Hands off Sankey

A visual representation of how votes flowed between the parties at the 2016 and 2019 elections, plus other observations from the Australian National University’s post-election survey.

First up, note that you can find Adrian Beaumont’s latest British election post immediately below this one, and that The Guardian has preliminary details of what will presumably be the last Essential Research poll for the year, which I will blog about this evening when the full report is available (suffice to say for now that it still doesn’t feature voting intention numbers).

Now on to some further observations from the Australian National University’s post-election Australian Election Study survey, at which I took a preliminary look at the tail end of the previous post. Over the fold at the bottom of this post you can find a Sankey diagram showing how respondents’ vote choices in 2016 and 2019 compared, based on the slightly contingency of their recollections of what they did three years ago.

These suggest the Coalition actually lost a sizeable chunk of voters to Labor – 5.1% of the total, compared with only 1.6% going the other way. I might take a closer look at the survey responses for that 5.1% one day, but presumably they were the kind of Malcolm Turnbull-supporting voter who drove the swing to Labor in affluent inner urban areas. The key point is that the Coalition was able to make good this loss out of those who were in the “others” camp (i.e. everyone but the Coalition, Labor and the Greens) in 2016 – both directly, in that fully 30% of “others” from 2016 voted Coalition this time (or 4.1% of voters overall, compared with 1.6% who went from others to Labor), and indirectly, in that their preference share from what remained went from 50.8% to 56.3%.

Before that, some other general observations based on my reading of the ANU’s overview of its findings:

• The survey adds context for some intuitively obvious points: that the Coalition won because self-identified swinging voters rated them better to handle the economy, taxation and leadership, and rated those issues the most determinants of their vote choice. Labor’s strengths were, as ever, health and environment, which rated lower on the importance scale, and education, which hardly featured.

• Coalition and Labor voters weren’t vastly in their opinions on negative gearing and franking credits, with support and opposition being fairly evenly divided for both. However, there were enormously divided on their sense of the importance of global warming, which was rated extremely important by 64% of Labor voters but only 22% of Coalition voters.

• A drop in support for Labor among women caused the gender gap to moderate compared with 2016, although the unchanged 10% gap on the Liberal vote remains remarkable by recent historic standards. The new normal of Liberal doing better among men and Labor among women only really goes back to 2010 – back in the Keating era, it was Labor who had the women problem.

• Scott Morrison trounced Bill Shorten on popularity, their respective mean ratings on a zero-to-ten scale being 5.14 and 3.97.

• The number of respondents professing no party identity reached a new peak of 21%, maintaining a trend going back to 2010.

• The 2018 leadership coup was received as badly as the 2010 coup against Kevin Rudd. The 2013 and 2015 coups were less badly received, but both scored over 50% disapproval.

• Long-term trends show a steady erosion in trust in government, satisfaction with democracy and belief government is run for “all the people”, although the 2019 results weren’t particularly worse than 2016. Satisfaction with democracy is poor compared to the countries with which Australia is normally compared – though slightly higher than the United Kingdom, which is presumably one symptom among many of Brexit.

Continue reading “Hands off Sankey”

UK election minus three days

The lower-educated appear likely to sink Labour in Thursday’s UK election. Also featured: a guide to how the results will come in on Friday (our time).

Seven UK national polls were released last week, with the Conservatives leading by eight to 11 points in five and by 14% to 15% in Survation and Opinium. There was little change since last week in most polls, but the Conservative lead was up five points in Survation.

Donald Trump was in the UK from December 2-4, and there was a head-to-head debate between Boris Johnson and Jeremy Corbyn on December 6. Polls listed above were taken after the Trump visit, and Johnson won the leaders’ debate by 52-48 in a YouGov “insta poll”.

In the latest YouGov poll, the Conservatives hold leads of 39-34 with the ABC1 social grade (upper to middle class), but by 48-31 with C2DE (working class). Sky’s Lewis Goodall has qualitative research into Labour’s problems with lower-educated voters. Facebook ads have damaged Corbyn’s popularity with these voters. The Conservative message that Labour has blocked Brexit is cutting through.

While Corbyn has a problem with the lower educated, he’s far from unique. Centre-left parties had unexpectedly dismal results in the UK (2015), US (2016) and Australia (2019), owing to swings towards conservative parties among the lower educated.

There are probably two ways for the global left to start winning elections consistently again. One is via a deep economic recession. The other way is via demographic change. Since 1940, educational attainment among those aged 25-29 in the US has surged. As the population becomes better-educated, the left is likely to do better – but not for a long time.

