Thursday, October 16, 2014

Polling methodology seems to be playing a roll in Massachusetts gubernatorial survey results

IVR polls are 6 points better for Baker/GOP
by BRENT BENSON

Polling averages of the Massachusetts gubernatorial race between Republican Charlie Baker and Democrat Martha Coakley show Coakley with a small 2 point lead, but there seems to be a major difference between the polls conducted using IVR (Interactive Voice Response) methods, and the polls that use traditional polling methods of operator-based dialing, or newer Internet-based survey techniques.

IVR polling has the significant disadvantage of only being able to reach respondents with land-line telephones—federal law disallows robocalls to cell phone numbers. IVR pollsters like Rassmussen and Massachusetts-based Emerson College Polling Society use various sampling and weighting techniques to attempt to compensate for their limited reach.

Breaking down all of the Massachusetts gubernatorial polls conducted since the September 9 primary by polling methodology (Internet, IVR, and Traditional) shows a large gap between the non-cell-phone IVR polls, which have Republican Charlie Baker up 3 points, and the traditional and internet-based polls, which have Democrat Coakley up by 3 points. The boxplot below gives a visual representation of the poll distributions.

There is some chance that these variations are due to sampling error or other causes, but the variations could very well be due to the IVR pollsters not reaching cell-phone-only voters who tend to vote more for Democratic candidates.
MA Gov Polls by polling method (graph)

MA Gov Polls by polling method (chart)

Monday, October 13, 2014

Recent polling shows a robust coalition against repeal of Massachusetts casino law

Most Republicans, Independents, and men support gaming law—Democrats are split down the middle
by BRENT BENSON

The casino issue has a history of splitting coalitions in Massachusetts with religious Republicans vs. big business Republicans and social justice Democrats vs. union Democrats, but recent polls show more consistency in pro-casino sentiment—likely enough to defeat an anti-gaming ballot initiative.

Opponents of the Massachusetts's gaming law, which allows for construction of up to three destination casinos in the Commonwealth, have sponsored Ballot Question 3 to repeal the law. However, casino supporters have lead in all ten of the post-primary surveys of the ballot question by an average margin of 11 points.

Polls show a strong cohort of casino supporters consisting of Republicans, Independents, and men who, on average, oppose repeal by 16 to 17 points, roughly 55 to 38 percent. Women, on average, oppose repeal by a smaller 6 point margin—48 to 42 percent. Democrats—who have strong idealogical factions both for and against expanded gaming—are split down the middle with approximately 45 percent on either side of Ballot Question 3.

Graph of MA Question 3 polls

Polls of MA Question 3 on casino law repeal

There have been ten public surveys published since the September 9 state primary that have tested the casino ballot question. All of the polls had sample sizes in the range of 400 to 504 likely voters.

The post-primary polls range from a 4 point margin against repeal in the 9/21-9/23 Boston Globe/Social Sphere survey, to a 20 point margin against repeal in the following 9/24-9/27 WBUR/MassINC poll. The time-weighted average, which gives higher weights to more recent polls, shows an 11 point margin against gaming law repeal—40 percent for repeal and 52 percent against.

There are strong majorities of Republicans, Independents, and men against repealing the casino law, with average margins of about 16 points across these voting subgroups. Over 50% of Independents opposed repeal in every single poll, and over 50% of men opposed repeal in all but one poll, the most recent Boston Globe poll, which showed 49% of men in opposition.

The polls show women opposing the repeal of the casino law by a smaller 6 point margin with 48% opposing repeal and 42% in support of repeal. More women were against, rather than for, the repeal in all but one survey, the closer-than-average Boston Globe/SocialSphere poll with 9/21-9/23 field dates, which had a 4 point margin for women who want to see the casino law repealed. There are a larger group of undecided women voters than most other categories (10%).

