Anatomy of a Modern Campaign
One of the biggest jokes politics plays on itself is “secrecy” - secret internal polling that apparently makes public polling wrong, secret campaign tactics that relegate Sun Tzu to
Apr 22, 2013
One of the biggest jokes politics plays on itself is “secrecy” - secret internal polling that apparently makes public polling wrong, secret campaign tactics that relegate Sun Tzu to
One of the biggest jokes politics plays on itself is “secrecy” – secret internal polling that apparently makes public polling wrong, secret campaign tactics that relegate Sun Tzu to the status of mere amateur, through to other various secret and supposedly profound insights into the electorate that are beyond the comprehension of normal mortals. Cryptic snippets of these Holy Grails of political knowledge find their way into the public space through deliberate leaks to the media, generally whispered by various political folks that either seemingly enjoy the ego inflation that comes along with being “holders of the secret information”, or believe that the stories generated from them actually make a rats arse of difference with the electorate.
The actual reality of this “secret information” is usually a fairly different story altogether, with 95% of it being pretty banal and ordinary – and that’s when the stuff that makes its way to the public space isn’t completely fabricated to begin with.
But hacks with delusions of being Bruno Gianelli will continue being so, and this sort of nonsense won’t disappear any time soon.
So lets kick the door down on the secrecy and see how a modern campaign actually runs – the research involved, the technology, the analysis, the logistics of the ground game and the capabilities that get brought to the table when all of these things become integrated. Let’s look at the anatomy of a modern campaign – not just any campaign, but one of the more sophisticated campaigns ever run and one that is happening in Queensland right now.
Some of the details involved here will surprise a lot of people, scare the bejesus out of others and demonstrate that there’s a lot of pontificating has-beens around the joint whose understanding of modern politics is pretty redundant and irrelevant.
First – a brief partisan backgrounder (feel free to skip it if your conservative sensibilities are feeling particularly delicate at the moment). When the Newman Government came to power in Queensland – off the back of making a set of campaign promises to the Qld public sector and wider Qld public on public services – they quickly set about breaking them. They sacked 14,000 public servants when they promised otherwise (costing themselves around $300 million more than they needed to, while ending up with a larger public service than they would have had if they simply sat around with their thumb up their rectum and did nothing). They’ve started attacking the delivery of a wide range of public services based on similar levels of competence demonstrated with their public sector “reform”, and have generally acted like a bunch of arrogant fools defined by their gross ignorance and public policy ineptness.
Enter stage far-right Peter Costello. The Newman Government, with delusions of grandeur stemming from their enormous victory, hired Costello to produce an alleged audit report that would provide a blue-print for the government’s policy direction over the next term (or five terms if you listened to some of the hubris coming out of them). Costello’s report basically recommends privatising everything the Qld government owns and outsourcing everything it doesn’t where possible – a sort of ideological wet dream that makes Kennett’s Victoria appear as a bastion of socialist endeavour..
Unfortunately for the government, the people of Qld have other ideas about that.
So Newman has promised that there will be no privatisation without a mandate from the Qld people at the next election (an election at some unspecified time) in an effort to curtail the backlash. While the summary report was released a while ago, the full report is due to be unveiled publically some time during the next few weeks.
Which brings us to today. A government with little public policy experience and a (albeit brief) history of general incompetence in the area, about to receive the centrepiece of its first term policy agenda that recommends undertaking the largest privatisation and outsourcing program in Australian history, but one that can only be fully achieved with a fresh mandate from the Qld people via a new election.
Then there’s us (my employer) – Qld’s public sector union – and our campaign.
From the very beginning – and this is going back to well before the last Qld election – we discussed what may or may not happen with the next government and the resources and capabilities we may or may not need. A fair bit of what you could describe as casual wargaming was had, and various capabilities were built up or continued to be built up that could be utilised or redeployed as necessary – particularly on the non-industrial side (the community campaigning and advocacy side). We also started to devise a fairly significant research program.
So we came into this with existing capabilities and knowledge – which is a fundamentally important, but often unappreciated component of any campaigning apparatus.
