Saturday, July 4, 2020

Eden-Monaro Late Live Comments And Post-Count

Eden-Monaro (ALP 0.85%) McBain (ALP) retains with 50.41% 2PP (-0.44)

Donations welcome!

If you find my coverage useful please consider donating to support the large amount of time I spend working on this site.  Donations can be made by the Paypal button in the sidebar or email me via the address in my profile for my account details.  Please only donate if you are sure you can afford to do so.


Some people are speculating about a recount.  A recount is automatic if the final margin lands below 100 votes, but that is now very unlikely.  For margins above this, recount calls are typically rejected.

Nationals And Shooters Preference Factors

There is a lot of speculation about whether Nationals preferences have caused Kotvojs to lose.  Estimates of the Nationals to Liberals preference flow have varied from 70% to nearly 90%.  Scrutineering of this flow across a large electorate is extremely difficult; it's best to wait until the AEC comes out with the final preference flow.

On figures current as of Tuesday morning, McBain leads by 775 and there are 5924 Nationals votes.  This means that unless the flow from the Nationals to Kotvojs exceeds 93.5%, Kotvojs would be now leading had she got every Nationals preference.  But that's not interesting, because 100% preference flows never occur and the Nationals running is not the cause of the "leakers" preferencing Labor (most would have done so anyway).  What is interesting is whether the flow has weakened since 2019 and done so by enough that the weakening of the flow becomes the cause of Kotvojs' defeat.  At the moment this is the case if the flow from the Nationals has shifted to below 80.7%.

Another question is whether the Shooters preferences have caused the outcome.  Estimates of the Shooters flow vary with some claims as high as two to one to Labor, but Liberal scrutineers have said 50-50.  As of Tuesday morning, for Shooters preferences to have alone determined the outcome, they would have to break 57.8% to Labor.  However, the flow of Shooters preferences to Labor will be boosted by the donkey vote, meaning that if even, say, 55% of genuine Shooters voters have preferenced Labor then that will have decided the seat.  For the Shooters how to vote card to have decided the seat, it would currently have to be the case that the flow from the Shooters (whatever it is) would have been 15.7 points weaker had the Shooters preferenced the other way.  Limited evidence does not support Shooters how to vote cards making that much difference.  This number may drop with further counting, but I'm not sure that it will.

A further possibility is that some combination of Shooters preferences and weakening in the Nationals flow could account for the result, with neither factor decisive by itself.

I will analyse preference flows fully when they are available.

Count Updates - comments scroll to top below the line


Late Friday: McBain up by 770 at the end of the 2PP count, any changes from here will be trivial and the seat will be declared on Monday.

Friday: McBain up by 764, 104 votes to go unless any more are added today.

Second Wednesday: McBain's lead has increased slightly to 754 with probably only 100 or so votes to go.  A few things I should mention:

1. The 2PP swing on postals is currently 3.9% to Labor, vs 0.4% to Liberal for non-postals.  Had the swing on postals been the same, Labor would be about 160 behind, because of the number of postals, and would lose.  The changed mix of postal voters and/or the strength of Labor's postals campaign has made the difference.  Perhaps the former was enough by itself or perhaps either would have been, it's impossible to say.

2. John Black has a terrible article in the Australian that does the normal John Black thing of finding demographic correlations with booth-by-booth results and declaring that the by-election was swung by people with certain wildly elaborate demographics (in this case "Howard battlers" primarily), and then telling a story round that.  In the process he combines the ecological fallacy (attributing characteristics of groups to individuals) with another fallacy of mistaking correlation for causation.  Black's 1418 words of nonsense, accompanied by three completely misleading graphs, was all unnecessary since the real story of this by-election can be seen by looking at Ben Raue's maps (switch 2PP swing on, switch 2PP result off).  Bega Shire swung almost entirely to McBain because she was an excellent local candidate.  The only parts that didn't were near Kotvojs' home base.  Without this factor Labor would have lost because of swings elsewhere, especially Queanbeyan.  There's no reason to believe the demographics of Bega Shire made it vote Labor when a candidate factor fits the geographic pattern so neatly.

Friday 11:45 Well this is very weird.  Yesterday's correction in Tumut PPVC has now been reversed and McBain's lead is back to 723.  Was the correction a mistake in itself?  Anyway it's still completely over.

Thursday 7:30 The last roll of the dice that was the Tumut PPVC has come up in Labor's favour with a 174 vote correction there so McBain leads by 905 and as a contest this one is now over to the power of over! Phew!  The Labor 2PP has come up to 50.48 (-0.37) and there will be very little change from there.   Kotvojs has conceded.  

