What was it that I said yesterday (in the immediately previous post) about not trusting exit polls in the political and security context of Palestine? What was it I said about the multi-seat plurality races being especially hard to estimate based on exit-poll sampling, even under the best of conditions for the polling itself?
From the Jerusalem Post:
Exit polls on the night of the vote gave the ruling Fatah Party a slight victory, a finding that was dramatically reversed on Thursday when Fatah and Hamas leaders said Hamas had won a clear majority of the 132-seat legislature.
The discrepancy may have been due to a reluctance by some voters to admit to pollsters that they were abandoning the ruling party. The polling errors appeared especially glaring in district races, where smaller numbers of voters were surveyed. [my emphasis]
Not only would a smaller sample be a problem, but so would the basic process of estimating from that sample, given that a typical pattern of multi-seat plurality systems (as I have noted in more detail in past posts) is that voters do not necessarily give all their ‘nominal’ votes to candidates of the same party or even use all of the nominal votes they are entitled to use.
The story goes on to say that Hamas apparently won a “large majority” in the district races, just as I suspected they would, because their voters would be more likely to vote the full slate than would voters of the internally divided Fatah.
In fact the scale of the Hamas win in the distrists vastly exceeds what I could have imagined. Another Post story says that Hamas won nearly all of them, and 75 seats overall. Yet another story I read (at Ynet) said that Hamas might have won 80 overall.
I have not yet seen updated votes totals, but remember–the vote estimates that were reported last night from exit polls might prove to be much closer to reality than the seat results, because the nominal (district) and PR-list seats are allocated completely separately (in ‘parallel’) in Palestine’s variant of a mixed-member system. A multi-seat district plurality system–Palestine’s nominal tier–can produce wide divergence between votes and seat shares, especially when one major party has a more cohesive electorate than the other. This is precisely what I have meant in past posts when referring to the votes-seats conversion probably benefiting Hamas relative to Fatah–and that was evident even in the estimates that the exit pollsters gave.
Moreover, this electoral system is a ticket-splitter’s dream. Not only do the list votes (which surely is what the exit poll was reporting) not have to match the candidate votes, but (reiterating here) within a district voters can split among candidates of different parties.
These factors compound not only exit-poll estimation, but also any systematic real-results relationship between votes and seats.
With a clear majority in the parliament, Hamas will now have to form the government. Palestine has a semi-presidential system with a relatively strong presidency, but requires that the cabinet have the confidence of the parliament. Already, the cabinet has resigned.
I will leave it to others more versed in the Middle East to assess the significance of this result.
Hi Prof. Shugart. Given what you’ve said about the electoral system and the nature of exit polling in that system, I curious to know whether you think exit polling is useful at all in this type of parallel system at all. Is there anyway to mitigate these factors, or is the nature of the system just unsuited for reliable results from exit polling?
I am also curious to know who (or which entities) commissioned polls. Sometimes US pollsters only have a cursory understanding of the electoral systems in other countries and don’t exactly ask the right questions. But given high profile of these elections, I would be surprised if unexperience pollsters were used.
Seed planted by Nina Merchant — 26 January 2006 @ 09:58
Nina raises good questions. I am not competent enough about sampling and extrapolating methodologies to be certain of the answers, but the Palestinian context had many factors (which I outline in the post) that would make proper estimation difficult. I recommend two posts from Charles Franklin on this topic: One from last night comparing the three separate exit polls (a luxury he notes we no longer have in the US), and a follow-up today.
My take on what we can generalize from this would be the following: Parallel system would tend to pose more challenges than a pure SSD plurality system, which in turn would be more difficult than MMP, which in turn is very slightly more difficult than pure PR. Don’t hold me to that precise ranking of difficulty, but I am reasonably confident that a parallel sysem is more challenging than the others just mentioned. I must admit that I know nothing about the record of accuracy of any exit polls in Japan since their parallel system went into effect, or other “simple” parallel systems–those with single-seat plurality districts and a separate PR tier (also found, in various forms in Russia and several other countries). If anyone out there knows about exit polls in these systems, please post!
A multi-seat plurality system would be even harder than any of the above. Well, Palestine combines both of those hard systems: multi-seat plurality and paralell allocation (with multiple votes). And then there are the problems of people not being willing to speak openly in face-to-face interviews, the lack of experience conducting polls there (because electoral democracy is new to Palestine), etc.
And it is not as though exit polls have an unblemished record even in the US, predicting a single national election result in a setting with lots of experience to build on and little fear (though perhaps growing reluctance) on the part of voters to speak to pollsters!
Seed planted by Professor Matthew Søberg Shugart — 26 January 2006 @ 10:20
Nina, most of the polls were conducted by Palestinian universities or public research agencies. For the most part, the Palestinian pollsters are experienced people with good reputations, and they have a decent track record at predicting the results of municipal elections. Whatever the errors in the polling process, they weren’t due to lack of experience or local knowledge.
