Polling of Massachusetts Governor's race still a story of mostly unknown candidates

Democratic ballot test performance driven by name recognition of candidate
by BRENT BENSON

On Thursday, March 20 the MassINC Polling Group released a new poll of the Massachusetts Governor's race commissioned by WBUR. While WBUR led with Attorney General Martha Coakley maintaining a large lead in the race, the most important aspect of the poll—the continuing low name recognition of the non-Coakley candidates—was only covered in passing in comments by pollster Steve Koczela in the on-air interview.

In fact, the performance of the Democratic candidates on the Democratic Primary ballot test is defined, and drastically limited, by low name recognition, as the following charts and graph will show. These early polls give us very little information about how these candidates will fair in the September primary or November general election after voters begin to pay attention to the race and get to know and differentiate between the top-tier candidates.

The key number is the 75% of respondents that have an opinion of Martha Coakley, which is the driver for her performance in the primary ballot test against the other Democrats, and in her 41% to 26% lead over Charlie Baker in the general election ballot test. Charlie Baker and Steve Grossman are the next closest candidates with 44% and 29% respectively. The remaining candidates are largely unknown to Massachusetts voters.

The next chart shows the share of the respondents that said they would vote for each of the candidates in the Democratic primary.

There is a strong correlation between the Democratic Primary ballot test percentages and the Has opinion column in the name recognition chart. A linear regression between the ballot test percentage and the name recognition percentage shows the strength of this relationship. While the amount of data is small, you would expect candidates to be well above or below the line if they were over- or under-performing within the pool of voters who knew them.

Martha Coakley's name recognition, combined with her favorable/unfavorable numbers are a positive sign for her prospects in the fall elections, and at some point low name recognition will be a problem for the lesser-known candidates. However, history points towards an tightening of the race as candidates who build statewide organizations have an opportunity to make their case to the voters when the election moves into its next stage and voters are actually paying attention.

A comprehensive history of NCAA Men's Basketball Tournament Final Four seeds from 1985 through 2013

Only 16 out of 116 Final Four teams have been seeded greater than four
by BRENT BENSON

I am taking my once-a-year break from writing about Massachusetts politics to write a sports analytics column, this one looking at how often teams with various seeds end up in the NCAA Men's Basketball Final Four and go on to win the championship.

While working on your bracket you may be asking yourself questions like, "Should I pick all number one seeds to go to the Final Four?," or "Should I only pick long shots to have a better chance differentiating myself from my bracket pool?" While I don't have definitive answers to these questions, it is instructive to look at how many teams of each seed have made appearances in the Final Four, have gone on to the final, and have won the national championship. This is a comprehensive history from 1985—the year the NCAA moved to a 64 team field—through last year's 2013 tournament.

There are a lot of 1s, 2s, 3s, and 4s in this table. In fact, out of the 116 teams that have made the Final Four since 1985, 100 have been teams with seeds 1 through 4, with only 16 teams having higher seeds. There has never been a team with a seed above 11 in the Final Four. It is important to note that this is likely to change as more games are played within this system—events with low probability do happen. However, teams with the lowest seeds are much more likely to make the Final Four.

Let's break this down by round. Here is a table of appearances in the Final Four by seed.

Of the 116 teams that have made the Final Four since 1985, 47 (41%) have held the top seed, 25 (22%) the 2 seed, 14 (12%) have been 3 seeds, and 14 (12%) have been 4 seeds. The remaining 16 teams (14%) have had seeds from 5 through 11. While it may seem mysterious that a 7 seed or a 10 seed has never made the Final Four, we are dealing with a sample size of only 116 and with a percentage that would likely be in the very low single digits—it is likely just a matter of chance. Later we perform a regression that gives us a feel for what expected frequencies for each of the seeds might be.

The median seed for this set of 116 Final Four seeds is 2, the mean of the seeds is 2.68, with a standard deviation of 2.26.

If we consider the seeds of the teams that have made the final championship game, the highest seed is 8 (compared to 11 in the Final Four). There is a drop off from the 1 seed to the 2 seed similar to that in the Final Four round, and only 10% of the teams had a seed below 4. The two 8-seeded teams that made the final game were Villanova in 1985—who went on to win the championship game in spectacular fashion against top-seeded Georgetown—and Butler in 2011. The 2011 Butler team lost to 3-seed Connecticut 53-41.