Democrats are almost evenly divided on the casino question with a 1 point average margin against repeal. There is a strong anti-casino coalition in the Democratic party which campaigns against the social costs of institutional gambling, while a largely union-based Democratic coalition supports casino construction as a job creation measure. These opposing coalitions seem to be canceling each other out. The comparatively large group of undecided Democratic voters (10%) may also indicate an ambivalence or divided loyalties on the casino question.

So far, the anti-casino wing of the Democratic party does not have enough support to overturn the Commonwealth's gaming statute. Pro-Question-3 groups will need to do a better job of making their anti-casino case to the public in order to achieve repeal on November 4.

Tuesday, September 30, 2014

Massachusetts 2014 Governor's race general election polling averages

A look at all of the public polls with topline, gender, and party averages
by BRENT BENSON

Last updated 10/24/2014 with Boston Globe/SocialSphere poll.

This page is tracking the public polls conducted on the 2014 gubernatorial election to replace Governor Deval Patrick. It includes all of the publicly-release non-partisan polls and has averages for the topline horse-race number, as well as averages for gender and party affiliation.

The topline results have Attorney General Martha Coakley clinging to a 1 point lead in the time-weighted average that gives more weight to polls closer to the election. The last five polls have shown Coakley +3, a tie, Baker +2, Baker +1, and Baker +9. The pendulum has been swinging in Baker's direction, but it will take additional polls with Baker up to move the average.

Poll Size Type Start End Baker Coakley
Total M F Dem GOP Ind Total M F Dem GOP Ind
WBUR 504 LV 9/11 9/14 35 42 28 11 74 42 44 37 50 68 12 35
Globe 407 LV 9/14 9/16 36 43 30 8 83 43 39 32 46 76 6 24
Rasmussen 750 LV 9/16 9/17 42 42
WBUR 502 LV 9/16 9/21 36 44 28 13 75 43 46 38 53 70 10 38
UMass 440 LV 9/19 9/23 46 55 38 7 99 68 45 36 53 87 1 19
Globe 400 LV 9/21 9/23 40 50 31 10 94 45 38 28 46 75 2 23
YouGov 2389 LV 9/20 10/1 41 50 34 8 91 52 47 41 53 84 4 30
WBUR 503 LV 9/24 9/27 41 46 36 20 75 46 44 39 50 68 13 35
Herald 500 LV 9/25 9/28 43 49 38 20 79 52 44 38 49 71 9 33
Globe 401 LV 9/28 9/30 39 42 35 11 89 44 36 33 39 71 6 20
UMass 414 LV 9/26 10/2 44 56 33 9 95 56 48 37 58 87 0 33
WBUR 504 LV 10/1 10/4 39 46 32 17 78 45 41 40 42 67 9 31
Globe 400 LV 10/5 10/7 34 37 30 11 87 36 39 36 42 69 5 26
Emerson 500 LV 10/6 10/7 45 49 42 22 74 56 40 35 44 68 18 24
WBUR 500 LV 10/8 10/11 39 45 33 18 71 46 42 37 47 67 14 32
Globe 400 LV 10/12 10/14 41 48 34 13 83 49 41 36 45 80 3 23
Rasmussen 980 LV 10/13 10/14 48 46
WBUR 501 LV 10/15 10/18 43 50 36 18 76 52 42 35 49 66 13 32
Globe 400 LV 10/19 10/21 46 55 39 13 93 57 37 30 43 74 3 20
TW Average 9/11 10/21 41 48 34 14 83 49 42 36 47 73 8 28
Difference 9/11 10/21 -1 12 -13 -59 76 21 1 -12 13 59 -76 -21

While the polls have been showing a very close race between Coakley and Baker with somewhat inconsistent results, the share of the vote held by the third-party candidates has remained very consistent around 5%.

MA Gov Polling graph

Gender

Attorney General Coakley has held a strong lead with women throughout the primary and general election. Charlie Baker has improved his margin with men, but Coakley is holding a slim lead in the gender gap, up by 13 points with women, while Baker has an 12 point advantage with men.

