other features in their governments
which would seem much more multiparty-genic than the USA with IRV added will ever have.
So you can be sure the USA with IRV would be 2-party dominated too.
E.g.
"The composition of IRV seatholders in Australia is very biased against
third parties: examining all 564 statehouse & federal IRV seats we found only a
single one occupied by a third party."
Australian analysts themselves agree that IRV leads to 2-party domination –
this is not just crazy old me talking. A quote from
http://www.australianpolitics.com/voting/systems/preferential.shtml
is IRV "promotes a two-party system to the detriment of minor parties and independents."
Because here is an example election
and here is another in
which, if some voters honestly order the candidates N>G>B where N=Nader is the "third
party" candidate, then N and G both lose, but if those voters
dishonestly vote G>N>B or G>B>N, then G wins (a better result in their view).
In these example elections, just like under the present plurality system, voting Nader
is therefore strategically foolish. A possible reason why the Australian
House remains 2-party dominated after 80+ years of IRV is that
these kinds of elections happen
often enough so that strategic voters feel logically justified in thus "betraying"
Nader. Hence: the third parties suffer
a tremendous disadvantage, hence they die off.
And hence voters observing this year after year
realize the third parties have no chance, which justifies
their strategic vote-exaggeration/betrayal decision all the more.
(Who cares if you betray N if he had no chance anyhow?
It is worth it if it increases the chance
G will win – or so they reason.)
Result: Vicious cycle – entrenched 2-party domination – third parties die.
We do not know that this is the actual reason, but we will say that
it would be entirely logical if it were; the only way it is not the actual reason is
because of the insufficient-logic of Australian voters. If we instead assume
more-mentally-challenged voters who don't know or can't understand this, then they
will likely just give the two-most-likely-to-win candidates top & bottom rankings
– without worrying about whether
that really is logical or not – to "max out their vote's impact."
That is intuitive and requires no thought at all
(but nevertheless is supported by deeper analysis)
and with IRV will prevent a third-party
candidate from ever winning.