Thinkyness Blog

What does an editorial in The Australian questioning ABC bias, articles analysing election results such as those regarding the recent WA Senate re-run, and analysis of the recently released Intergovernmental Panel on Climate Change (IPCC) report have in common?  All of them make use of the word ‘significant’, which, like the use of  ‘mandate’, is a word used by writers to convey the importance of something to give gravitas to possibly weak arguments.

Looking first at The Australian article published April 5^th 2014. The two paragraphs I noted were:

This simply ignores the recent opinion polls that demonstrate that a significant proportion of Coalition voters regard the ABC as biased or politically imbalanced: 35 per cent according to Newspoll and 46 per cent according to AC Nielsen.

This is a serious problem, which Scott refuses to acknowledge. Not surprisingly, Greens voters are the most supportive of the ABC.

According to the author 35% (according to Newspoll) and 46% (according to AC Nielsen) of Coalition voters consider the ABC as biased or politically imbalanced. The author goes as far as to suggest in the following paragraph this is ‘a serious problem’. I tweeted at the time:

@therealpbarry @marygarden How is 35% of LNP voters regard ABC biased or politically imbalanced more significant than 65% who don't?

— Paul (@davispg) April 5, 2014

I am not suggesting that 35% or 46% of a sample is not a significant result. It may very well be, however significant when used in the context of data or results generally implies ‘substantial’. This means the movement (here a percentage) is significant with regard to all of the data. This distinction, all, is important, because the percentage referred to by the author is a subset of the data, in this case, Coalition voters. 35% or 46% may be significant to the subset (Coalition voters), but if for arguments sake Coalition voters only represented 40% of the total responses, the percentage of voters of the total sample who are both Coalition voters and think the ABC is biased is 14% and 18.4% respectively. Is 14% and or 18.4% substantial and, therefore, significant?

Moving onto election analysis. With the recent WA Election there has been quite a bit of analysis despite the result not yet being finalised. Notably, much of the analysis has centred on the swings away from the Labor and Coalition Parties. What I found interesting is the first preference votes are shifting to parties other than the two majors (Coalition and Labor). At the time of writing (acknowledging the count has not completed), the first preference votes by group were:

Party	Votes	Percentage
Liberal + The Nationals	336,037 + 30,634 = 366,671	33.80 + 3.08 = 36.88
Australian Labor Party	216,190	21.74
The Rest (The Greens, PUP etc)	411,385	38.51
Informal	29,332	2.87

Note the total votes counted so far to everybody else except the Coalition and Labor is 411,385, or 38.51%. I would argue this is more significant than whether the Coalition or Labor have lost a couple of percentage points. The total number of people voting for somebody else other than the “two party preferred” is 38.51% of the votes counted so far. Are we seeing a breakdown of the two-party preferred oligopoly?

Of course it is possible this is just a ‘protest vote’, a swing from both parties to be expected in a ‘bi-election’. This spin was started by both majors early into election night. So lets look at the House of Representatives. The graph below is based on data from the Australian Electoral Commission.

Note the grey line for “The Rest”. In the 2013 election more than 20% of the electorate voted for somebody else other than Labor or the Coalition. The majority of the vote has come from Labor, however the Coalition is not immune as the rate of the Coalition vote increase since 2007 (the slope of the line) is less than that of the votes to “The Rest”. So what is more substantial, both Labor and Coalition being a few percentage points lower, or the size of the votes to “The Rest” growing year-on-year?

Lastly we’ve got the latest Inter-governmental Panel on Climate Change (IPCC) Report, and the analysis that has surrounds the report. Articles included Climate change will damage Australia’s coastal infrastructure (Guardian), Climate change could be putting human health at risk (ABC), Climate change intensifies risk of conflict, migration (SBS), and how humans are changing the climate (SMH). The silence from some of the media was deafening, except for of course the "opinionators". There is nothing like a high-quality, scientifically produced report to bring out the opinionators and it would be reticent of me if I didn't mentioned current king of Australia’s online opinion scene, Andrew Bolt. He published a blog about how the IPCC report was doctored to scare you that appears to rely on arguments contained within a UK Sunday Mail article that was later blasted by the IPCC for being completely misleading. Each of these articles rely on words such as ‘significant’; but is climate change significant?

In terms of the climate system everything is connected. The best scientists in the world are suggesting an increase in global temperatures of just a few degrees will disrupt the earth’s weather patterns. In this example significant (in terms of temperature) is measured in 1 and 2-degree increments, not the large percentages of the first and second example. It seems, according to the IPCC report, human amplified climate change is well underway and, contrary to the naysayers, is already having a significant impact on the planet: more extreme storms (polar vortex anyone), worse fires (Australian fire season starting in October), etc. So, on balance, the use of the word significant in these reports seems to make sense.

I find if I read a document and an author has used the word ‘significant’ or ‘ substantial’ I ask ‘what makes this thing significant/substantial’. If the argument is based on a subset (first example – Coalition voters), the question is how ‘significant’ is the thing against the entire sample. If the argument is based on a percentage (second example – “The Rest”), see what the other percentages are, why has the author mentioned the percentages. If it is based on movements of small amounts (third example, temperature), ask what is it about ‘the system’ that makes it this way.

I find when I do this I find sometimes the author is trying to give gravitas to a possibly weak argument, or there is more to the story.