Seed planted by Jonathan Edelstein — 26 January 2006 @ 11:03
Thanks for the info Prof. Shugart. Prof. Malesky was entertaining us the other night with stories about selection bias in US exit polls. Just for fun, the night of the election, he looked up the data for exit polling in Ohio and found that females, aged 18-30 were way overrepresented. Which, of course, we all know now, but it must have been fun to pinpoint it then. It seems that hiring male college students to do the interviews will do that.
I just brought up the pollster question becuse USAID often contracts with US-based democracy and governance development NGOs or for-profit development firms to commission public opinion and exit polling. The results from these can sometimes be mixed.
Seed planted by Nina Merchant — 26 January 2006 @ 13:05
“hiring male college students to do the interviews will” lead to “females, aged 18-30 [being] way overrepresented.”
That’s great! (Not for the accuracy of the results, however!)
Seed planted by Professor Matthew Søberg Shugart — 26 January 2006 @ 13:36
We should be grateful to Palestine for giving the world a lesson in the perils of the parallel system.
Hamas just got 45.35% of the vote, and should not have had a majority.
For the first real Palestinian election, all parties other than Fatah had asked, of course, for proportional representation. Fatah, of course, being the dominant party, wanted to keep FPTP. So they compromised on a half-proportional system. “Parallel” like Japan. Half the seats by PR lists, and half by single seats. But unlike Germany and the other MMP jurisdictions, the list seats were not compensatory.
The idea was, Fatah would sweep the single seats as usual, enjoying the “winners’ bonus,” but letting the smaller parties have a taste of proportionality.
But Hamas swept the local seats, and got the “winners’ bonus” that Fatah had expected.
Hamas got 30 of the 66 list seats, but 46 of the 66 local seats, total 76 of 132..
http://www.elections.ps/pdf/result_seat_distribution_EN.pdf
Fatah got 42.08% of the vote, and 27 of the 66 list seats, plus 16 of the 66 local seats, total 43.
Four local seats went to independents, nine list seats to smaller parties.
If they had MMP, the results would have been:
Hamas: 58 (46 local, 12 top-up list)
Fatah: 54 (16 local, 38 top-up list)
PFLP: 6 list
Badil (“The Alternative”): 4 list
“Independent Palestine” (left wing of Fatah): 3 list
The Third Way: 3 list
Local independents: 4
No doubt Fatah would have led a coalition of those willing to recognize and negotiate with Israel.
What an electoral tragedy.
Seed planted by Wilfred Day — 26 January 2006 @ 17:08
Clarification: The local seats (nominal tier) in Palestine are not FPTP. They are multi-seat districts, averaging about 4 seats each. This would tend to increase disproportionality relative to FPTP–if most of the leading party’s voters gave used all their votes on candidates of that party. We will need detailed data to know, but I suspect most did, hence the large seat bonus.
An additional caveat: Is the 45.4-42.1 vote split party-list vote? Presumably. It is possible that the candidate votes were somewhat different. Still, a large majority of seats on 45% of the vote sounds about right, given the rules, which I agree were a very bad choice. Bad both for being parallel, as you note, and worse still for having a multi-seat plurality (block vote) nominal tier instead of FPTP.
Seed planted by Professor Matthew Søberg Shugart — 26 January 2006 @ 18:24
Of course the clarification is correct (with the minor quibble that 3 of the local seats are indeed single-seat districts). My Canadian bias crept in here; from time to time in our electoral history we have had some multi-seat districts, within the context of the FPTP system, so I think of them as part of the genus FPTP (if you will excuse the botanical analogy.) But that’s no excuse for writing “single seats” when I meant “local seats.”
Seed planted by Wilfred Day — 26 January 2006 @ 19:45
Hamas actually got almost exactly 42 percent of the nationwide vote to Fatah’s 38.98 percent. The results table on the CEC site only lists the parties that made the proportional representation threshold, so it doesn’t reflect the total number of votes cast (which was 1,035,164).
But I agree with the basic premise of Wilfred Day’s comment – Fatah outsmarted itself big time.
Seed planted by Jonathan Edelstein — 26 January 2006 @ 19:56
I’m assuming the nominal tier was block vote?
This is interesting: rather analogous to Taiwan’s last two legislative elections that produced widely divergent results in terms of legislators elected for each party/bloc with rather similar votes for each party/bloc; but I’d guess there’s a lot more room for coordination failure for poorly organized parties if there are multiple ballots.
Seed planted by kao_hsien_chih — 27 January 2006 @ 05:10
[...] The following started as a further comment to propagate some ideas already planted beneath the post on Palestinian exit polls. [...] It is specifically in response to the point raised by Kao Hsien Chih [...] regarding coordination failure. [...]
Scion grafted by Fruits and Votes » Blog Archive » The genus, Plurality — 27 January 2006 @ 08:15