I may not follow the remaining scraps of this count too closely, but I expect to be back with a full article on the preference flows and their implications when they are out.

Thursday 4:20 McBain wins today's postals 313-310.  McBain leads by 730.

Thursday 12:15 The number of postals available today is down to 680 with the remainder being rejected.

Thursday 10:30 No counting yet but 746 postals have now arrived leaving 3804 potentially outstanding.  

Wednesday 5:30 
A strong batch of postals for Kotvojs today, splitting about 447-354 (55.8%) leaving Kotvojs now 719 behind (50.38% to McBain).  Two booths have not been rechecked, one of them being Tumut PPVC.  In theory 4565 postals could still appear but many of those won't; even assuming all did with 2% rejected and 3% informal then Kotvojs would need over 58%, barring corrections or gains on declaration prepolls.

Wednesday 12:20 No counting yet but some updates on the postals tally.  889 postals are now awaiting counting and the number potentially outstanding (many of which will never arrive) is 4501.  Processing of provisionals has also started.

Tuesday 7:45 Only one significant PPVC and six other mostly small booths left to recheck now.  McBain up by 816.

Tuesday 5:00 McBain gains about 100 votes on rechecks, mainly a 96-vote error in the Queanbeyan ordinary booth, with various other errors cancelling out.  Now only 16 booths to recheck with two PPVCs.  The largest remaining PPVC has gone (McBain lost 40 there.)

Tuesday 2:45 New postals broke 263-221 (54.3%) to Kotvojs, closing the gap to 729.  A theoretical maximum of 5936 postals still to count; it won't really be nearly that many.  28 booth rechecks to go.

Tuesday 2:30 Not much to see today so far, only four booths showing as newly rechecked leaving 31 to go.

Tuesday Two more booths rechecked so far today and McBain leads by 766. 592 postals are now awaiting counting with a maximum 5322 still to arrive (though many of those won't).  Really we are only waiting for the four big PPVs to be checked and then if nothing happens there Kotvojs should concede.

Monday late night Some more booths were rechecked tonight.  There are now 36 to go including four substantial PPVs, and McBain leads by 775.  Extremely unlikely now there could be errors out there big enough to make a difference.  

Monday 7:20 No further booths checked since 3:30-ish today.

Monday 5:30 Forty booths remain to be rechecked, including six substantial prepoll booths and two little ones.  I don't see the margin finishing below 500 without corrections in Kotvojs' favour in this process.    

Monday 4:10 More postals have been counted with 1166 counted today splitting 52.3% to Kotvojs.  The current split on all postals is 5549-4820.  Labor leads by 749 having presumably made some small gains on rechecking elsewhere.  Booths rechecked now include Yass PPVC.  In theory there are still up to 5703 postals out there; even if all arrive Kotvojs now needs 57% which would not happen.  

Monday 1:20 Jerrabombera PPVC has been rechecked with McBain gaining six on the margin there.  Movement on the postals front: there are now 1283 envelopes awaiting processing with a maximum 5845 yet to arrive.  

Monday 12:20 Adaminaby has been rectified (on rechecking Kotvojs won that booth by 96).  Another four little booths have also been rechecked.

Monday 11:40 The margin is currently showing as 835 but this is because the 2PP for the Adaminaby booth has been taken offline or mistakenly erased during rechecking (Kotvojs won this booth by 100).  

Monday 11:10: Margin down by seven on rechecking at Bredbo and Nimmitabel.  Booths requiring rechecking: 69 including nine substantial PPVCs.  

Monday 10:30: Labor's lead has increased by five votes to 742. I think this may be on counting of the small number of postals awaiting counting yesterday.

Monday 10:10:  The only news so far today is that the number of postals awaiting counting has increased to 360.  6790 (thousands of which will probably never arrive) remain unaccounted for.  Over the next few days we should keep an eye on which booths have and have not been rechecked.  At this stage the following booths have had updates today: Bowning, Central Tilba.  The following had updates on Sunday: Araluen, Thredbo, Tathra, Jerrabombera, Canberra South PPVC, Merimbula PPVC, Bodella, Queanbeyan West, Narooma PPVC, Queanbeyan Heights Central, Queanbeyan City Central, Jindabyne PPVC, BLV Eden Monaro PPVC (no votes).  The remaining booths have had no checking since Saturday and the nine PPVCs in this list could be fertile ground for further errors.  However these probably would not be as large as the ones found so far.  It is also worth tracking the total informal vote: it is currently 6038.  As votes are checked, the informal vote tends to rise.


