Newspoll attitudinals

I do not normally pay much attention to the attitudinal polling. I find it interesting, but not overly informative. The following charts are the latest attitudinal results from Newspoll, published in the Australian.

First, let's look at satisfaction with the Prime Minister.

Second, satisfaction with the Opposition Leader

Now let's compare the net satisfaction results.

And finally, the better prime minister result.

Newspoll 42 to 58 in Coalition's favour

There is a bit of noise in the most recent polls, but the overall message is a move away from the government.

• Newspoll has moved 6 points to the Coalition over the previous fortnight. But if you concede that the previous fortnight was a little rogue, it has only moved 3 points on the previous month.
• Morgan has moved away from the government two points on the previous week, but it is only half a point down on a fortnight ago.
• Galaxy is unchanged on three weeks ago.
• Essential was unchanged on the previous week, but up a point for the government on the previous fortnight, and up two points on a month ago.

Graphically, this looks like ...

Let's have a closer look at that Newspoll.

Turning to the Bayesian aggregation, it looks like we have a new minimum, and the potential for an even larger landslide election result than we have seen for some time.

A galactic aggregation

Today, the Murdoch press had the latest Galaxy poll. It was pretty much unchanged from the previous Galaxy poll.

My workspace

From time to time I am asked how I do my charts. I have answered this question at length on my other blog. In this post, I want to focus a little more on the hardware side of things.

Currently, I use a late-2012 iMac with a 3.4 GHz processor and 16Mb of ram. This is my third Apple Mac computer. The previous two were MacBooks. Before that I was a Linux/Ubuntu fan. (I still use Linux occasionally to manipulate data on the Parallels virtual machine).

Connected to the iMac are two Apple Thunderbolt Displays and one generic 27 inch monitor. The generic monitor connects through a J5 Create USB3 to HDMI adapter.  Often I use one of the monitors as a TV with an Elgato Eye-TV tuner.

For sound, I have two floor-standing speakers (at each side of the picture) and a sub-woofer (under the desk), all driven by a Denon 2113 AV Receiver (on the filing cabinet). The remote control for the Denon is on the desk. I use a Toslink optical cable to get the sound from the iMac to the AV Receiver.

On the desk you can see a DVD drive, a blu-ray drive, a silver 2TB USB3 back-up hard drive (time machine), and an orange 256MB Thunderbolt SSD drive (where most of my charting software lives, so I can take it with me when I travel with the MacBook).

Also on the desk there is a Microsoft wheel mouse (my preferred pointing device), an Apple magic mouse and an Apple track pad. Harder to see is an iPhone 4S on charge and a USB2 hub.

Just visible on the bookcase to the right is an Apple Airport Extreme (there is an Airport Express elsewhere in the house that ensures WiFi coverage from the front of the House to the rear). Hiding behind the left desk monitor is an ADSL modem. Behind the AV Receiver is an 8-port hub that distributes internet to other rooms in the house.

What's not in the photo? Behind me is a network printer and a windows machine (only used occasionally). On each side are six bookcases (three a side) spanning my eclectic interests: politics, economics, mathematics, computer science, anthropology, and sci-fi/fantasy fiction. On top of the bookcases are boxes of computer parts (from many years of building my own machines, something I no-longer do). Also not in the photo is my iPad (which gets used every day). 

This is where I prepare my charts.

Recap: the last days of Kevin Rudd (2010)

With all the drama of this week, let's look at the polls in the six months prior to Kevin Rudd's removal from office on 24 June 2010. All of the data for that period is strongly suggestive that Rudd was recovering from his May 2010 slump. This recovery, between May 2010 and June 24 is most evident in the Bayesian aggregation for the period.

The June 2010 recovery is pretty easy to see in the raw data.

At the very least (after adjusting for collective house effects) it would appear that Keven Rudd was removed from office on that fateful morning in June 2010 on a poll winning 52 per cent of the two-party preferred vote. We know what happened next.

Update

Kevin in the comments below makes a valid point. The last few data points above slide into Julia's elevation bounce. I have changed the cut-off date from 24 June to 20 June and re-run the analyses. The key charts with this change follow.

I should point out that the observation count in the LOESS charts is the number of LOESS data points in the analysis. Where there are two polls on a date, only one point in the LOESS series is calculated.

Morgan's multiple modalities

Morgan has described four of its polls to date as “multi-mode”.

• The first multi-mode poll was published on 5 March. It had a whopping sample of 9101 respondents. It comprised “web-based surveying combined with face-to-face surveying.
• The second multi-mode poll, published on 12 March, also comprised “face-to-face interviewing and … online surveying”.
• The third multi-modepoll, published on 20 March, saw a different methodology. It was "conducted via face-to-face interviewing, online surveying and via SMS polling". 
• The fourth multi-mode poll, published on 22 March, gave rise to a third methodology. It was solely conducted by SMS polling.

Analytically, this flexibility (some would say inconsistency) in methodology makes it difficult to benchmark the house effects that might arise in Morgan's multi-mode polls. In plain-English, with each Morgan multi-mode poll we may well be comparing apples and oranges. 

Because these flexibilities affect the underpinning logic for the Bayesian aggregation, I am seriously thinking of removing all Morgan multi-mode polls from the aggregation.

Up, down or round and round

What are the polls doing? Is Labor heading south, scraping the bottom or on the rebound? I ask the question because today's Nielsen looks like a bottom feeding story and today's Essential is a rebound (at least over the last two independent polls).

On the raw data, the individual polling houses tell all of these stories. Morgan and Galaxy have Labor heading south. Nielsen is scraping the bottom. Newspoll and Essential are rebounding.

Regular readers of this blog know that individual poll-to-poll movements are mostly meaningless noise. That is why I aggregate the polls. When I aggregate, the most likely story continues to be one of bouncing along the bottom. The polls are no longer heading south, but nor are they rebounding in any substantial way.

Nielsen update

Today saw the Fairfax media publish the latest Nielsen Poll. It was very much a status quo result. According to Nielsen, the middle of March looked much the same as the middle of February.

Dropping these figures into the Bayesian aggregation, the combined polling story appears to mirror the Nielsen result. Following the voting collapse at the start of 2013, the middle of March looks much like the middle of February.

Newspoll 52-48

The latest Newspoll has a three point bounce in the Government's favour.

I am always amused at how the papers interpret poll-to-poll bounces. Most of the time, most of the movement in a poll-to-poll bounce is random noise. It is without meaning. Most likely, today is no different.

All of the polls agree that the Government is recovering from its February nadir, just not at the rate of three percentage points in a fortnight. The latest Bayesian aggregation follows.

Note of caution: the Bayesian aggregation is very good at picking turning points in the trend. It is less accurate at picking the actual vote share the parties enjoy at a particular point in time. The aggregation model assumes the population trend is the average of the bias across the (now 7) polling houses. My suspicion is that this average is currently biased.

Update

Morgan has released another multi-mode poll. This one was not as favourable for the Government as today's Newspoll. In terms of the how respondents said they would preference, the result was a cataclysmic 57.5 to 42.5 in the Coalition's favour. Using preference flows from the last election, the result was only stonkingly bad for the government: 55.5 to 44.5 in the Coalition's favour. It is this latter figure I use in my Bayesian aggregation.

It is clearly too early to have a view on the stability and inherent house biases of this new polling technology from Morgan. Nonetheless, the Bayesian aggregation has shifted somewhat with this latest data point. The notion of a polling recovery for the government is no longer as clear as it was prior to receiving the Morgan multi-mode poll.