Making adjustments

School’s back, people have returned from their holidays, and soon we might even see the odd politician on the telly. Tomorrow sees the start of September, and the silly season (and dearth of real opinion polls) is coming to an end. Last year, MRBI published a poll in the first week of September, and it will be interesting to see whether they do this again this weekend.

This will be the first foray from MRBI since they announced that they’d be changing how they adjust their base figures. All companies do this to account for the differences in turnout of different party supporters, and, more importantly, the fact that “Don’t Knows” never split in the same proportion as those voters who, er, “Do Know”.

The MRBI method has been remarkably accurate over the years, and wouldn’t be amended now, if it weren’t for two things. Firstly, since Labour has taken off into the stratosphere (and FF sunk to a lower, darker, and hotter place), the degree to which this has happened has been greater in MRBI than in RedC, the other company with a reputation to maintain in this field. Given even the lesser swings have been dramatic enough, party hacks and loyalists antagonistic to this have derided the more spectacular change recorded in MRBI – interestingly in greater numbers among the FG supporters, on politics.ie at any rate. If Labour did get to within a couple of percentage points of FG on polling day, greater transfers from SF and OTHs would, most likely, see them eclipse FG as the first party win more seats than FF in a GE (well, since the 1920s).

When, at the turn of the year, Labour were at 24-25% in MRBI, there was a storm of people saying that the figures were nonsense, and RedC were the only one to believe, giving Gilmore’s crew 17%. By June of this year, there was a brief period where it looked like they had converged, with RedC moving LP up to the mid-20s, but then MRBI jumped up to 32% (29% before adjusted). RedC still trails this level of LP support, but at 27%, it still is well up on the MRBI figures previously cited as unreliable.

The other reason, and it at least depends on more than an exasperated cry of “that can’t be right!”, is that the adjustment ,while appropriate when FF were on 40% and LP on 10%, is no longer useful when LP are ahead by up to 10% on the raw figures (this argument at least doesn’t suggest that the pollsters are just asking the wrong people). Evidence cited for this is Labour’s (relatively) poor performance in the 2007 Local/Euro elections, polling 14-15%, compared to 23% in MRBI, and 18% in RedC.

Of course, non-Dail elections always return quite different results to opinion polls, as people see it as quite a different question. In the locals, many of the FF-to-LP swing voters stuck with the local man, or went for the Indo. In the Euros, LP got as many votes as it needed to elect it’s 3 serious candidates, with many other LP supporters switching to other candidates, particularly in Dublin, where De Rossa was safe, but they wanted to see FF lose their only seat. Certainly, I was tempted. There were others who voted for Joe Higgins, SF, or even Patricia McKenna in protest at Labour’s support for Lisbon – a big issue in the Euro-campaign, and something up to 40% of LP voters still had misgivings about. In a General Election, you don’t really have that luxury. You’re electing a Government, and in the current climate, it seems unlikely to me that many will be wasting their vote in an election as important as the next one, particularly if a handful of seats could be the difference between Kenny and Gilmore being Taoiseach. This point in particular is a narrative that FG strategists are not keen to see take wings.

The truth is that we’ll not know if LP can translate their opinion polling into reality until a GE is actually held. But it will be interesting to see what difference, if anything, the new MRBI adjustment will make to the differential with RedC. My own view is that it’s been a red herring, and the two companies results are closer than people suggest. Red C polls this year have averaged FF at 25%, FG at 33%, LP at 21%, SF at 8.5% and GP at 4.7%. MRBI in contrast have shown FF at 20%, FG at 31%, LP at 27%, SF 8.3% and GP on 3%. For all the hype, there’s not such a big difference, and if in reality FF were at 22-3%, FG 32%, LP 24%, SF on 8-9% and GP on 3-4%, the variances remain within the margin of error. The big issue is that, if LP are even marginally ahead of FG in one set of polls, the choice being between Kenny and Gilmore for Taoiseach will become a more accepted one, and that’s likely to move more voters into the labour column than FG.

We’ll not know the effect of the new adjustment until we have a post-summer poll from both organisations – a dip in support for LP in the MRBI could be based on the adjustment, but could also be based on them dipping from a previous high (quite likely during a Dail recess, given how much of their rise is apparently due to Gilmore’s performances), and a rise in FF could be a result of no politics as usual during the Summer months and disaffected voters drifting back home. My own view is that the change in that methodology won’t make a big difference, the big changes will come from the general drift – if LP were 29% on the raw figures last time, there’s little reason to believe that their likely performance among undecideds wouldn’t have pushed them over the 30% barrier, whatever the method used. Setting myself up for a fall, I think the next MRBI will show FF at around 21%, FG at 32%, and Labour on 25%. RedC may see FF about 25% and Labour around 23%, with FG no different to MRBI.