Poll Size Type Start End Baker Coakley
Total F M Total F M
WBUR 504 LV 9/11 9/14 35 28 42 44 50 37
Globe 407 LV 9/14 9/16 36 30 43 39 46 32
Rasmussen 750 LV 9/16 9/17 42 42
WBUR 502 LV 9/16 9/21 36 28 44 46 53 38
UMass 440 LV 9/19 9/23 46 38 55 45 53 36
Globe 400 LV 9/21 9/23 40 31 50 38 46 28
YouGov 2389 LV 9/20 10/1 41 34 50 47 53 41
WBUR 503 LV 9/24 9/27 41 36 46 44 50 39
Herald 500 LV 9/25 9/28 43 38 49 44 49 38
Globe 401 LV 9/28 9/30 39 35 42 36 39 33
UMass 414 LV 9/26 10/2 44 33 56 48 58 37
WBUR 504 LV 10/1 10/4 39 32 46 41 42 40
Globe 400 LV 10/5 10/7 34 30 37 39 42 36
Emerson 500 LV 10/6 10/7 45 42 49 40 44 35
WBUR 500 LV 10/8 10/11 39 33 45 42 47 37
Globe 400 LV 10/12 10/14 41 34 48 41 45 36
Rasmussen 980 LV 10/13 10/14 48 46
WBUR 501 LV 10/15 10/18 43 36 50 42 49 35
Globe 400 LV 10/19 10/21 46 39 55 37 43 30
TW Average 9/11 10/21 41 34 48 42 47 36
Difference 9/11 10/21 -1 -13 12 1 13 -12

There has been a great deal of variability in the gender breakdowns as you can see in the graph.

MA Gov Polling by Gender graph

Party

The post-primary polling predicts that the party breakdown of the voters on November 4 should be about 37% Democratic, 13% Republican, and 50% Unenrolled. In order to win, Charlie Baker will need win a large majority (probably two-thirds) of the Unenrolled voters and to make some inroads with Democrats. The polling so far is promising for Baker with an average of 14% of Democrats and climbing towards 50% of independent voters. On the other hand, if Coakley can consolidate the Democratic vote and keep it close with independent voters, she should be able to win in November.

Poll Size Type Start End Baker Coakley
Total Dem GOP Ind Total Dem GOP Ind
WBUR 504 LV 9/11 9/14 35 11 74 42 44 68 12 35
Globe 407 LV 9/14 9/16 36 8 83 43 39 76 6 24
Rasmussen 750 LV 9/16 9/17 42 42
WBUR 502 LV 9/16 9/21 36 13 75 43 46 70 10 38
UMass 440 LV 9/19 9/23 46 7 99 68 45 87 1 19
Globe 400 LV 9/21 9/23 40 10 94 45 38 75 2 23
YouGov 2389 LV 9/20 10/1 41 8 91 52 47 84 4 30
WBUR 503 LV 9/24 9/27 41 20 75 46 44 68 13 35
Herald 500 LV 9/25 9/28 43 20 79 52 44 71 9 33
Globe 401 LV 9/28 9/30 39 11 89 44 36 71 6 20
UMass 414 LV 9/26 10/2 44 9 95 56 48 87 0 33
WBUR 504 LV 10/1 10/4 39 17 78 45 41 67 9 31
Globe 400 LV 10/5 10/7 34 11 87 36 39 69 5 26
Emerson 500 LV 10/6 10/7 45 22 74 56 40 68 18 24
WBUR 500 LV 10/8 10/11 39 18 71 46 42 67 14 32
Globe 400 LV 10/12 10/14 41 13 83 49 41 80 3 23
Rasmussen 980 LV 10/13 10/14 48 46
WBUR 501 LV 10/15 10/18 43 18 76 52 42 66 13 32
Globe 400 LV 10/19 10/21 46 13 93 57 37 74 3 20
TW Average 9/11 10/21 41 14 83 49 42 73 8 28
Difference 9/11 10/21 -1 -59 76 21 1 59 -76 -21

While there have been some differences in the poll-by-poll numbers, the stratification of the party breakdown has been remarkably consistent since the primary.