A hope for UK Labour is that Johnson’s ratings in an Ipsos poll slumped 22 net points to -20 since November, while Corbyn was up 16 to -44. Something could go wrong for the Conservatives with Johnson that unpopular. In YouGov, Johnson’s net approval was down nine points since last fortnight to -13, Corbyn down five to -47 and the Liberal Democrats’ Jo Swinson down 18 to -36.

As I wrote previously, there are three ways Labour could defy the polls. A fourth can be added: late deciders. While the Conservatives lead by 43-33 in YouGov, they only lead by 33-26 including “won’t vote” (8%), “don’t know” (13%) and “refused” (3%). If late deciders break to Labour, it will be closer than current polls suggest.

At the UK’s May European elections, pollsters tended to overstate support for the Brexit party, Conservatives and Labour, and understate support for the Lib Dems and Greens. The Brexit party was the clear choice for Leavers at that election. In general, the performance of UK pollsters has been poor.

A guide to election results day (Friday)

UK polls are open from 7am to 10pm Thursday local time. Unlike Australia, where small booths report quickly, the UK has no counting by booth. Instead, votes from all booths within a seat are transported to a central counting centre, and counted there. Postal votes must arrive by election day. Barring a recount, seats are declared once the vote count finishes. It takes far longer to get a good idea of the result than in Australia.

To follow their elections, the British need to pull an all-nighter. In Australia it’s easier, with polls closing at 9am Friday Australian Eastern Daylight Time (AEDT). Here is my guide to the events in Australia; all times are Friday AEDT.

9am: Polls close and The Exit Poll is released (intentional capitalisation). In the last three elections, The Exit Poll has given seat results which greatly disagreed with pre-election polls and expectations. In all three cases, The Exit Poll was far closer to the mark than pre-election polls. Only seat counts are given, not vote shares.

11am: According to this article about the 2017 election, only three of 650 declarations are expected by this time.

1pm-3pm: These two hours should be the heaviest for declarations.  Initial results will be biased to Labour as the Conservative heartland regional seats take longer to gather their votes. The key is to watch the changes in vote share, and whether seats are being gained or lost.

6pm: Only a few seats will not be declared by this time. Very close seats can take longer to declare owing to recounts. If there’s snow on the roads, results will be delayed.

From the editor

Below is an update of the poll tracker I published on Friday, with nine new polls added. It maintains a trend of steady improvement for Labour at the expense of the Liberal Democrats, while the Conservatives hold steady. The latest trend result is Conservatives 42.7% (up 0.3% on 2017), Labour 33.7% (down 6.3%), Liberal Democrats 12.3% (up 4.9%) and Brexit Party 3.4%.

Newspoll: 52-48 to Coalition

Another modest Coalition lead from the second poll in a new-look Newspoll series, which also finds Scott Morrison rated well for strength, vision and experience, but higher than he’d like for arrogance. Also featured: a quick early look at the ANU’s deep and wide post-election survey.

The second Newspoll conducted under the new regime of online polls conducted by YouGov records the Coalition with a 52-48 lead, out from 51-49 a fortnight ago. On the primary vote, the Coalition is up a point to 42%, Labor is steady on 33%, the Greens are down one to 11% and One Nation is steady on 5%. Both leaders’ personal ratings are improved after weak results last time, with Scott Morrison up two on approval to 45% and down four on disapproval to 48%, and Anthony Albanese up two to 40% and down four to 41%. Morrison’s lead as preferred prime minister is out from 46-35 to 48-34.

Respondents were also asked to rate the leaders according to nine attributes, eight positive and one negative. Morrison scored higher than Albanese for the experience (68-64), decisiveness and strength (60-51) and having a vision for Australia (60-54), while Albanese had the edge on caring for people (60-55). There was essentially nothing to separate them on understanding the major issues (57-56 to Albanese), likeability (56-56), being in touch with voters (50-49 to Albanese) and trustworthiness (49-48). However, Morrison’s worst result was his 58-40 lead on the one negative quality that was gauged – arrogance. The Australian’s paywalled report of the results is here.

In other poll news, a uComms poll (apparently minus the ReachTEL branding now) for the Courier-Mail ($) suggests Queensland’s embattled Deputy Premier, Jackie Trad, is in grave danger of losing her seat of South Brisbane to the Greens. The poll shows the Greens on 29.4%, Labor on 27.5% and the Liberal National Party on 26.6%, with 10.4% undecided. Labor is credited with a 52-48 lead on respondent-allocated preferences, but this may flatter Labor given the LNP’s announcement that they would direct preferences against them. No field work date is provided that I can see, but the sample size was 700. The deficiencies of automated phone polls in inner city seats were noted by Kevin Bonham, among others.