The median seed of this set is again 2, and the mean of the seeds has inched down to 2.26, with a standard deviation of 1.69.

The seeds of the teams that have won the NCAA championship since 1985 are an exclusive set, with only the 1985 Villanova team having a seed greater than four. Well over half the teams that have won the championship (62%) have been 1 seeds. There have been 4 teams (14%) that were 2 seeds, and 4 teams (14%) that were 3 seeds.

The two teams with 4 seeds that won the championship were Danny Manning's 1988 Kansas team, and the 1997 Arizona Wildcats who came back twice from 10 point deficits in earlier rounds. The median of the championship seeds is 1, the mean is 1.86, and the standard deviation is 1.53.

The seed data from all of the rounds comes remarkably close to fitting an exponential decay regression of the form (1/2)^seed. Here are the same data with a fitted regression line. The regression lines indicate some possible outliers like over-performing 4 seeds in the Final Four, and underperforming 2 seeds for the final championship.

The 28 years of NCAA 64-team-bracket historical data show clearly that choosing a team with a seed greater than 4 to go all the way and win the championship is hoping for a true long shot. However, chances are that a 5 or 6 seed will win a championship if the tournament continues in its current form.

Fifth Middlesex Senate seat likely to remain blue

Redistricting changes removed GOP stronghold of Lynnfield and added Democratic precincts in Winchester and Melrose
by BRENT BENSON

On Tuesday, March 4, Winchester State Representative Jason Lewis won the Democratic special state primary for the 5th Middlesex State Senate seat, which had previously been held by Congresswoman Katherine Clark. Lewis will face Melrose Republican and Alderman-at-Large Monica Medeiros in the general election—all signs point towards a win by Representative Lewis on April 1.

That is not to say the 5th Middlesex Senate District is a bastion of liberalism. A strong GOP candidate who out-worked and out-raised a Democratic candidate could very well take this seat. However, the lean of the district has moved to the left since it was held by GOP State Senator Richard Tisei, Lewis showed strength in the Democratic primary, and Lewis shows no signs of being out-worked or out-raised.

Much of the pushback on my piece predicting a win by Lewis in the primary and in the general was based on an assertion that the district has a Republican lean and that the previous incarnation of the district was help by Republican Richard Tisei for 20 years. I would like to present some specifics about district's shift to the left in the 2011 redistricting that makes it a tougher win for the GOP.

The Middlesex and Essex District

Congresswoman Clark was originally elected to the State Senate in the Middlesex and Essex State Senate District, the previous incarnation of the 5th Middlesex. Clark was elected to replace Republican Richard Tisei who had served the district for 20 years and who left to run for Lieutenant Governor as Charlie Baker's running mate in 2010.

Before redistricting, the Middlesex and Essex State Senate District consisted of Lynnfield, Malden, Wards 1 through 5 of Melrose, Reading, Stoneham, and Wakefield. We can look at the 2008 performance of President Obama in the precincts of the district to see how they compare to the rest of Massachusetts. The first column shows President Obama's 2008 margin in district precincts in each of the municipalities; the second column subtract's Obama's statewide margin, normalizing it for the Commonwealth; and the last column shows Katherine Clark's margin against Republican Craig Spadafora in the 2010 election to replace Tisei.

Lynnfield is one of the most Republican towns in Massachusetts. Reading, Stoneham, and Wakefield are split between more liberal and more conservative precincts. Melrose could be considered about average for the Commonwealth, and Malden is fairly liberal. The district as a whole had an Obama margin of 18.7, which was 7.7 points below the state average of 26.4 points.

In 2010 Clark outperformed Obama in her hometown of Melrose and came out 4.6 points above her Republican opponent Craig Spadafora district-wide. Clark's 4.6 point margin was 12.3 points above the normalized district score of -7.7, a margin we can project into the new district to judge how a Democrat might fair post-redistricting.