MA Gov Polling by Party graph

Thursday, September 25, 2014

Globe poll showing Baker lead has a surprisingly large leap in male support

Poll shows Coakley/Baker male gender gap greater than any statewide GOP winner in last 16 years
by BRENT BENSON

A new poll from the Boston Globe conducted by SocialSphere shows Charlie Baker with a two point lead over Martha Coakley among 400 likely voters in the gubernatorial race to replace Governor Deval Patrick. This is the first time Baker has led in a public poll.

While a key finding in all previous polling has been Martha Coakley's large lead over Baker among women (varying from 15 to 25 points in post-primary polling), the new Globe poll shows a striking leap in Baker's lead with men of 50% to Coakley's 28% for a gap of 22 points.

 Poll Size Type      Start        End  Baker       Coakley      Diff
                                           M    F        M    F    M    F
 WBUR  504   LV 2014-09-11 2014-09-23   42.0 28.0     37.0 50.0   +5  -22
Globe  407   LV 2014-09-14 2014-09-16   43.0 30.0     32.0 46.0  +11  -16
 WBUR  502   LV 2014-09-16 2014-09-21   44.0 28.0     38.0 53.0   +6  -25
Globe  400   LV 2014-09-21 2014-09-23   50.0 31.0     28.0 46.0  +22  -15
                     Average            44.8 29.2     33.8 48.8  +11  -20

This 22 point lead among men is twice the average lead of 11 points seen across all of the WBUR/MassINC and Globe/SocialSphere post-primary polls.

The largest margin seen by a statewide Republican winner in Massachusetts in the last 16 years was 15 points by Paul Cellucci in 1998. Romney's male margin was 13 points over Shannon O'Brien in 2002, and Scott Brown's margin among men was 14 points over Martha Coakley in the 2010 Senate special election. A 22 point margin is unprecedented in recent history.

Coakley/Baker polls by gender graph

It remains to be seen in subsequent polls whether Baker will continue to show this level of support with likely male voters. In the meantime, it is safer to look at the averages showing Martha Coakley with a small, shrinking, but likely lead.

       Poll Size Type      Start        End  Coakley  Baker  Other  Undecided
       WBUR  504   LV 2014-09-11 2014-09-23     44.0   35.0    4.0       15.0
      Globe  407   LV 2014-09-14 2014-09-16     39.0   36.0    6.0       19.0
 Rassmussen  750   LV 2014-09-16 2014-09-17     42.0   42.0    5.0       10.0
       WBUR  502   LV 2014-09-16 2014-09-21     46.0   36.0    4.0       12.0
      Globe  400   LV 2014-09-21 2014-09-23     38.0   40.0    5.0       18.0
    Average           2014-09-11 2014-09-23     41.8   37.8    4.8       14.8

Tuesday, September 16, 2014

The rise of internet polling—more adults have internet access than a landline telephone

One surprising graph tells the story
by BRENT BENSON

There is an ongoing battle in the polling and survey community about the relative benefits of live caller polling vs. automated polling vs. internet polling. Live caller polling is considered the gold standard polling, in large part because human survey operators are allowed to call cell phone numbers, giving them a more representative sample of respondents. Pollsters who use automated systems are prohibited from calling cell phones, and therefore resort to extensive demographic-based weighting to compensate. However, live caller polling is expensive.

There has been an uptick in internet-based polling by pollsters like YouGov that have their own sampling problems, but are able to reach large panels of respondents and use weighting to achieve accurate results. There is a nice overview of internet-based polling by Nate Cohn in the Upshot.

The following graph shows why Internet polling may be the wave of the future. The percentage of US adults that use the internet passed the percentage of adults that have a landline phone in 2010 and is approaching 90%, while the number of adults with a landline has fallen to 60%.

Landline phones vs. Internet use graph

The combination of lower response rates for tradition telephone-based polling, the steady rise of internet usage by the U.S. population, and the refinement of online panel weighting techniques, makes it likely that we will see increased reliance on internet polling in the coming years.