5:40 Another batch of postals added - these broke 55-45 to Kotvojs taking the overall break on postals back to around where it was.  The live 2PP is now down to 50.4, a lead of 737.  There are in theory up to 7151 postals still remaining, and there will also be some provisionals and declaration prepolls that may aid Labor's cause by fifty votes or so.  Even assuming all the remaining postals arrive and are all accepted with 3% informal, Kotvojs would still need over 55%.  More likely she needs close to 60%.  

The counting errors pointing to a closer result now mean the Coalition has a slight edge on bragging rights.

5:15 And now a break for Labor, another 3000 postals are gone and they weakened compared with the first lot, breaking 52.5% to Kotvojs.  As concerns the result we can probably go back to sleep, but the live 2PP is down to 50.5% and it's unlikely to improve from here.  So it looks like a 2PP swing to the Coalition of a few tenths to half a point.

4:30  A big unknown in a contest that now has perhaps a flicker of life in it is how many postals are still to arrive.  6899 are unaccounted for, as well as 4751 received and awaiting processing.  If the 2019 return rate holds up, then around half of those will never arrive, which would leave Kotvojs needing about 56.7% 2CP on formal non-rejected postals - very unlikely given the weak break on the ones counted so far.  However, if, say, the raw non-returned number is the same as in 2019 (extremely unlikely) then the asking rate drops to about 55.5%, though even that is probably out of reach.  It's still very hard to see Kotvojs winning unless there is another counting error.  

3:35 Another error in the Merimbula PPVC booth cuts a further 310 off Labor's margin, and while they are still projecting to win I would think there is now some vaguely tangible if slender chance they might not, especially if any further errors are found.  

The error can be seen in this chart of preference flow by booth on the previous figures:

Merimbula PPVC doesn't look like such a big outlier visually because of the much larger Tanja outlier - Tanja is a tiny booth with a ridiculously high Green vote.  But it is.  The error has now been fixed.

2:00 Figures for the postals added by the AEC do show some degree of counting error as they were reported as 2464-2394 last night but have been posted as 2605-2231.  However, Labor is ahead by 1335.  Kristy McBain has tweeted that she is 1000 ahead with "only 4,000 left to go" based on 94K counted - suggesting that there are more votes not yet added - but I think the number to go may be somewhat larger.  It is still possible Labor's final margin could be very narrow, and fairly likely that there is a swing against Labor in 2PP terms.

10:30 Labor has done better than my projection expected on preferences at Narooma PPVC, mainly as a result of there being a very low Nationals vote there.  This puts Labor's lead at 1718, or notionally 1648 after the first batch of postals.  Meanwhile 190 postals have been rejected.

End-of-night wrap

It's been a long night but it looks like Labor has retained Eden-Monaro, with very little swing either way.  The news of a very weak break to the Liberals in the first 5000 postals has put the seat beyond realistic reach barring major counting errors, though there is still potential for the result to be very close.

The average swing to Oppositions in their own seats at by-elections contested by the government is 1.25%.  (Ignore all comparisons to the c. 4% average for all by-elections.) Labor should get a much smaller swing if they get any swing at all, but the high personal vote of outgoing ALP MP Mike Kelly is probably enough to explain that difference by itself.  All of the other things that could have influenced the outcome: bushfires, coronavirus, Coalition infighting, Labor scandals in two states - there is no evidence yet that any of them did.  It may well be that they did, but that they cancelled each other out.

The total primary vote recorded by left-ish candidates slightly exceeded that recorded by right-ish candidates, something true to much the same degree in 2019.  Fiona Kotvojs has been very successful indeed at concentrating the right-wing vote in her own pile, whereas Labor seems to have lost some primaries to the Shooter and left micros in the absence of Mike Kelly.  It's no surprise that a major party should experience a primary vote swing against in a field of 14 candidates.

Kotvojs has been less successful on preferences, but that doesn't (yet) mean that shifting preferences of any particular party's voters have made the difference.  It may just be that the number of votes cast for left-wing minor parties, and the very small vote for right-wing minors except the Nationals (I regard the Shooters as intermediate) has made the difference to the preference flows.  Anecdotally, however, the flow from the Nationals may have been weaker than the very high 87% last time, and that will be watched with interest when the final preference flows arrive in about four weeks' time.

The Greens' result is poor.  The swing against them is around 3%, most of it apparently going to left-wing micros.  The party competed more successfully against those micros at the 2019 Senate election in this seat, so it is possible that it also lost votes to Labor, perhaps because McBain appealed to some Green primary voters.

