## Photo finishes: Indi

This post will be progressively updated to follow the late counting in the undecided seat of Indi.

*Wednesday 4pm.* Another 69 absents, 56 provisionals, 87 pre-polls and 70 postals have trickled, making little difference to a Cathy McGowan lead that currently stands at 395. The bigger news is that Sophie Mirabella has conceded defeat.

*Tuesday 6pm.* A variety of additions to the count today – 132 absents, 116 pre-polls and even 132 provisionals – which have chipped 18 votes from McGowan’s lead, now at 387. Mirabella still needs the dwindling stock of outstanding votes, now at around 1750, to break 60-40 her way.

*Monday 4pm.* Another 1611 pre-polls have favoured Mirabella 899-712, but they’ve been partly negated by 577 absents which continue to favour McGowan, this time by 327-250. Mirabella’s net rate of gain is less than what she needs, putting her on course for defeat by about 400 votes.

*Sunday 7pm.* Pre-polls are continuing to heavily favour Mirabella, a batch of 1431 breaking 842-589 her way. This is a ratio of 59-41 in a situation where 57-43 should be sufficient to win her the seat. However, the remaining votes include at least 1000 absent as well as 2000 pre-poll votes, and those have been favouring McGowan. A further 971 of the dwindling stock of postal votes was slightly below par for Mirabella in breaking only 520-421 her way.

*Saturday 8pm.* Along with some rechecking, just 475 absents added to the count today and they’re continuing to favour McGowan, breaking 252-223 her way and putting her lead at 837. My projection of the share of the outstanding vote Mirabella will need is up from 56.3% to 57.1%.

*Friday 6pm.* The first 930 absent votes have been added and are good news for Cathy McGowan, adding 98 to her lead. Postals continue to flow solidly to Mirabella, a further 1918 boosting her by 290 votes, but there can’t be too many of those left. Her overall deficit is 897 and with about 7000 votes to come, she needs something of a miracle from here.

*Friday 3pm.* Another 1918 postals have been added and they’ve favoured Mirabella by 290, with McGowan’s lead down to 791. Yesterday’s projection that Mirabella needed about 55% of the outstanding basically still holds. She’s doing slightly better than that on postals, but absents and pre-polls might behave very differently and single one of either has been counted yet.

*Thursday evening.* The addition of 1930 postal votes has clawed back 318 for Sophie Mirabella after the identification of yesterday’s polling booth error blew out Cathy McGowan’s lead from a bit under 500 to 1471. Still to come are about 2750 postals, 4500 pre-polls, and 2500 absents, which will need to break at least 55-45 Sophie Mirabella’s way. Her share of postal votes so far is 57.6%, but absents and perhaps also pre-polls will presumably be less unlike ordinary votes.

*Wednesday 4pm.* A sensational development today with the emergence of a counting error at the Wangaratta pre-poll vote centre which was costing Cathy McGowan a bundle of 1000 votes. As David Barry notes in comments, McGowan had hitherto been doing 15% worse in the Wangaratta pre-poll booth than in the ordinary booth vote compared with a more typical 7% at other pre-poll centres, a discrepancy which disappears if 1000 votes are added to McGowan counts. Where McGowan’s lead had dramatically scaled back to 498 votes on latest counting, it can now be regarded as out to 1498, which almost certainly decides the result.

Monday 5pm. All the fixed polling both two-candidate counts are in, giving McGowan a solid looking lead of 24,840 to 23,086. Better modelling of my own in comments variously gives her 50.5% and 50.3% after pre-poll voting centres are added. However, there remains to come the very large number of postals, pre-polls and absents which seem likely to rein that in, the question being by how much. So a fair bit of entertainment left to come, it would appear.

Monday 3.30pm. The AEC is conducting a two-candidate preferred preference count between Mirabella and McGowan, going through the booths in alphabetical order. Since the largest centres in the electorate start with a “W” (Wangaratta and Wodonga), larger towns are unrepresented in the sample so far, which has Mirabella leading McGowan 14850 to 14569. I have attempted to model the relationship between the preference split and a given booth’s combined Labor and Greens preference flow, without a huge amount of success – the relationship is highly significant, but it only explains 24.4% of the variation. For what it’s worth though, the resulting model (preferences to McGowan equal 0.73*0.331x, where x equals the combined Labor and Greens vote) gives McGowan a 690 vote lead, remembering that this doesn’t include postals, pre-poll and absent votes.