Let’s jump ahead to the beginning of this year. We knew what Newman was doing, we knew what he was preparing, we also knew that powerful parts of the LNP were starting to push the idea of an early election – so we flicked the switch to “on”.
You cannot manage what you do not know
It’s not 2001 anymore. It’s not 1992. It’s certainly not 1985
You simply cannot manage what you do not know.
We knew from existing research that the Qld public was opposed – vehemently opposed – to the privatisation and outsourcing of public services. More so than any other state.
But was that opposition homogenous? If not, where was it strongest and weakest by geography, by demographic cohort, by political cohort – and what were the spreads involved?
What was the strength of that opposition? How much was passive and how much was active opposition – as in how much was it a generically passive “Nah – I don’t like it” type thing compared to how much was more akin to “I’m storming the fucking barricades if you try that one on mate!” type opposition?
What was the relationship between privatisation and voting intention? How much was one going to drive the other? How much did it now? How much could it? What were the sensitivity thresholds involved? What was the distribution of the dynamics of those issue/voting intention relationships across the major demographic, social, cultural and electoral cohorts?
We found out.
To do so took the largest single polling and research program in Australian history.
Dual use Robocalling
After the summary of the Costello report was released and the media had spread around the message of its contents for a week, we launched the first phase of a two phase research program, but one that could be harnessed for a second use – canvassing and campaign recruitment. The timing was also important, as the government was just about to start considering the contents of the report. We wanted the electorates to give some pretty solid feedback immediately before their local LNP members went into the consideration phase. Suffice to say that was successful.
We knew Megapoll was coming (more on that later) which would provide a massive amount of statistically accurate data on opinion, yet we also knew the statistics involved with ringing random Qld households i.e. if you ring 10,000 random households and someone picks up the phone, we know what percent will be men, what percent will be women and what percent of each of those will have what ages as a series of probability distributions.
So we bombarded strategic electorates (at first) with robocalls using ReachTEL (The Peoples Pollster) that were unweighted. We didn’t need the results to contain any demographic information that enables it to be weighted as all orthodox polls are, as we could deal with that at the end with some clever maths, to derive real world estimates that were within acceptable error parameters for us with this part of the research program (which saved money, increased completion rates and delivered more calls).
The dual-use robocalling was a simple message and question combination that went (with slight rewording each day to account for what day the calls were made on):
This is a message from Together. On Monday, the Queensland state government is considering a report which recommends the privatisation of public services including health and hospitals, disability services, child protection, corrective services and schools. It means these public services could be handed to large corporations to run
Do you support or oppose the privatisation of these public services?
People heard the message, which was important in and of itself, but for those that answered “Oppose”, a follow up question was asked:
Would you be willing to be contacted to join a community campaign to stop the privatisation of these public services?
The people that answered “YES” to this then had their numbers recorded. We are one of the few unions with our own dedicated outbound call centre, so we then used our call centre to contact these new recruits and bring them into the community campaign we were planning.
It turned out to be so successful that we continued to run it for the next week across every LNP held electorate in Qld, hitting over 130,000 people with the full message, getting over 50,000 household responses and garnering over 10,000 contacts of which about 75% were in places and electorates that were “campaign viable”.
On the research side, we then broke those results down by electorate and used a big regression/simulation analysis (combined with Megapoll and recruit data) to find the passive/active spread of opposition to generic privatisation for each electorate.
From this, we not only added thousands and thousands of new recruits to our community campaign from demographic cohorts and occupational backgrounds and electorates that generally never join anything, but we also know the passive/active split of opposition in each electorate – so we know what proportion of people are passively opposed (those that simply don’t like it) compared to those that are actively opposed (those that don’t like it and are willing to participate in activities in order to do something about it) in every LNP held seat in Qld. This is one of the few pieces of data (one of only three in fact) from our mountains of data that we will keep secret – such is its high value. Suffice to say that some electorates have active opposition rates well over 20% of the population – which is just extraordinary in historical terms. That means over 1 in 5 people in those electorate are so pissed off with the idea of privatising public services that they are willing to actively participate in stopping it. We also know that it never drops below 7% in any electorate, including those on the Gold Coast which are notorious for not giving a shit about politics.