UPDATE: In better poll news still, the results from the post-election Australian Election Study survey are available in all their glory, courtesy of the Australian National University. You can view the ANU’s overview of the findings here, but the real fun of this resource is that it allows you to cross-tabulate responses to 3143-respondent survey across a dizzying range of variables. The survey also includes demographic weightings that presume to correct for the biases introduced by the survey process. The survey also addresses a long-standing criticism by including a component of 968 respondents who also completed the 2016 survey, allowing for study of the changing behaviour of the same set of respondents over time.

Rest assured you will be hearing a great deal more about the survey going forward, but for the time being, here’s one set of numbers I have crunched for starters. This shows the primary vote broken down into three age cohorts, and compares them with the equivalent figures from the 2016 survey. This produces some eye-catching results, particularly in regard to a probably excessive surge in support for the Coalition among the middle-aged cohort – mostly at the expense of “others”. By contrast, the young cohort swung heavily to the left, while the boomers were relatively static.

British election polling in five charts

A quick and dirty guide to what pollsters seem to be telling us a week out from an unpredictable British election.

With a week to go until Britain’s election, polls continue to credit the Conservatives with a lead over Labour which, although substantial, is not so great that they can lock in a parliamentary majority. Furthermore, the trend of polling is somewhat in favour of Labour, with the Conservatives apparently having sucked dry the short-lived insurgency of the Brexit Party, while Labour continues to pick off support from the floundering Liberal Democrats. This is illustrated clearly enough in the poll trend chart below, which combines the work of nine polling series.

While all the pollsters have the Conservatives well ahead, the size of the lead covers a wide range, from as little as 6% by the reckoning of the most recent poll from BMG to as much as 15% from Opinium. Given the assurance that the Scottish National Party will dominate north of the border, the former end would certainly be weak enough to leave the Conservatives short of the formidable hurdle that either major party must clear if they are to win enough seats in England to score a majority. The next chart shows local area regression trends for each pollster’s reckoning of the Conservative lead (a little more on the methodology is explained later in the post).

Labour optimists are hanging on to the notion of a pro-Labour “youthquake” that will up-end the pollsters’ turnout models. The importance of the age distribution of the voting population is forcefully illustrated by the chart below, which shows voting intention by age cohort (off very small sub-samples) from a recent poll by ICM Research.

Pollsters are varying quite substantially as to age distribution, as illustrated in the next chart, which has been derived from the weighting data provided as standard from pollsters in the UK (in Australia we can only imagine such things). Kantar looms as an outlier in its expectation that fully 49% of voters will be 55 and over, compared with just 19% for the 18-to-34 cohort. The other pollsters range from 25% to 29% for 18-to-34 and 37% to 43% for 55-plus. None of this has any obvious bearing on the pollsters’ leanings, perhaps with the exception of ICM, whose young age profile has been reflected by relatively modest Conservative leads.

Then there’s their modelling of the population by vote in the 2016 Brexit referendum, which I’m slightly puzzled by in that there is dramatic variation in the size of the various pollster’s “did not vote” cohort (which appears not to be related to their age distributions). In the cases of YouGov, ComRes, Survation and Deltapoll, I had to infer this total from what was left over after “leave” and ”remain” were removed from the total, which may be causing me to miss subtlety. Whatever the case, let the peculiarity of Kantar again be noted in that it proposes a majority “remain” population, despite having a age distribution that skews old. Conversely, Opinium’s weighting to “leave” may explain its apparent lean to the Conservatives.

Now for a quick introduction to the British polling fraternity. First up, the following table shows bias adjustments that have been used to standardise the poll trend measures at the top of this post. These were achieved my comparing their results to a straightforward trend measure of all the polls entered into the model (142 polls from nine pollsters). The results are actually fairly modest as these things go, contrary to the impression given by the range of results in the “Conservative lead trend by pollster”.

All these polls are of the online panel variety, with two exceptions: Survation, a phone poll, and Ipsos MORI, which despite being a big name has only published one poll since the campaign began. Survation is also unique in that it includes Northern Ireland in its polling, whereas the others stick to England, Scotland and Wales.

The pollsters are variable in how they structure their voting intention responses, particularly in relation to the Brexit Party, which is not running in most Conservative-held seats. Before that they had to struggle with which minor parties to include among the initial list of responses and which as a follow-up for those who chose “other party”. It is routine for the Scottish National Party and Plaid Cymru to be included as topline response options in Scotland and Wales respectively.

