06 November 2018

The MQO poll and Party Choice #nlpoli

MQO's quarterly omnibus poll shows some curious changes in public opinion about provincial political parties.  The province-wide numbers are not curious:  the changes in a couple of the regions are.

Let's take a look first at the provincial numbers.  As usual, SRBP presents the results as a share of all responses, including the refused/undecided/no answer folks.  That's why the numbers here are different from the ones used by MQO in its own release and reported by local news media.  Go to the end of the post for a brief discussion about the way the data is presented.

MQO kindly provided the data tables for this analysis free of charge.  The data collection methodology is theirs.  Interpretation of the data here is solely SRBP.  There is no business or other relationship between SRBP  and MQO to create a conflict of interest.

Contrary to media reports, support overall in the province for the governing Liberals didn't change in the last two quarters of 2018.  The Liberals had the support of 29 percent of respondents in Q3 and 28 percent in Q4.  Support for the Progressive Conservatives declined by four points - 25 to 21 - while support for the NDP went from 11 percent to nine percent.  The number of respondent who refused to answer,  were undecided, or planned not to vote grew by five points (from 36 to 41 percent of responses).

Right off the bat you can see why it's important to look at all responses, not just those who have made a definite choice for one party or another.

Support for the Liberals did *not* grow.  It stayed the same.  Support for the opposition parties dropped slightly from quarter to quarter, with those folks apparently becoming undecided in their party choice.

We don't have any clue from the polling data why the drop happened in the opposition. If you think about events in the province over the past six months, the only significant positive change - potentially - for the Liberals would be the introduction of legal cannabis.  That should have caused a boost in their numbers, not a flat line.

Otherwise,  the news might lead you to expect a downward trend overall in support for the Liberals. Polling started just after Thanksgiving and continued for about a week after the House of Assembly opened to deal with accusations of harassment involving two former members of the Liberal caucus.  On the other side,  PC leader Ches Crosbie won a by-election in Windsor Lake district before polling started for the omnibus.

Yet while we could imahine and increase or decrease in Liberal support the poll showed no change.


The Longer Trending

While there wasn't much change from quarter to quarter according to MQO,  you need to see party choice numbers from all the polls over a longer period.

Liberals might be comforted by the notion that they have opened a wider gap with the PCs.  That would be a mistake.  They are less than 10 points in front and that's a distance that can change quickly during an election.  Since there's between six and 12 months before the next election,  there's plenty of time for things to change and wipe out that lead the Liberals enjoy now.

Since the middle of 2016,  Liberals have been consistently below 30% in every poll except one.  That's clearly an anomaly.  Notice that there have been three periods where the Tories were in the lead. The Conservatives have been on a downward trend since June but that could change as rapidly as it changed twice before since the last election.

Curiouser and curiouser

Some strange things happened in the MQO poll over the past two quarters.  The Liberals and PCs swapped places in St. John's(Libs + 5, PCs -7) shortly after voters in Windsor Lake returned the PC leader as their member in the House of Assembly.  There was also a curious climb by the Liberals and a corresponding drop by the PCs in the Avalon outside St. John's.

That St. John's change is a head-scratcher. Polling before (MQO Q3 and CRA results) and during the Windsor Lake by-election (MQO) suggested the Liberals were ahead of the Liberals overall and doing well in the by-election.  While we can waste time pondering the cause of the flip, let's take them at face value.  Take it as a reminder that kind of swing seen here - 10 points in a short time - can happen.  And it would be enough on a provincial basis to make the next election into a minority government easily enough for either the Tories or the Grits.

If you think the St. John's result is weird, take a look at the MQO regional results for the rest of the province.  Notice especially western Newfoundland and the four districts in Labrador.  This has been the Liberal base, with the province in the last general election being deeper and deeper red the farther west you travelled.

MQO found a 20 point drop in Liberal support from Q3 to Q4, a slight drop in PC support,  a slight growth in NDP support but with most respondents becoming undecided, refusing to answer or saying they planned not to vote.

That's just breath-taking. Maybe it helps explain last week's announcements on the west coast during November when Corporate Research Associates is collecting data for its quarterly poll.

Meanwhile, in Central,  the changes were minor and left the Liberals and PCs with the support of 27 percent of respondents each.  That would likely put a few Liberal seats in jeopardy if those numbers turned up during an election.

The Undecideds

Pay attention to this category of poll results.  No one should discount them.

In a poll, they are undecided,  refused to answer, or don't know.  In the voting world, they are a mix of two types:  folks who do not vote at all and folks who usually vote but who are right now not happy with any of the three choices.

When the UND is around 40%,  just split it in half.  That means about 20% of respondents are potentially available to the parties as persuadable voters.  They won't necessarily split the same way as existing decideds, although many polling firms work on that assumption.  Put that 20% in play and you have the potential for a surprise ending to the next election.  Put just 10% in play and you could make a huge difference in a seat here or there.

Always remember,  the Liberals received less of the popular vote in 1989 than the Tories but won 34 seats.  That's because the Liberal vote was distributed more evenly - it was more efficient - than the Tory vote.  Same thing could make a difference in the next election even if only a fraction of what shows up as Undecided in a poll turned up at the voting booth.

The Take-Aways

1.  With so many people unwilling to state a preference  and with the example of big swings on a regional basis, no one can rest on their laurels.

2.  By the same token, no one can predict the future with any confidence.

3.  Notice,  though, that the PCs have an apparently firm base across the province.  With the exception of St. John's,  support for the party has remained relatively steady over the last two quarters in the MQO polling.

4.  The volatility in the MQO numbers do not put the governing Liberals in a very comfortable position if - as the scuttlebutt has it - the party is looking at a spring election.

5.  The New Democrats are not a factor anywhere except in a few seats in St. John's. There's little chance the party leadership is ready to make the changes needed to be genuinely competitive even in the metro St. John's area.

6.  With the polls running like this, we can look forward to the CRA quarterly omnibus the first week of December.

A note on polls and results:  No polling company in Canada polls voters.  They poll people who are eligible to vote. The popular convention is to report only those who made a specific party choice and equate that with the popular vote results on polling day.  Even with that convention, a string of general elections in the past decade have shown that popular polls that show a horse-race type result have a limited value beyond the most general when it comes to forecasting what people will do in an actual election.  

SRBP uses all responses as valid, not just those that express a clear choice for a political party.  This allows us to see the ebbs and flows of opinion more clearly and, when possible, identify possible correlations between shifts in opinion and events that could be driving changes.  

All those responses that do not select a party are taken together as "Undecided".  Historically in Newfoundland and Labrador,  this broad category includes individuals who have never voted or who have voted extremely infrequently in their lifetime.  No one has ever quantified chronic non-voters in Newfoundland and Labrador.  SRBP uses the estimate that they make up about 10 to 15% of the "Undecided" to represent people who have never voted or who vote infrequently (every 10 to 15 years or when there is a major issue).  This is not supported by any evidence.  It is  - at best - an guess informed by observation over time.  As such, it is prone to error.

The "Undecideds" also include people who usually vote but who, at the time of polling, have decided not to express the choice they otherwise make or would make. They tend to vote although they may have sometimes not voted on one or more occasion in the preceding 20 years.

MQO samples seem to be weighted to include more older people than the population as a whole.  This should make them more useful for estimating vote behaviour since research suggests that older people are more likely to vote than younger people.