Phase two of the research program was undertaking the largest proper political poll in Australian history with a sample of over 36,000. We polled Qld at the electorate level, for every electorate, with a series of questions and demographic information. You can read all about the Megapoll results here – it’s the state level results with links to the micro-regional breakdowns (groups of seats between 3 and 8 in number). I’ve also got it in a condensed version here. The seat level data as a whole is the second piece of data we’ll keep to ourselves, though we’re releasing a lot of it to various local media outlets so they can get a handle on the views of their local communities. The demographic cross-tabs contains extraordinarily valuable information not only for the campaign, but our broader analysis and continuing research.
You cannot manage what you do not know?
We now “know”.
Between the Dual-Use Robocalling program, Megapoll and our other research, we can now answer every single one of the questions under the “You cannot manage what you do not know” heading above. Every. Single. One.
Before we get onto some juicy bits about how this works in practice, it’s worth noting about now that among the lesser political commentariat – the types of dismal people that make a partial living providing the elevator music of political commentary – we were accused of push polling.
This accusation mostly comes from various drones that couldn’t comprehend why one might need such a large poll, or folks thought Megapoll was actually the dual use robocalling, but were too dim to understand either program, or too lazy to even bother asking someone to find out (we’re hard to find apparently!) before pontificating on matters they know not an ounce of shit about.
Well, except for Lawrence Springborg who has made the allegation on a number of occasions, but he appears to be perpetually confused, so we’d really expect no less from him.
Push polling, by the way, is a bit of a myth in Australia (as it is in most parts of the world) – and it doesn’t mean what most people think it does. What happened to John McCain in South Carolina is push polling . Seeing a poll you don’t like the wording of, or don’t understand the purpose of, isn’t.
Meanwhile, this isn’t the only data we have. One of things that all Federal members have access to (which basically means every semi-advanced level political hack in the country has access to as well) – is a statistical estimate of the breakdown of the most recent election not just by booths, but booth catchments (areas around a booth that catch where people that vote in each booth live), which are then broken down further into census level blocks at the Census Collection District level (and now moving to what’s called the SA1 level that replaced them in the 2011 Census). It’s provided by the Parliamentary Library via work produced by the AEC, that they undertake primarily for reasons of electoral redistribution processes and analysis. Basically it anchors the booths people vote at to their home addresses, aggregated into CCDs. That might not sound like much, but it’s actually an extraordinarily useful piece of information.
The third piece of data that we keep secret and not for public release, is that we’ve built a very similar thing, but at the Qld state level for Qld state electoral data by combining a massive amount of AEC, ECQ, census data, polling data and about half a dozen other datasets into a large model – what separates it from the norm is that it focuses on the probabilities of vote estimates over the simple “estimated” variety . That lets us deal with uncertainty in a much more sophisticated way. As far as I know, we’re the only mob to have such a thing in Qld that is not a complete dogs breakfast .
We have two layers of this data – the original 2012 State election data at SA1 level or combinations of SA1 level data, and a second layer of the data which is dynamic and evolves according to our ongoing polling research. Because we can measure which demographics are shifting in what seats, across large inter-temporal samples, we can project estimates of the changes of this data into the SA1 level booth catchments (but more importantly, project them as probability distributions of the vote) that aggregate up across cohorts and geographies to equal the seat wide level data results – something you can’t even begin to do unless you undertake the type of programs we’ve undertaken. Which, it should be noted, no one else in the country ever has.
Many people stuck in the 1980s questioned why we undertook such large samples – this is a key part of the answer.
This lets us know where the weaknesses are in every electorate for every sitting member, not only in terms of the most recent election result (telling us where swing voters actually live – because you can compare the election result before last at this level, to the most recent election result at this level – all down to areas between about 10 to 40 street blocks in resolution), but also how those swing voters are currently behaving by utilising the current dynamically projected estimates and attached probabilities to find new areas of weakness, or recognise areas that have slipped out of our cohort targeting window.