Election night. Antony Green has Sophie Mirabella leading Cathy McGowan 29567 to 27880, which is presumably a projection because the AEC has no preference count and will presumably do one today or tomorrow. The primary votes are 44.4% for Mirabella to 32.2% for independent Cathy McGowan, which will be hard to overhaul with the available preferences from Labor (11.3%), the Greens (3.2%) and 9% others.

Mirabella should benefit from postals and have a high enough primary vote to just fall over the line

Very hard for McGowan. Currently needs 76% of preferences and a substantial minority don’t come from ALP/Green, which makes it so much harder. Indeed she’s bombed the Greens into micro status, they won’t even get their deposit back.

Maybe doable on this asking rate if the voters really hate Mirabella but very likely postals will make it harder.

http://kevinbonham.blogspot.com.au/2013/09/2013-federal-election-late-counting.html

1. Mirabella is gone in the long run, whatever happens. There’s been a high level of discontent with her amongst the local party faithful for years – they’re used to local members who actually show some interest in the electorate – and there’s been talk of de-selecting her in the past. This will give them the ammunition they need to get rid of her (she’ll probably retire to spend more time with her family…)

2. Last election, Labor (i.e. me) picked up 1.5% after the postals and absentees were counted. If that happens this time, it’s enough to get McGowan across the line. On present numbers, however, I’d be tipping Sophie’s back.

Re 2. that reminded me I should have checked the post-count swing in Indi. From the total of ordinary votes to the final total the Coalition’s 2PP improved by 0.32%. Non-ordinary votes include out-of-electorate prepolls as well as postals, absents and provisionals.

TCP figures trickling in, I make it a 78% preference flow to McGowan so far – this is very close.

Not much of a relationship between McGowan’s primary vote percentage and the rate of preference flow to her.

I have set-up a similar regression model for Indi to the one I made for the close count in Denison at the 2010 election. Whilst the model is not yet complete, it is currently predicting a final 2PP number for Cathy McGowan of 50.51%. The standard error, however, is currently an unhelpful 1.37%.

McGowan is going to very badly lose the Benalla PPVC and Wangaratta PPVC booths which are large booths. A number of PPVS booths have not reported yet and while McGowan is doing well in some she is doing very badly in others. They could decide the result. Still a long way to go for McGowan to win.

I just had a look at the booths remaining with 70 booths included. At that time McGowan was behind but just looking at the primaries they projected her to the lead, which I see she has since taken.

They’re going so fast there’s not much point me modelling it – I was out on work commitments for a few hours. Looks like McGowan will be in a position after ordinary votes that might be just holdable.

Sophie looks home to me.

281 votes in front on 2PP from 92/103 booths

Sophie now behind by ~1,700 votes. She will win from here.

*whoops Cathy I meant.

my model predicts (McGowan 2PP):

Benalla ppvc (45.71%)

Div Office prepoll (51.56%)

Harrietville ppvc (41.17%)

Merrijig ppvc (37.37%)

Mount beauty ppvc (55.96%)

Wangaratta (43.60%)

Wodonga (51.55%)

sykesie

Not sure what the underlining assumptions are but Wodonga and Wang seem a bit low so if able too what are you basing that on.

Mexicanbeemer; simply basing it on a regression analysis of a relatively large number of booths already reporting.

3 variable model.

Model says:

McGowan 2PP% = 57.59 + 0.45*(McGowan primary) -0.60*(Mirabella primary) + 0.47*(ALP primary)

The model R2 is 0.98.

Errors in the coefficents are:

57.59 (+- 7.50)

0.45 (+- 0.08)

0.60 (+- 0.08)

0.47 (+- 0.11)

Thanks Sykesie

Sykesie is McGowan still tracking for 50.5 or so?

50.3 after pre-polls from my regression which is less sophisticated than sykesie’s (just uses Mirabella primary). Judging by the 2010 absentee and postals, I don’t think it’ll quite be enough for McGowan to hold on.

Are they doing the PPVC booths today?

Kevin – just got back from giving a 4pm spectroscopy lecture. I’m about to input a whole heap more data and we will see what happens to the model. It seems fairly robust at this stage.

My info is that postal vote counting will begin tomorrow.

OK, so the model has now been updated and predicts the following:

1) For the 92/103 booths the model predicted a final McGowan 2PP vote of 50.66% (the actual one according to the AEC is 51.83%). So the first point is that my model seems to UNDERPREDICT McGowan’s 2PP vote. This may be due to booths where the green vote is higher (green primary is not included in the model).

2) If I take the primaries of the 7 pre-poll booths that do not currently have 2PP counts and project a 2PP count based on the model, I get Mirabella pulling back the lead by a further 863 votes.

