On retention

When I posted Saturday about what seems to me to be one major barrier in the way of any discussion of an NDP/Lib coalition discussion, my focus was on the apparent lack of a partner which the NDP can expect to deal with in good faith. And in comments, Malcolm from Simple Massing Priest noted equally serious questions as to whether the NDP and Libs actually share enough values to carry out an effective coalition/merger, and whether such a pact would be less than the some of its parts.

But Malcolm passes along some numbers based on the 2008 vote which call into question whether there's any point to a coalition even if the NDP, Libs and Greens all join forces:

I've done a quick tally of constituencies where the combined Liberal / NDP / Green vote exceeded the votes of the winner. My tally shows a total of 64 seats.

There are
* 45 of these seats where Conservatives won and the Liberals were the leading "coalition" party,
* 8 seats where the Bloc won and the Liberals were the leading "coalition" party,
* 8 seats where the Conservatives won and the NDP were the leading "coalition" party,
* 2 seats where the Conservatives won and the Greens were the leading "coalition" party,
* 1 seat where the Bloc won and the New Democrats were the leading "coalition" party.

I also calculated what proportion of the other "coalition" parties' votes the leading "coalition" party would need to retain on a net basis in order to win.

For example, in Simcoe North, the Liberals trailed the Conservatives by 22.0% of the vote. The combined New Democrat and Green vote was 22.7%. Therefore, the Liberals would have had to retain (net) 96.92% of New Democrat and Green voters. I say "net," because it is likely that some undetermined number of the other "coalition" party voters would actually vote for the Conservatives or Bloc as their second choice. In Simcoe North, if even 2% of NDP and Green voters were to support the Conservatives, it would be impossible for the Liberals to make up the difference. Similarly, if 4% of NDP and Green voters simply didn't bother to vote, the Liberals would be unable to make up the difference.

...(T)he Canadian Election Study suggests that the proportion of voters who would simply not vote is much higher than 4%, and the percentages of NDP and Green voters who would be inclined to choose the Conservatives over the Liberals is larger than 2%.

I trust, on that basis, that even the most rabidly pro-coalition advocate will concede that there will be some "coalition" party voters who will not conform. The argument is about how many, and whether the coalition is still viable.

I then went through and calculated the seat changes based on the leading "coalition" party retaining 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100% of the other two "coalition" parties" vote.

Frankly, the "coalition" did better than I had expected, achieving a combined plurality (ie, Lib + NDP + Green > Con) at the 30% net retention mark.

Of course, this does not take into consideration how the "coalition" would affect the vote in the other 244 seats. In some seats narrowly won by the NDP or Liberals over the Conservatives or Bloc, for example,it could actually result in a loss of seats where, for instance, Liberal voters may actually split in favour of the Conservatives in the absence of a Liberal candidate, as the CES suggests. I have not examined that question at all. Nor have I considered how the condemnation from the Conservative campaign would move Lib-Con or NDP-Con swing voters.

Net retention	Con	Bloc	Lib	NDP	Green	Total coalition
10%	136 (-7)	47 (-2)	84 (+7)	39 (+2)	0 (0)	123
20%	132 (-11)	45 (-4)	88 (+11)	41 (+4)	0 (0)	129
30%	128 (-15)	44 (-5)	92 (+15)	42 (+5)	0 (0)	134
40%	123 (-20)	44 (-5)	96 (+19)	43 (+6)	0 (0)	139
50%	119 (-24)	44 (-5)	99 (+22)	44 (+7)	0 (0)	143
60%	115 (-28)	44 (-5)	102 (+25)	45 (+8)	0 (0)	147
70%	108 (-35)	41 (-8)	111 (+34)	45 (+8)	1 (+1)	157
80%	101 (-42)	41 (-8)	117 (+40)	46 (+9)	1 (+1)	164
90%	94 (-49)	41 (-8)	123 (+46)	46 (+9)	2 (+2)	171
100%	88 (-55)	40 (-9)	130 (+53)	46 (+9)	2 (2)	178

Note that the model, while interesting, has serious flaws. In particular:

* people don't behave with the kind of marked consistency used in modeling the result shifts
* no consideration has been given to how the strategy affects the outcome in NDP-Conservative marginals or Liberal-Conservative marginals. Given that the CES clearly shows the Conservatives as the more popular second choice of Liberal voters, it could well mean the loss of ridings like London-Fanshawe, Sault Ste. Marie, Thunder Bay - Superior North, Welland, Edmonton-Strathcona, Western Arctic, Burnaby - Douglas, New Westminster - Coquitlam and Vancouver Kingsway,
* it also does not consider the potential for an electoral coalition to depress turnout for any of the "coalition" parties.

Finally, please note that the shifts modeled below reflect NET gain to the leading coalition party - meaning that it is the margin of votes gained by the largest coalition party over the votes gained by the Conservatives. By illustration, in a seat where the Liberals are the largest "coalition" party, and 30% of NDP and Green supporters don't bother to vote, a net 30% gain means a split of 50-20 of the NDP and Green voters that DO vote - which is actually a 71.4% share of those who end up voting.

As Malcolm notes, there are serious issues with using the 2008 baseline without accounting for shifts in voter preferences in the meantime.

But the above numbers still speak to the remote prospect that a coalition would actually manage to consolidate enough votes behind the participating parties to change the outcome of an election. A net retention rate somewhere between 20-30% is needed before a coalition would even exceed the Cons' seat count; and a 70% net retention rate would be needed both to push one coalition party alone past the Cons' number of seats, and to provide the coalition with a majority of seats in the House of Commons. Which is in stark contrast to the actual polling numbers which suggest that Lib votes would likely bleed equally to the NDP and Cons, while Lib votes would actually move primarily to the Cons to make for a negative net retention rate.

In contrast, results similar to those at the highest retention rates are entirely plausible within the realm of simple shifts in public preferences. The relative balance of power between the Libs and Cons in the 40-50% net retention range is virtually identical to what it was based on the 2006 election results, while the effective results based on an entirely implausible 70% net retention rate could match those arising from any of the frequent Lib/Con polling stalemates.

So comparing the minimal benefit to a coalition at any plausible range of net retention to the chances of achieving similar results simply by keeping pressure on the Cons, the better course of action looks to be for each party to keep competing to show why Con MPs need to be replaced with its own candidates. And if we end up with another election result which allows for a coalition, then that will be the time to work out a governing arrangement.