There is no adverse message about the national polls in this result.  There might have been some reason to suspect Newspoll was inaccurate if Labor won hugely or the Coalition won comfortably, but neither happened and the result is perfectly in line with historic relationships between polling and 2PP vote.

The uComms seat robopolls for various organisations have probably modestly overstated Labor's primary vote and slightly overstated the 2PP, but provided Labor wins it still seems they have done pretty well by seat poll standards, with all four tipping Labor for the win.  The Nationals internal polling, however, has turned out to be nonsense.

Provided Labor does win and it doesn't come down to, say, 200 votes, I see the result as saying very little about the condition of either major party.  As for the next general election, we saw Labor narrowly retain Braddon at a by-election then lose it at the general, so the Coalition would be hopeful of the same in this example.

Comments will follow over the next few days as the count continues.


1:00 Antony Green pretty much called this just after midnight and after running my own calculations I can't see how the Liberals can win this.  This is why.  Labor currently leads by 2020.  Off the two prepolls to report 2PPs I project them to drop 480 of that lead.  Another 70 is lost on the first 5K of postals.  That leaves 1470.  Let's assume Kotvojs does as well on the remaining postals as she did on all postals last time, though this is extremely unlikely given what we have seen from the first 5000.  In that case she gains at 0.143 votes per vote.  Ignoring provisional votes she then needs 10280 formal postals to win.  But allow for a few percent to be informal and this increases to, say, 10650.  There are at most 11840 remaining if all are returned and accepted, but that never happens as there are always voters who ask for postal votes but fail to vote or vote another way.  So even with a highly unlikely assumption there will just not be enough votes left to catch up.  It might still finish very close, but also given what we've seen on postals so far it might not.  [Edit: Jindabyne is in and the Liberals did a whole nine votes better than I projected there.]

12:15 A heap more numbers in with only Jindabyne and Narooma PPVCs still to report 2PP, so better than expected on that front.  Now Labor leads with 51.3% which the AEC projects to 50.67%.  I'll be doing some calculations to see if there is any hope of this being caught.  General consensus is there isn't. 

11:30 A big boost for Labor with the news that the first batch of postals has broken 2464-2394 to Kotvojs, a swing of 6.4% to Labor from the overall 2019 postals.  That greatly strengthens Labor's chance of holding the seat unless there is something odd going on.  

11:20 Nothing new while I did that at all.  I'll be continuing commentary but comments will be sparse over the next 40 minutes.

11:00 Soon I expect to be on with Ross Leedham over here:

10:40 Labor dropping back on projections as unfavourable PPVCs arrive, with one Queanbeyan booth still to come.  

10:25 Counting to stop at around 11 pm for OHS reasons.  AEC expecting all primaries in but many PPVCs will be missing 2PPs.  A report that 

10:20 The fizz is possibly going out of the prepolls with Labor doing well in Tumut on preferences despite primary vote swings.

10:05 Kotvojs bucked the trend of prepoll results matching the booth swings by doing OK on primary vote swing at Bega.  There are some prepolls to come in areas where Labor polled badly in the booths.   There are also some Canberra prepolls which I'd expect to be juicy for Labor.   

10:00 Labor has copped a bit of a whacking in Tumut and Yass PPVCs with both showing primary vote swings to Liberals, 4% in Yass and 6% in Tumut.  

9:45 Labor is up 6% in the Merimbula prepoll according to Antony Green but is also up in the Merimbula booths.  I have heard unofficially that in Yass prepoll the Liberals are doing well (where Labor did badly in the booths.)  Projections will bounce around as good and bad prepolls go in.  

9:26 Kristina Keneally suggests we will not get 2PPs for some of the prepolls tonight.  We'll see ....

9:20 A bad result for the Greens, back to their old ways of having their vote gouged by every random similar micro that comes along.  It does seem though from scrutineer gossip that Nats preferences are flowing weakly to Kotvojs which may be helping Labor.

9:15 Betting has Kotvojs in front.  Ross Leedham's projections have them well in the mix too but most projections have Labor ahead.  A huge unknown is how many postals are yet to arrive.  The proportion that are not returned may be higher than normal.  If this is still close at the end of the night we need to have a careful look at this.  

9:00 Not much going on at present, waiting for some prepolls.

8:30 I've done this graph of the share of ALP preferences by ALP primary vote.  In general Labor is getting more preferences where its primary vote is higher.  Two outliers, Tanja and Adaminaby (very strongly left and right respectively and both very small) have been removed.

This means that all else being equal, if Labor's primary declines, its preference share will too.  I haven't however compared this to 2019.  At the moment this chart doesn't ring alarm bells for Labor; it still projects that a uniform swing on primaries in the remaining votes would not cost them the seat.  