Data

  • Internet usage data: Pew Research - Percent of U.S. adults who use the internet
  • Landline telephone usage data: CDC - Percent of U.S. adults with landline phone
  • Raw data spreadsheet

Wednesday, September 10, 2014

Charlie Baker's underperformance was three times greater than Martha Coakley's

Baker underperforms polling by 24 points, Coakley by 8
by BRENT BENSON

The performance of the primary candidates for Massachusetts Governor on September 9 seem to indicate a voter turnout similar to that predicted by the Boston Herald/Suffolk University poll of the primary released on August 25. The poll had a very tight screen for likely voters, requiring that the respondents be able to identify—in some fashion—when the primary was occurring.

The Herald/Suffolk poll showed a tighter race than other polls on both the Democratic and Republican sides. While all of the polls got the ordering of the results correctly, the Herald/Suffolk poll came closest to the actual margins.

After the election pundits like Scot Lehigh posited that Coakley's underperformance compared to the polls indicates weakness, while there is very little talk about how much Charlie Baker underperformed against Tea Party rival Mark Fisher when compared to the same polls.

Let's look at the numbers.

Performance on nominees compared to polls

If you take the Herald/Suffolk poll results and normalize them into election results by removing the undecideds, it predicts Coakley at 48% and Grossman at 34% for a 14 point win. The preliminary results show Coakley with 42% and Grossman with 36%, a 6 point winning margin.

On the GOP side the poll predicted Baker with 86% and Fisher with 14%, a whopping 72 point win. The preliminary results have Baker with 74% and Fisher with 26%, a 49 point margin.

If pundits are going to make an issue of Coakley's 8 point underperformance, then it also needs to apply the same standard to Charlie Baker who underperformed by three times as much.

Tuesday, September 9, 2014

What are the regional strengths of the Democratic gubernatorial candidates?

What polling tells us about areas of strength and weakness for the primary candidates
by BRENT BENSON

It is primary election day in Massachusetts. While we wait for tonight's election returns I decided to look at the regional performance of the Democratic candidates for Governor to see if there were any surprises.

I used the cross-tabs from the last three Boston Globe/SocialSphere polls because they used the same regional breakdowns and gave a reasonable sample size when averaged across all the three polls. I could have gone back farther to get a larger sample, but I wanted a relatively recent snapshot.

My table shows how each candidate performed across the complete set of three polls and all regions (the All column), and for each region provides the candidate's percentage, and a ± number showing how regional performance compared to overall performance. There is a proportionately-sized blue bar for over-performance compared to the candidate's overall performance, and red bar for under-performance.

MA Gov Dem candidate regional performance

The Metro Boston region was closest to the state as a whole with Martha Coakley even, Steve Grossman -2, and Don Berwick +1, and also an almost average percentage of undecided voters.

Martha Coakley did well in the Cape and Islands, Merrimack Valley, and South of Boston regions (+5, +6, +7) and poorly in the Inside 128, Metro West, and South Shore regions (-7, -7, -10).

Steve Grossman excelled in the Merrimack Valley, Inside 128, and South Shore regions (+5, +9, +12), and had more trouble in Western Massachusetts and in the Cape and Islands (-8, -9). Grossman actually tied Coakley on the South Shore, where he was +12 and Coakley was -10.

Don Berwick did exceptionally well in the Metro West region at +9 and tied Steve Grossman who was -4 west of Boston. He did not do as well on the North Shore or South Shore (-6, -8).

The Western Massachusetts region had the highest number of undecided voters at 31% (+12), while the Merrimack Valley region had the fewest with 9% (-10).

We can keep an eye out for unexpectedly high turnouts on the South Shore—which would improve Steve Grossman's overall performance—or in Metro West—which would be great for Don Berwick. However, Martha Coakley's polling margins are big enough in enough regions that strong regional turnouts are very unlikely to change the final result.