Anomaly watch: Nova Scotia goes to the polls today

UPDATE, 14 June, morning and again in the afternoon:

The CBC link (first one below) has been updated with early vote count results, which suggest the Tories will retain their plurality in the assembly, 23 seats, or a loss of two. The big gainer was the NDP, going from 15 to 20 seats, while the Liberals–whose leader declared the party would double its seats, instead lost three (including that of the leader himself!) and now has nine.

So, does this result qualify as anomalous? The incumbent party lost seats but may remain in office unless the other two join against it when it faces parliament the first time. The second party gained the most, but can’t form a government unless the Liberals back it.

In the votes breakdown, it looks to be: Conservative, 39.6%; NDP, 34.5%; Liberal, 23.6%. So both leading parties gained about three percentage points at the expense of the Liberals, yet one of them lost two seats (3.8%) while the other gained five (9.6%). Very odd indeed. Ah, the fun of three-party FPTP elections!

Don’t miss Rici’s comment below. Among other interesting details, including an ‘experimient’ with regional PR, he notes that the result it not very anomalous and unlikely to give any impetus to an electoral-reform push. I certainly agree on that latter point. In fact, one of the things I am saying at this weekend’s conference in Montreal is that there is no clear path to PR from minority governments. That may seem like a paradox–after all, minority governments already introduce an element of interparty cooperation in parliament unlike the typical Westminster–FPTP pattern, and they almost by definition give leverage to smaller parties that might be expected to prefer PR. Yet if the theory in my own paper [PDF] is anywhere near accurate, minority governments are unlikely to provide the conditions needed to generate a reform process.

In a second and very rich and interesting comment below, Rici saves me the trouble of applying the seat-vote equation to this election, and notes that according to the equation, this is another unexpected minority government.

Nova Scotians vote today in a provincial legislative election. The current government is a Conservative minority. Nova Scotia is one of the provinces that has a real three-party system. Many districts are expected to be close, which could cause (as in past elections) odd relations between votes and seats.

What follows is a re-post of my previous discussion of the province’s elections (originally posted 14 May).

Nova Scotia Premier Rodney MacDonald has called a provincial election for June 13.

The last general election for the provincial assembly was in 2003, and it resulted in a narrow miss of a majority for the Conservative party: 25 of 52 seats. With the economy doing well, MacDonald hopes this time the voters will get it “right” and give his party a few more seats. MacDonald’s decision, as a premier heading a minority government, to call an early election puts him in good company with his co-partisan counterpart, Stephen Harper, as Wilf Day put it in a very interesting seed beneath an earlier planting on the Canadian federal minority government. Wilf’s comment offers a quite appropriate lesson from Irish history, and asks why the media seem to think it is perfectly normal and understandable that a minority government ought to go to the polls early in search of a majority.

The question is relevant because it is not as if the governments of either MacDonald or Harper obtained anything just short of a majority of votes. Like Harper’s federal Conservatives in January of this year, MacDonald’s Nova Scotia party obtained only around 36% of the votes in the 2003 election. (For more on minority governments and FPTP, go to my Canada subdomain and scroll down to early February and late January.)

In fact, the Nova Scotia result in 2003 was rather anomalous:

    Party, votes, seats
    Conservative, 36.3%, 25
    Liberal, 31.5%, 12
    NDP, 31.0%, 15

This is arguably a worse result than the federal election, in which the two two parties’ votes percentages were almost the same as in Nova Scotia, but the leading party won just over 40% of the seats–much closer to its actual voting result. Moreover, in the federal election, the third party (also the NDP) was much farther behind in votes and seats, whereas in Nova Scotia, the third party (in votes) actually obtained more seats than the second party.

Despite the rather anomalous nature of the Nova Scotia outcome, in my current research on “systemic failures” of plurality electoral systems and moves towards proportional representation, Nova Scotia does not show up as a severe or even moderate case of failure. I define the inherent conditions for reform as chronic under-representation of the second party (second in seats, that is), based on expectations derived from the seat-vote equation. When this underrepresentation occurs in a very close election–a contingent factor–it becomes noticeable and puts reform on the agenda, although the initiation of a reform process happens only after an alternation to that (now former) second party–a further contingency.

Based on the parameters of the seat-vote equation–the size of the assembly, the number of voters, and the actual ratios of the leading parties’ votes–the 2003 Nova Scotia election might have been expected to produce seat percentages for the top three parties of:

44.5 — 28.4 — 27.1,

instead of:

48.1 — 23.1 — 29.0.