Ross Leedham has suggested a lot rests on just how many prepolls and postals are included.

Note that the informal vote is 6.54% currently, but it may very well go up in the counted booths with rechecking.  

8:20 William and the AEC are still both projecting below 51 while the ABC is projecting higher.  We're now waiting for prepolls to come in to see what they do on the 2PP.  In particular, will preference flows in the booths be more or less the same in the prepolls? 

8:10 7% 2PP to Coalition in Cobargo; who'd have thunk it?

8:00 William Bowe's projection has come back to 50.5 and the AEC's is on 50.7 so there is still quite a lot of life in this one though Labor are ahead.  Noteably the Labor primary is overall down to an extent that would cost them the seat if the overall preference flow was the same as in 2019 but this is evidently not so far the case.  This is probably mainly because the micro-party mix is different with a large number of left-wing outfits contesting.  I think we can call last place for Jason Potter of the Australian Federation Party at this point, but there's a vigorous fight for second-last.  This includes the Christian Democrats who have performed abysmally.

7:55 The ABC was showing very small booth swings but that is not correct.  Nonetheless overall the swing is small.  Note that the average swing in an Opposition by-election contested by the government is only 1.2% to the Opposition, not the 3.8% of all by-elections.    No prepoll booths are actually in with the AEC at 2PP level. 

7:45 Home now and just taking stock of the situation - have seen a tweet from Antony Green projecting Labor to 51.7% with 26.2% counted. William Bowe has 51.3%.  Obviously not callable yet.  Will be taking stock over the next 10 minutes.  

I will be posting some comments on the Eden-Monaro by-election tonight, but because of another commitment I do not expect to be online before about 8 pm.  Even if the by-election has by that stage been called (and I agree with it) there should still be something to discuss!  Once comments start they will appear here; refresh every ten minutes or so for updates.  This post will continue with updates through the postcount over following days, the volume of updates depending on whether the outcome remains in doubt.  In the meantime for those looking outside the obvious ABC, keep an eye on Poll Bludger and I believe Ross Leedham will be doing something.  Also check Tally RoomMy guide page for this by-election is here.

Note: I have redone comments to scroll to the top.


  1. Don't apologize for not starting till 8pm Kevin. At general elections all the channels start their "coverage" soon after 6 and spend anything from 60 to 90 minutes saying "it's early days yet.... early days yet... early days yet". It's brilliantly summed up in the Election Grandstand sequence in The Rise and Rise of Michael Rimmer, where the compere asks a pollster what he thinks of the situation and he replies "Well frankly Stephen I think we've started an hour too early". I'll send you a video clip, which should be shared with all election pundits.

  2. What are you expecting when the pre-polls and postals come in? Labor has increased its postals 100x (40 to 4000). So, that's likely to cause a swing in postals. If that's not accounted for presently, I'd expect this to blow out the result.

    1. I have never seen a scientific analysis of the impact of party postals on postal vote flow patterns. One would suspect there is some impact, but a party postal is just one collected by a party, it does not commit the voter to voting for that party. As for prepolls we will see, I see no particular reason for them to swing differently to the booth votes, but they may. (There wasn't much in that in the Queensland by-elections.)

    2. My prediction is looking better with the first batch of postals. --DVC

  3. Who do you think will take out 3rd place out of Shooters, Nats and Greens once all the preferences are thrown? There's dispute over whether the Greens had a very bad night, or if it all be explained by drift to single issue parties for the byelection (with those votes presumably coming back in the general).

  4. Latest update Thursday 5:30? Have you dragged out the ouija board or the crystal ball? On a serious note has there ever been a by-election where the winner could have brought down the government or result in a hung parliament if the governments candidate had lost?

    1. Ta, I'd noticed and fixed that one while you were commenting. The 2018 Wentworth by-election did cost the government its majority leaving it with 75/150 seats assuming that you treat Kevin Hogan as a member of the government despite him sitting on the crossbench (I do as he remained a member of the National Party). The 2017 Bennelong by-election could have had the same effect but didn't.

      After the 1940 election the Menzies government held 36/74 and relied on two independents. Two by-elections were held that could have reduced that number to 35, which might have brought down the government or forced it to go to a snap election. The government won both but was eventually brought down anyway. In the parliaments following the 1913, 1961 and 2010 elections no government seats went to by-elections. The Hughes government following the 1919 election was one seat short of a majority and lost another in a by-election before regaining one, but at no stage were the numbers close to there to remove it.