If you’re running a community based campaign ground game, an increase of just 10% or 20% efficiency in targeting the right people with the right message at the right time is a massive difference. If we increase our targeting efficiency by just 400%, it will be a disappointment.
That’s where this matters.
Before we get to some examples of how it all comes together, we also have our campaign recruitment. Currently we have about 35,000 people involved with our community campaign apparatus called Working For Queenslanders (Feel free to join!)
Yet having volunteers to campaign is not enough – you need organisation to make it all work. It’s all very good and well getting volunteers to go to some place to do some particular thing, but unless there’s also people on the ground to guide it, that are able to deal with the complexities that often arise and can enable the whole exercise to work – well, you’re really just a clusterfuck waiting to happen.
We already have a substantial number of people in Working For Queenslanders with training and organising capabilities, but we also have an enormous in-house training capability which we’ve developed over many years. One of our biggest strengths as a union is our campaign training capability – we often train third party groups. Our folks tell me we can easily train over 1000 people a month at the drop of a hat, more if we had to.
This is all without having pressed the button on any major recruiting program of our own 40,000 odd members for the campaign yet (many of which already have training that can be deployed instantly, are highly intelligent and, let it be said, are completely pissed off with the government).
And this is just us *so far*. It doesn’t include the members or resources of any of the unions we regularly campaign with like, say, United Voice, let alone the wider labour movement in Qld.
That sort of gives you an idea of the numbers involved here and the information and analytical capabilities that make up a modern campaign.
Now, it’s all good and well “knowing” about these things and having built up a campaign volunteer base, a structure and an organisational capability to enable it from top to bottom – but what does it mean in terms of doing? Info without action is just a spreadsheet.
Let’s take a hypothetical, integrated example – but instead of using the State level booth catchment data broken down into the SA1 census level or its current projections (because that’s one of our 3 actual secret pieces of info), we’ll use the Federal equivalent. So the voting data here is 2010 Federal election data (broken down into the old SA1 equivalent called CCD level), and we’ll use some 2011 Census data too – simply to show how it all comes together.
And we’ll use the seat of Moreton, in the southern area of Brisbane to keep it relevant.
Let’s say the government launched a policy that materially affected the welfare of families with dependent children in such a way that the more dependent children a family had, the larger the overall deleterious effect on the household’s finances and general well being. To keep it visually simple, we’ll just use Moreton as an example here, but it’s infinitely scalable in real time, so this can happen across every part of every electorate we happen to want to target.
Here’s just a corner of the Moreton electorate and the booth results of the 2010 election. Blue booths are where the Coalition won a two party preferred majority, red booths are where the ALP won a two party preferred majority:
Now let’s say we’re only targeting swing voter areas, but with an initial focus on LNP leaning swing voter areas (where the message would be more persuasion driven) as we would have a separate program for ALP leaning swing voters (where the message would be more reinforcement driven). So let’s add the CCD level booth catchment breakdowns for these swing voter areas held by the LNP on margins of 4% or less.
The blue shaded regions are the areas where the estimated ALP two party preferred vote sits between 46% and 50%. Now let’s overlay the census data, where the proportion of the population that are dependent children is 30% or higher (showing us areas where there are large numbers of dependent children for the average family)
The hatched areas which sit on top of the blue areas would be the locations which produce the most efficient hit rate for the swing voter/dependent children target matrix. Also remember that we could widen or tighten our targeting specificity at whim here – this is just an example.
So next we’d overlay the addresses of our campaigners over the top of that map (we’d have a couple of hundred campaigners in the maps shown). We’d then use email and robocalling to contact those campaigners that live either within or next to those targeted areas to ask if they’d be willing to do mail drops into their local neighbourhoods. Those that accept would receive the material with a map of their drop target.
Total turnaround time between policy announcement and customised material hitting the homes of targeted cohorts? Between 18 and 36 hours.