3) The AEC currently has Mirabella behind by 1754 votes.

If we apply the model change and subtract the 863 votes McGowan is predicted to drop for the pre-poll votes, then McGowan’s lead drops to 891. However since the model under-predicts McGowan’s vote, it may be a little better than that for McGowan.

4) So after the 7 pre-poll booth 2PP votes are added in, Mirabella will need to pick up 800 or so votes from the remaining (postals??).

5) I reckon that means Mirabella would need to win the 2PP for the absentees by about 55-45. It might be worse than this because of the model under-prediction bias. Doable? Sounds pretty tight to me …

Mirabella is definitely in for the fight of her life, but I can’t rule out that she might overturn it on very favourable postals. I would suggest she is on the precipice.

The Sports guy seems quite reasonable.

He’s copping a pasting from Steve Price and taking it very well.

I’m going to have a look (when I get home from chess) at Mirabella’s performance at PPVCs compared to ordinary booths last time and this to see if it might be a fair indicator of whether she is likely to do better than normal at out of electorate PPVCs.

Dio# 23

Yeah but what did the detectives say; when did he walk into the room; etc etc etc

KB #24 I had also been thinking about Mirabella’s relative performance with ordinary votes, prep old and postal votes at previous elections. I’ve had a look back at the 2004, 2007 and 2010 results. The thing that stands out is that, while she did significantly better on postal votes than in ordinary votes, this was offset by a poorer performance in absent and declaration votes. At two of the three previous elections, Mirabella only performed marginally better on prepoll votes than ordinary votes on polling day.

In all three previous elections, Mirabella’s overall two candidate preferred result improved on the ordinary vote by less than McGowan’s current lead in percentage terms.

2004 ordinary vote: 66.13% final vote: 66.29%

2007 ordinary: 58.79% final: 59.19%

2010 ordinary: 59.62% final: 59.94%

My final observation for the moment is that the two candidate preferred figure the AEC has published does not seem to include all the ordinary votes – 47,926 counted of 73,313. I can’t identify what is missing.

Prep old = prepoll

I did a very quick, very rough cout and there are 23000 PPVC votes that have been counted and showing first preferences but are not 2PP counted yet. Is that the missing votes you are talking about?

I forgot to make the further point that for some reason the AEC Indi Divisional Office has not provided the declaration vote details so we can see how many prepoll and postal votes there were. I thought I saw this over the weekend, but perhaps I’m thinking of another electorate?

Yes there are a lot of counted PPVC primaries that are not yet included in the 2PP.

I am a bit cautious re including pre-2010 figures on post-count performance because of the changes that came in when PPVC votes became ordinary votes that were counted on the night.

Can say that Mirabella’s net post-count gain of 0.32% 2PP in 2010 was almost right on the national average (0.28.) Doesn’t seem to be an electorate that lends itself to a huge rate of post-polling catchup.

McGowan is doing better than I expected on the preference flow; the voters for other parties must be totally fed up with Mirabella.

Edi_Mahin #28 that may be it. Last election there were 71,000 ‘ordinary’ votes and 4,400 ‘Early Vote (pre poll)’. Perhaps the PPVC votes are displayed as ‘ordinary’.

Postals issued were about 6,600. This is smaller than I would have expected. If it is true, the gain for Mirabella, based on usual differences between Postal and Ordinary, won’t be large

The pre-poll primaries vote count (it is hard to find- look in the CSV download page) is currently about 22,000. Mirabella is polling 45.7% there, compared to 44.0% in the booths. If the preference flow pattern remains the same, this would probably mean Mirabell could creep up by about 400 votes in the TCP, which would not be enough.

About 4,400 pre-polls counted so far this morning and Cathy’s lead has increased by 1 vote. The first 2,000 postals are also being counted this morning. Perhaps they are in the totals already. Unless the remaining pre-polls and postals are weirdly different, I think we can say Cathy’s win is just about locked in.

Modelling 2PP to McGowan in each booth as a linear function (zero intercept) of 1st preference percentages for: McGowan, ALP, Greens and other (other is a sum of all 1st prefs for candidates other than McGowan, Mirrabella, ALP and Greens) gives coefficients of: McGowan 0.996, ALP 0.928, Green 1.033, Other 0.518. So there are strong 2PP preference flows from ALP and Greens, but roughly 50:50 from the other candidates. R^2 for that model is 0.999 (adjusted R^2 is 0.988).