Still, we’ll see eh?

Labour poll in Dun Laoghaire

I’ve heard from a pretty good source that Labour ran a poll recently in Dun Laoghaire.  Only figures I heard were FF, which were 7% Andrews and 6% Hanafin (whatever the sample size that’s well within the margin of error, but good-ish news for Andrews, given the 2007 result).  If anyone reading has had sight of the overall results, please share!

13% for FF is pretty much the territory my spreadsheet was putting them, which is interesting (I had suspected that the swing away might be slightly less, given the social profile of the constituency).  It would leave them fighting for the last seat, but depending on a lot going right for them to avoid falling short.  A 50/50 split like that isn’t great for them, the 13% would be expected to drop to about 11.5% after transfer leakage (if lucky).  If the GP were about 4-5% (again hopeful), they could expect to be brought to about 13% on their transfers, and they could expect about 1.5% in SF transfers, pushing them to 14.5%.  Poor vote management by Gilmore could see them pick up about 2% in leaked transfers from there.  If that put them ahead of the second FG candidate (which it probably would),  Sean Barrett would be elected on other FG transfers, and there’d be a good 5-7% surplus at least going, and 2-3% of those going to FF would probably get them scrape the last seat (without reaching the quota).

The effect of this poll if true (and I assume if I’ve heard it, so has Hanafin) will be interesting.  This is a hard constituency to call.  Many would consider it FG heartland, and yet in 2002-2007 they held none of the 5 seats.  It’s had 2 left TDs before, at a time when the left vote nationally was significantly lower than current polls suggest, and some would suggest that with Cuffe, there were two of the current incumbents elected on a broad left platform.  PBP’s Richie Boyd Barrett makes it all even harder to call, given the irrelevance of the “Others” national total to many micro-parties’ fortunes.  Add to that 3 Ministers, all fighting for their political lives, including one who is considered a potential (if long-shot) chance of Taoiseach if there’s an anti-Cowen heave and Lenihen isn’t keen, add the loss of a seat in the revisions…

Very hard one to call, so it would be nice if anyone out there could share the full figures…..

poll of readers

June2010 RedC

Previously posted by me on politics.ie in June, following a Red C Poll with the following figures;

FF 24% NC.
FG 33% +3.
LAB 27% +5.
Greens 2% -3
SF 8% -2
Ind 6% -3.

—————————-

Welllll……

One of the reasons I was impatient for this poll was that I’d come up with a second (alternative) base for the swings, and was going to run both of them for the first time on an actual poll. The new base takes the ’07 figures, but adjusts for (a) regional swings, and, (b) some adjustment of the method of the base to account for a different LP swing where they come from a higher or lower base, which effectively ‘flattens’ a bit the uniform swing (using a swing of what % are NOT voting LP). In dry runs its shown a more similar pattern, on a constituency basis, than uniform swing, compared to the changed distribution in ’92 over ’89, which was the last time LP gained so much in a single GE.

In theory it should make very little difference to the overall number of TDs each part gets (and it doesn’t this time, re-assuringly enough), but gives a better idea where those gains will be.

Now ….. great poll for LP (best ever Red C and up 10% on what they appeared to be stuck on until recently – itself over two-thirds up on the last GE). Very high preference for Gilmore as Taoiseach (not just ‘approval’ rating’) means they’ll do well on Dks and also on transfers.

FG, while they’ll be relieved, are pretty much what I predicted – the Oscar Wilde maxim, and Kenny finally looking like a leader (although, as I said before, I think he’s not bad behind the scenes – I did say they’d regret it if they replaced him with RB). If they can knit a decent FB together, they are probably over the worst of it -a decent poll like this means that if there are any who still resent him (well there are I’m sure) they are utterly outflanked. The Shinners continue to pop between 6-10%, and 8% I have to say looks about right to me.

FF will shrug and say “nothing to see here” and effectively they are right. But the GP will be shocked. They’ve been rock solid at 5-6% in all RedC polls forever now. Could be a once off, given the margin of error, but still, they’ll be nervous in the extreme until another poll puts them in 4-5% territory. Certainly, they’ll not be pulling the plug any time soon.

Finally, of interest is a very poor showing by “Others”. Again, hard to say why, but maybe the talk of Gilmore as a possible Taoiseach is drawing votes from ‘protest’ candidates, where they may be seen as wasted, to the LP column.

So what are the spreadsheets saying? First figure is uniform swing from 07, second is swing from the adjusted figures based on regional swings etc, you can choose which suits your prejudices, but both are very close.