With the key point being nearly all of this is automated and where the number of seats we can cover is only limited by the amount of material we can print. We have a lot of printers.
Now imagine each targeted seat having a number of these neighbourhood based campaign programs happening every week, on multiple issues, in real time as they arise. And not just with previous election data like what is above, but using the current dynamic voting data. And that’s on top of our other activities.
“What sort of activities” I hear you ask?
Well this sort of thing doesn’t just apply to mail drops, but also to the important yet often forgotten area of face to face voter contact. We not only know where to send campaigners to door knock, but who to send, to deliver what messages (often according to who answers the door – male, female, age, with/without dependent children etc), on which particular areas of privatisation that resonate the most when it comes to voting intention of that particular person.
Yet our campaigners aren’t only campaigners, they’re also an enormous local intelligence network.
For example, let’s say a local LNP member in a targeted seat sets up for a community stall somewhere to meet and greet their constituents. A general Saturday morning political institution in the suburbs across Australia.
We’ll know about it within minutes.
We’ll email and robocall every campaigner within 5 km of that LNP Member’s stall and within 20 minutes of finding out, have a line of local constituents 20 deep asking them important and uncomfortable questions, pretty much all day.
Within an hour, we’ll have an anti-privatisation stall set up right next to them. Stalls that are set up in areas where local residents have high opposition rates to privatisation and exhibit (through our polling) a high coefficient for privatisation issues on their voting intention – i.e. privatisation has a high vote driving effect – then the larger those stalls will be and the more material and human resources will be put into them.
Yet we not only have to react, we can take the campaign to MPs. For example, let’s say an MP says something about the privatisation of aged care. They can expect to see a bus full of ladies over 70 years of age drawn from the local community, occupying the MPs office and having a rollicking good time in the process until the Member answers their questions. Unlike most campaigns, we have a large number of older Australians wanting to be politically active over this – often for the first time in their lives. They’d be pretty keen to get on TV.
These are just a few simple examples of what are really an almost infinite number of campaigning options that utilise the full spectrum of information we have – we already have a formidable list drawn up that’s being added to every day, some simple, some complex, some tactical and some strategic. Now that “we know”, the things we can and will do are not just much larger in scope, but also more efficient and focused and much more sophisticated.
We know who to target, where, why, with what message based on their demographics, their geography, using which specific areas of the privatisation of public services that are most salient and relevant to their lives, at the local community level, delivered by their neighbours, by their friends and by their families across all of Qld.
But most importantly, at the very center of it all will be local constituents, campaigning locally, defending what is a *super majority* of Queensland public opinion.
On the more strategic side, our research program has also allowed us to identify those seats where our campaign can make the difference between the seat remaining LNP or moving to anyone else.
Putting large amounts of resources into seats that will fall anyway would be a waste. Instead, we’ll be hunting up the pendulum – sometimes so far up the pendulum you need to take a packed lunch – because those are the seats that the research suggests can be moved. We’re not sure how many we’ll target yet – that call will be made soon enough – but it’ll likely be somewhere between 20 and 40.
This is just the ground game aspect of the campaign. There’s also the usual advertising, focus groups, comms research, media stuff, online etc etc. The bits and pieces that every campaign has.
However, the thing every campaign doesn’t have is a formidable, research intensive, community level ground game consisting of tens of thousands of active local participants.
Essentially we’re just applying the best of the resources of the early 21st century, to early 20th century community organising. But what that actually requires under the bonnet is a level of political information and analytics that’s never really been needed by a campaign before, so that it can be transformed into the knowledge required to enable the people with the real skills – our union campaign folks and our organisers – to achieve magnitudes of order more than they ordinarily could do without it.
We might be the first organisation to do this in Australia – but we’ll only be the first.
While that’s pretty exciting in many ways, the downside is that the political knowledge gap between the public (including political observers) and key parts of the political system will grow even further apart.
And that is probably not a very good thing at all. So this is our contribution to prevent that very thing from happening as best we can – the anatomy of a modern campaign.