Using that model to predict 2PP in the major PPVCs from 1st preferences, McGowan is predicted to win the Wodonga and Mt Beauty PPVCs and the Divisional Office prepoll, but lose the other PPVCs. The net prediction is McGowan to be ahead by 600.

Since building this model, the following PPVCs have results for 2PP: Divisional Office (PREPOLL), Harrietville PPVC, Merrijig PPVC, Mount Beauty PPVC. Across those 4, McGowan has 4 fewer 2PP votes than predicted by the model (out of a total of 4457 votes). Results of the bigger PPVCs for 2PP are yet be reported.

Re: Geoff Lambert comment 33. Results for the biggest PPVCs are not in yet. Mirrabella has a large 1st preference count in Wang and Benalla (51.5% and 49.7& of 1st preferences) which are two of the three largest PPVCs, so things might be different (but perhaps not different enough for Mirabella).

Now 101 of 103 booths counted and McGowan 676 votes ahead. Of the final 2, 1 is the Wodonga PPVC:

http://vtr.aec.gov.au/HousePollingPlaceFirstPrefs-17496-31542.htm

Doesn’t look like a bad booth for McGowan so she should (I guess) extend her lead slightly.

Results for Wang and Benalla PPVCs are now in. McGowan did lose those, but not by as much as my model predicted – across the pre-polls so far in, McGowan has got 96 more 2PP votes than my model predicted. Wodonga is the other large PPVC without 2PP results so far, where my model predicts a small win for McGowan.

And the Blind/Low Vision (BLV) centre will presumably be low numbers and may in fact have already been incorporated into the Divisional Office count.

OK, just had a call from one of my guys at the count in Wangaratta, reporting that McGowan ahead by over 1,000, with postals about to be counted.

He also reports overhearing Mirabella supporters conceding defeat.

Amazing performance by McGowan on preferences; I thought the small but substantial preference of minor right parties in the preference flow would probably make it too hard.

I looked at PPVC/ordinary comparisons that might assist in projecting the out of electorate prepolls but there were no serpents in the water there; if anything they weakened Mirabella’s chances.

I’d like to see some postals first but this is looking very very bad for the incumbent.

Praise the Lord!!! (and I’m an atheist).

McGowan has tweeted that she’s 1049 votes ahead before they start counting postals this afternoon.

There are 8844 postal votes received so far (it’s a bit frustrating that my comments are still awaiting moderation – things are moving fast!), so Mirabella needs ~56% of those on 2PP to win. Last election she got about 65%, so the swing of around 10% if realized in postal votes will make it very tight.

Prepoll votes all now counted, assuming nil from the blind/low vision centre. AEC shows McGowan 1051 ahead.

There were 15,275 absent, early vote and postal votes issued. Leaving aside declaration votes, very few of which will be counted, Mirabella needs to get 8164 of the 15,275 to win. That’s only 53.45%.

At the last election, Mirabella got less of the 2CP absent vote than she did of the ordinary vote. She got more of the early vote and almost 6% more 2CP from postal votes compared to ordinary votes. If the 2CP swing against McGowan holds into the 15,275 remaining votes, but they otherwise bear the same relativities to ordinary votes as in 2010, I calculate Mirabella will win by 21 votes.

To get to this, the benchmark I’d worked out above is 55.15% of postal votes, 52.1% of early votes and 41.59% of absent votes.

Right now I’d be calling for my party’s best scrutineers and thinking about grounds for disputing the result.

Sorry ‘declaration’ should be ‘provisional’.

McGowan has 50.72% 2PP at what I take to be the close of ordinaries.

In 2010 the Coalition managed a post-count swing exceeding .72 in twelve electorates nationwide. Five were Victorian: Hotham, Holt, Flinders, Dunkley and Melbourne Ports (the latter because of the high Jewish vote).

Kevin this time round there are a lot more postal votes – 10,465 compared to 6,388 last election.

canberra boy is right – the breakdown of remaining votes as postals/absents/EVPP is radically different to in 2010 and that changes things considerably and makes it much more likely Mirabella can pull this back. It’s not also the number of non-ordinaries but the different pattern of non-ordinaries that is crucial. So this is very unclear.

But if that is the case in Indi how many other seats is it also the case for? Are we going to see massive late Coalition pullbacks all over the place in post-counting?

Is it possible to find information on what portion of postal vote envelopes issued were returned last time?

Mirabella is still going to need Very favourable preferences on the remainder. Vastly different to what we have seen so far. Not impossible based on the type of votes remaining, but still not that likely in my opinion.