FG 62 or 56
LP 52 or 58
FF 43 or 43
SF 5 or 7
GP 0 or, eh, 0
OTH 4 or 2

As you can see, there’s a net 6 seat gain for LP over FG based on adjusting for regional swings and ‘flattening’ the swing, which would nudge LP ahead (although the number of seats going back and forward between the two methods is much more). I prefer to keep with the old method, which is the uniform swing method, as it is ‘purer’ (and therefore less open to human error) but there’s certainly much for LP to consider here. Even in the latter figures, there are 8-10 seats where a less pessimistic transfer rate would see them pull in seats (they don’t include Roscommon in either, for example, despite the addition fo a high profile candidate with a good electoral record). There are, however 62 seats they would win in either one or the other method.

So what would the implications be if something like this panned out? Well, the pretty obvious one is that FG and LP are neck and neck for seats, (FG slightly ahead) with FF a poor 3rd. If this became evident over the course of the election, one could see FF hit even harder, as the choice would no longer be between them and FG, but FG and LP, and one assumes that a number of those who want Gilmore to be Taoiseach are currently in the FF column). Another thing is that, on these figures, a left-led govt is not possible (unlike the MRBI one), given the more modest SF performance, and collapse of the GP and OTH vote.

Of course, it’s just one poll. What matters is the trend. Not long ago, we were being told that MRBI polls were ‘meaningless’ because they gave LP 24% after the adjustment, and Red C, giving LP 17%, were spot on. Now Red C are saying 27% (MRBI 32%) and the argument is how close they and FG are, and who is nudging ahead of who, with FF (currently) being in third place a foregone conclusion.

Right. Make of that what you will. I’m off to put the kids to bed, open a bottle of a Sicilian Red, and watch the second half of the Footy. See you when I’ve a few glasses under my belt!

———————-

NB – Since then I’ve honed the adjusted spreadsheet a bit more, and the overall results come out even closer to the ‘brute force’ uniform swing method while at the same time coming a lot closer to plausible local results, so I’ll be using that most of the time going forward.  Dotski 22/08/10

Welcome to Irish Polling Report

Dotski here.  If you’ve found this blog, you’re probably a regular on politics.ie.  I’ve been known to post there a bit, particularly following Irish opinion polls, when I crunch the figures through my trusty spreadsheet.  While I’ll keep posting there, it occurred to me during the recent (2010) UK General Election that a dedicated site for discussion of polls (as opposed to slagging off each other’s parties), along the lines of the excellent UK Polling Report,  would be a good thing.  So rather than wait for someone else to do it, I thought I would myself.

While I’ll be running the spreadsheet simulations to calculate the number of seats each party is likely to win based on each reputable poll, I’ll also be interested in any ‘internal polls’ that people get their hands on, so if any readers get their hands on these, please let me know, and we can consider the implications they’d have in that constituency if true, and what sort of national figures they would be consistent with.   While many of these are notoriously unreliable, so is 99% of the internet, and unlike most of that they can be quite fun!  And genuine ones, in fairness, often are interesting in that they can indicate what parties are likely to do in terms of candidate strategy, particularly when national polls are indicating figures very different to any previous General Election (I write this in August 2010).

So what are the rules for anyone wishing to post?  Well, no political point scoring.  Politics.ie is a perfectly good enough a place for that.  The idea of this space is proper analysis of polls.  Also, play the ball, not the poster.  If you think someone has made a crazy extrapolation, say why, don’t say it’s because they are clinically insane (even if you know it’s true.  Actually, especially then).

I’ve no idea yet if this will be read by many (edit: 6 weeks on it is, which is much appreciated!), but I would hope that, particularly as we approach the next election, this will be a useful place for people trying to work out what will happen in what promises to be the most interesting – and unpredictable – Irish election since the 1940s.  You can post anonymously, but if you’re a politics.ie regular it’d be nice if you used that name here.  If you’d like to contact me, you can get me at irishpollingreport@gmail.com

While there’s an archive of posts on the right hand side, you might be interested in the following posts/threads;

My most recent constituency-by-constituency analysis based on national polls

Analysis of constituency polls in Dublin SE Donegal SW, Cork SC, Cork NW, and Dun Laoghaire

There also stuff in there specifically on the relative merits of the polling companies, notably the RedC/ MRBI and Landsowne discrepancy, and somewhat odd aspects of Quantum Research’s work.

And for those who just think the polls must be wrong, because there will always be ‘local factors’ and rural issues that make big changes impossible, you may find this interesting.

So, in summary, I hope you enjoy the posts, that anyone who can will share interesting analysis and information, and spread the word!

Dotski.