Winners and losers

Reading between the lines of the Liberal Party’s post-election reports for the federal and Victorian state elections.

In the wake of Craig Emerson and Jay Weatherill’s federal electoral post-mortem for Labor, two post-election reviews have emerged from the Liberal Party, with very different tales to tell – one from the May 2019 federal triumph, the other from the November 2018 Victorian state disaster.

The first of these was conducted by Arthur Sinodinos and Steven Joyce, the latter being a former cabinet minister and campaign director for the conservative National Party in New Zealand. It seems we only get to see the executive summary and recommendations, the general tenor of which is that, while all concerned are to be congratulated on a job well done, the party benefited from a “poor Labor Party campaign” and shouldn’t get too cocky. Points of interest:

• It would seem the notion of introducing optional preferential voting has caught the fancy of some in the party. The report recommends the party “undertake analytical work to determine the opportunities and risks” – presumably with respect to itself – “before making any decision to request such a change”.

• Perhaps relatedly, the report says the party should work closer with the Nationals to avoid three-cornered contests. These may have handicapped the party in Gilmore, the one seat it lost to Labor in New South Wales outside Victoria.

• The report comes out for voter identification at the polling booth, a dubious notion that nonetheless did no real harm when it briefly operated in Queensland in 2015, and electronic certified lists of voters, which make a lot more sense.

• It is further felt that the parliament might want to look at cutting the pre-poll voting period from three weeks to two, but should keep its hands off the parties’ practice of mailing out postal vote applications. Parliament should also do something about “boorish behaviour around polling booths”, like “limiting the presence of volunteers to those linked with a particular candidate”.

• Hints are offered that Liberals’ pollsters served up dud results from “inner city metropolitan seats”. This probably means Reid in Sydney and Chisholm in Melbourne, both of which went better than they expected, and perhaps reflects difficulties polling the Chinese community. It is further suggested that the party’s polling program should expand from 20 seats to 25.

• Ten to twelve months is about the right length of time out from the election to preselect marginal seat candidates, and safe Labor seats can wait until six months out. This is at odds with the Victorian party’s recent decision to get promptly down to business, even ahead of a looming redistribution, which has been a source of friction between the state and federal party.

• After six of the party’s candidates fell by the wayside during the campaign, largely on account of social media indiscretions (one of which may have cost the Liberals the Tasmanian seat of Lyons), it is suggested that more careful vetting processes might be in order.

The Victorian inquiry was conducted by former state and federal party director Tony Nutt, and is available in apparently unexpurgated form. Notably:

• The party’s tough-on-crime campaign theme, turbo-charged by media reportage of an African gangs crisis, failed to land. Too many saw it as “a political tactic rather than an authentic problem to be solved by initiatives that would help make their neighbourhoods safer”. As if to show that you can’t always believe Peter Dutton, post-election research found the issue influenced the vote of only 6% of respondents, “and then not necessarily to our advantage”.

• As it became evident during the campaign that they were in trouble, the party’s research found the main problem was “a complete lack of knowledge about Matthew Guy, his team and their plans for Victoria if elected”. To the extent that Guy was recognised at all, it was usually on account of “lobster with a mobster”.

• Guy’s poor name recognition made it all the worse that attention was focused on personalities in federal politics, two months after the demise of Malcolm Turnbull. Post-election research found “30% of voters in Victorian electorates that were lost to Labor on the 24th November stated that they could not vote for the Liberal Party because of the removal of Malcolm Turnbull”.

• Amid a flurry of jabs at the Andrews government, for indiscretions said to make the Liberal defeat all the more intolerable, it is occasionally acknowledged tacitly that the government had not made itself an easy target. Voters were said to have been less concerned about “the Red Shirts affair for instance” than “more relevant, personal and compelling factors like delivery of local infrastructure”.

• The report features an exhausting list of recommendations, updated from David Kemp’s similar report in 2015, the first of which is that the party needs to get to work early on a “proper market research-based core strategy”. This reflects the Emerson and Weatherill report, which identified the main problem with the Labor campaign as a “weak strategy”.

• A set of recommendations headed “booth management” complains electoral commissions don’t act when Labor and union campaigners bully their volunteers.

• Without naming names, the report weights in against factional operators and journalists who “see themselves more as players and influencers than as traditional reporters”.

• The report is cagey about i360, described in The Age as “a controversial American voter data machine the party used in recent state elections in Victoria and South Australia”. It was reported to have been abandoned in April “amid a botched rollout and fears sensitive voter information was at risk”, but the report says only that it is in suspension, and recommends a “thorough review”.