The five percentage-point shortchanging of the second party is high, but not anything like extreme, compared to other countries and provinces, or even compared to Nova Scotia in the 1970s. (The Liberals obtained more than 20 percentage points less than expectation in 1967!)

Thus, Nova Scotia does not look like a candidate for a serious electoral-reform movement–yet. However, if MacDonald gets his wish–a seat majority–without a very large boost in his party’s votes, and if the Liberals thus fall farther behind even without a major votes loss, then the province could go on my “watch list” for likely electoral-system change.

With a three-party system despite plurality elections, Nova Scotia looks like a good candidate for PR. But my research shows it is not three-partism, per se, that generates serious reform processes. Rather, it is underrepresentation of the second party and close elections that do so.

0 thoughts on “Anomaly watch: Nova Scotia goes to the polls today

  1. Seems like the results were not “anomalous”. As you say, the 2003 results were, so it is not really fair to look at the relative vote/seat swings between 2003 and 2006.

    Based on the numbers and regions on the CBC web-page, I did a quick PR simulation assuming list-PR in each of the regions shown. (I have no idea whether that’s a logical division or not, but it was handy.) The district magnitudes are 4, 5, 6, 9, 11 and 17. That would give: PC 21, NDP 19, Lib 12, which is not that far off from the actual results (PC 23, NDP 20, Lib 9).

    The apparent similarity between the simulation and the actual results hides a difference in the regional allocation of seats in the PC and NDP caucuses. In the actual election, in the Northeast (where the PCs won a bit over 50% of the vote), the allocation is PC 9, NDP 2, Lib 0; the simulated result is PC 6, NDP 3, Lib 2. On the other hand, in Metro Halifax, the actual results are PC 2, NDP 13, Lib 2 while the simulation gives PC 5, NDP 9, Lib 3. In other words, the PCs are significantly underrepresented in Metro Halifax and significantly overrepresented in the Northeast. (In the smaller districts, the simulation is pretty similar to the actual results.)

    The actual election has a small bias introduced by the fact that the Liberals did not run a candidate in one constituency (Queens) where the NDP won by a very small majority; it could be that had the Liberals fielded a candidate there, the final results would have been PC 24, NDP 19, Lib 9. I tried adjusting the simulation to take this into account, but it made no difference to the actual results even if 20% of Queens’ vote was shifted from the NDP to the Liberals.

    Conclusion: although the election results are reasonably proportional, and accurately reflect the overall “shape” of the vote (i.e. a minority government where the PCs have a small plurality over the Liberals*), a regional list-PR system would provide better regional balance.

    I can’t see that providing a huge impetus for the adoption of PR, though.

    [* That should be NDP, right?–ed.]

  2. Indeed, that should have been NDP. What was I thinking?

    I tried your seat-vote equation, although I’m troubled that the sum of the predicted seat percentages ended up being 102%, which either means that I did something wrong or that the equation needs some sort of adjustment. Anyway, that would yield a Conservative majority, so in terms of your interesting paper the actual election qualifies as an unexpected minority, but with an underrepresentation (for the PCs) of 80%, which does not meet your criteria for failure (75%).

    The weakness of the seat-vote calculation, as you point out in your prediction of the Canadian election (, is that it fails to take into account regional variations. The simulation which I referred to above shows that there are significant regional differences in Nova Scotia. In particular, the Liberals came second in three of the regions, with the consequence that they are “overrepresented” (in the sense of expected behaviour of plurality) by 180%. (The model predicts five seats rather than nine.)

    Although this effect is most clearly identified as “regional”, in fact it is possible for a bias to be introduced by any division of constituencies in which there is a marked concentration by some party. For example, there might be a marked concentration of one party in rural constituencies and by another party in urban consistuencies, regardless of regions. This is far from uncommon, and is often exacerbated by disproportionate allocation of seats favouring rural voters; the effect can be seen both in plurality systems (this was the subject of electoral reform in Canada, but I believe it persists to a minor degree) and in proportional systems (it is quite clear in Spain and Perú, for example).

    Even when “regional” concentration does not distort the overall results (for example, in the Nova Scotia election, where the Halifax/NDP vs. Northeast/PC biases cancelled each other out), it may have pernicious effects and possibly deserves more prominence in the impetus for electoral reform.

    In Nova Scotia, only 2 members of the PC caucus of 23 (that is, 8.7%) represent metro Halifax, which is about one-third of the province’s population. This suggests that Haligonian issues will be underrepresented in the government’s agenda. Nor will the PCs have much motivation to favour Halifax; they will do better concentrating their resources (that is, government programs now, and campaign funds in the next election) on parts of the province where they have more potential.

    On the other hand, disaffected Haligonians cannot do much to change the results of a future election; the majoritarian effects of FPTP have already given opposition parties almost 90% of the Halifax seats. But it is hard to see an NDP caucus of whom 65% represent Halifax favouring a proportional system given that such a system would cause quite a number of them to lose their seats (the compensating seats which would be gained elsewhere are not members of caucus).

    That said, multiple-district PR is not necessarily going to produce better results (although it might in the case of Nova Scotia) because of the district magnitude effect. The district magnitude effect can be minimised by making all districts roughly the same size, or by making them all quite large, but the same logic in which all major parties are willing to cooperate in gerrymandering plurality ridings could well operate on districting decisions, particularly when the consequences of such decisions are not well understood.

    I apologise for hi-jacking this comment for such a long ramble 🙂

  3. Rici, excellent comment. Yes, with that votes ratio we would certainly expect a majority govenment. But, given the regionalization (NDP concentration in Halifax and Tory in the northeast), Nova Scotians probably would not expect a majority with such a votes distribution.

    And it is true that the s-v equation contains a weakness in that it considers only national fractionalization of the vote, and not regionalization. However, I would note that in fact, few FPTP systems have an effective number of elective parties (N-v) at 3.0 or above without having considerable regionalization. A party contributes more to N-v if it presents candidates and gains votes throughout a jurisdiction, but it probably exists as a major party only if it has some regional strongholds. That is, it’s unlikely to remain a force for long if it can’t at least win some seats and come close in others.

    In other words, regionalization is almost assured whenever N-v is high in comparison to the supposed “norm” of 2-something as the effective number of parties.

    It is not unusual to have the sums of “expected” seat shares add to more than 100%, given the error introduced by putting the effective number of elective parties (which is just a summary of the overall fragmentation) into the equation. One can adjust the expected shares by weighting the “overage” by those shares. That is, if party A is “expected” to have 54% of the seats, but the total adds to 102%, debit its expectation by (.02 x .54)=.01, or in other words, it is now expected to have 53%. (As you can see, it is hardly worth the effort; 102% isn’t worth being troubled over!)

  4. I would say that regionalization of party support makes a move to PR less likely, precisely for the reason Rici alludes to when he says, “it is hard to see an NDP caucus of whom 65% represent Halifax favouring a proportional system given that such a system would cause quite a number of them to lose their seats (the compensating seats which would be gained elsewhere are not members of caucus).”

    In other words, regionally concentrated parties actually benefit from FPTP–or at least their incumbent members (pretty much by definition) do. If a party is really concentrated–especially if it does not even try outside its strongholds–it may be over-represented by FPTP even though is “small” jurisdiction-wide. The Bloc Quebecois is, of course, an excellent example of such a party, as are several Indian state-based parties.

  5. I agree that “regionalization is almost assured whenever N-v is high in comparison to the supposed “norm” of 2-something as the effective number of parties.”

    But doesn’t that cast some doubt on the value of the s-v equation? It attempts to model the influence of N-v, but then it is almost certainly wrong in the case where N-v is much higher than 2 (in the case of Nova Scotia, it’s 3.01).

    So perhaps what is needed is some sort of metric of regionalization and a model which attempts to correlate that with N-v. Unfortunately, plurality systems with high N-v are comparatively rare, so there are not a lot of data points.

    As you say, a high degree of regionalization probably militates in favour of plurality systems. So a corrected s-v equation which somehow took regionalization into account would (presumably) show this election as less anomalous, and therefore less likely to encourage political parties to propose electoral reform.

    On the other hand, if my analysis above is correct, PR would be a good thing for Nova Scotia. (But then I would think that, given that I’ve always been a supporter of PR.)

  6. The National Party of Australia also confirms the thesis. They only contest (or win) rural seats in our IRV lower chambers while they’re almost nonexistent in the PRSTV upper houses in NSW and the Commonwealth.

    They invaribaly enter conservative Coalition Governments as a junior partner. They did manage to control the state of Queensland for an extended period, but there’s no prospect of that recurring.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.