• Other recommendations are that the party should write more lists, hold more meetings and find better candidates, and that its shadow ministers should pull their fingers out.

UK election minus nine days

Three ways Labour could outperform the polls, and defy the Conservatives’ lead. Also featured: recent developments in Germany. Guest post by Adrian Beaumont.

Guest post by Adrian Beaumont, who joins us from time to time to provide commentary on elections internationally. Adrian is an honorary associate at The University of Melbourne. His work on electoral matters for The Conversation can be found here, and his own website is here.

There were seven UK national polls released last weekend. In five of these polls, the Conservative lead was six to ten points, but it was 13 points in Deltapoll and 15 in Opinium. While Labour has gained since last week, the Conservative vote is holding between 42 and 46% in six polls. The exception is BMG, where the Conservative vote was just 39%. I believe a Conservative vote below 40% is the target for a realistic chance of Labour forming the next government at the December 12 election.

On November 26, the Chief Rabbi accused Jeremy Corbyn of antisemitism, and the BBC’s tough interviewer, Andrew Neil, interrogated Corbyn. Boris Johnson has not scheduled an interview with Neil. Polls listed above were conducted after the Corbyn/Neil interview.

On November 29, there was a terrorist attack on London Bridge in which two people, not including the attacker, were killed. Prior to the 2017 election, there were two major terrorist attacks in the UK, but Labour performed much better than expected.

Donald Trump will be in the UK for a NATO summit this Monday to Wednesday. This could assist Labour by focusing more attention on their November 27 claim, using leaked documents, that the Conservatives are planning to sell the National Health Service (NHS) to the US to clinch a US/UK trade deal. On Friday UK time, there will be a head-to-head debate between Johnson and Corbyn.

There has been some commentary that suggests Labour would be 20 points ahead if not led by Corbyn. But the main reason for people to vote Conservative is to Get Brexit Done. No Labour leader could match the Conservatives’ Brexit rhetoric. Labour’s 2017 performance was partly due to Corbyn being pro-Brexit. In addition, the latest jobs data indicates the economy is good for most people: unemployment is just 3.8% and real wages growth is 1.7%.

It is likely a centre-left, pro-Remain Labour leader would be destroyed by accusations of betraying the Brexit referendum.  Labour’s NHS scare campaign is their only realistic chance to regain enough lower-educated voters before the election.

Three ways current polling could understate Labour

Last-minute tactical voting: The Liberal Democrats still have about 13%, but in seats that are clearly Conservative vs Labour, it is plausible that many Lib Dems will vote Labour when faced with the ballot paper. While there is no love for Corbyn among hard Remainers, there is extreme antipathy to Johnson’s hard Brexit and illiberal agenda.

Differential turnout: It’s wrong to say that one side or the other benefits from high overall turnout. Where one side benefits is when their demographics vote at a higher rate than the other side’s demographics. In the past, better-educated and older people voted Conservative. As these demographics are more likely to vote, Conservatives tended to outperform their polls. But while the Conservatives retain an advantage among older people, they have lost it among the better-educated. It will be cold with short daylight hours on election day. Complacency among Leave voters could mean that relatively few vote. If there was relatively heavy turnout among Remainers, the Conservatives would likely perform worse than expected.

Vote efficiency: In this Conversation article, I wrote that Labour won over 70% in 37 of the 650 seats at the 2017 election, while the Conservatives had no 70%+ seats. Labour could lose many votes in their inner city strongholds to the Lib Dems and Greens, and still hold easily. This was the pattern during the Blair government years: high Lib Dem votes in the inner city, but low in the Labour/Conservative marginals. Meanwhile, the Conservatives could waste many votes owing to the Brexit party’s withdrawal from Conservative-held seats.

Nate Silver’s theory is that polling errors go in the opposite direction to the conventional wisdom’s expectations. The commentariat expect a healthy Conservative majority.

Germany: left wins SPD primary

On Saturday, the left-wing candidates defeated the moderate candidates by 53.1-45.3 in a membership postal vote for leadership of Germany’s centre-left SPD. The SPD is likely to abandon the grand coalition with the conservative CDU that has governed Germany for three of the last four terms since 2005. The next German election may need to be held before September 2021. Once one of the two major German parties, the SPD currently has about 14%, level with the far-right AfD and well behind the CDU and Greens.

See my personal website for the Swiss upper house results. At the October 27 Argentine election, the left won the presidency, but will need help from regionalists to pass legislation.