Middlesex and Essex precinct-by-precinct Clark margins
(click map for interactive version)

The Lynnfield portion of the map is the darkest red—Clark lost Precinct 4 of Lynnfield by a whopping 35 points. Clark lost Reading and Stoneham by small margins, and won the remaining municipalities.

Redistricting and the new 5th Middlesex Senate District

Each of the United States are required to redraw election district boundaries every 10 years based on population changes shown by the U.S. Census. A holistic view of Massachusetts redistricting shows few signs of gerrymandering, but population shifts did require changes making some districts more conservative, and some more liberal. The Middlesex and Essex State Senate District lost all of Lynnfield, its most conservative piece, and picked up more liberal precincts in the remainder of Melrose and half of Winchester, and was renamed the 5th Middlesex.

While Katherine Clark did not have an opponent in the first post-redistricting election in November 2012, we can look at President Obama's performance in the 5th Middlesex compared to the state average to evaluate the changes. The Projected Dem Margin column in the chart adds Clark's 2010 12.3 point differential to Obama's normalized performance, giving an idea how a strong Democratic candidate might fare.

In 2012 President Obama's Massachusetts statewide margin was down to 23.6 points (from 2008's 26.4 points). The 5th Middlesex redistricting added more precincts close to the statewide average—the rest of Melrose and half of Winchester—bringing the district's Obama margin within 3.5 points of the quite-liberal state average. A simple projection of Clark's 12.3 points above the state average gives a ballpark of a Democratic win of 8.8 points for the district, while projecting Democratic losses in Reading, Stoneham, and Wakefield.

It is unlikely that this projection will be realized exactly, resulting in an 8.8 point win by Rep. Lewis. However, the 4.2 point difference between the Middlesex and Essex District's score of -7.7 and the 5th Middlesex District's score of -3.5 can be seen as an approximate measure of the district's move to the left in redistricting.

Had Clark not run for Congress, she would have greatly benefited from the additional precincts in her hometown of Melrose (in addition to the loss of conservative Lynnfield). GOP candidate Monica Medeiros is from Melrose and will likely benefit from having all of her home town in the district. Jason Lewis is likely to get a hometown boost and perform on par with Obama in Winchester.

The following map leaves out these municipality-specific differences and simply gives the precinct-by-precinct projection based on Obama's performance, giving a feel for a generic Democrat's performance in each precinct. While the map contains a lot of red, it will be hard for the GOP to overcome the deep blue of Malden, the left-leaning precincts in Melrose, Winchester, Stoneham, and Reading. One recipe for a Medeiros win would be to win her hometown of Melrose and hold down Lewis's margins in Malden.

5th Middlesex precinct-by-precinct projected margins
(click map for interactive version)

The 5th Middlesex State Senate seat will be won by the candidate who does the best of job of connecting with voters in the next few weeks and getting them to the polls on April 1st. The electoral makeup of the district makes this a readily manageable task for the well-organized Rep. Lewis.

A rundown of the five March 4 special state primaries

One primary is inconsequential, the remainder will likely determine winner of general election
by BRENT BENSON

On Tuesday, March 4 there will be five special state primaries: one to fill Katherine Clark's State Senate seat, and four to fill the State Representative seats of Donald Humason (recently elected to the State Senate), Kathi-Anne Reinstein (left for the private sector), Marty Walsh (recently elected Mayor of Boston), and Eugene O'Flaherty (serving as Boston Corporate Council under Walsh). The primaries are uncontested for the Humason seat, but the winner of the remaining Democratic primaries will likely win the general election.

5th Middlesex Senate

The 5th Middlesex State Senate seat was held by Katherine Clark, who was elected to the U.S. Congress to replace Ed Markey, who was elected to the U.S. Senate to replace Secretary of State John Kerry. The ripple effects continue as State Representatives Christopher Fallon and Jason Lewis face HomeStart non-profit director Anthony Guardia in the Democratic primary. Alderman Monica Medeiros is unopposed in the Republican primary.

Candidate	Pty	City/Town	Position	Fundraising
Christopher Fallon	D	Malden	State Representative	$61,350
Anthony Guardia	D	Wakefield	Director of HomeStart non-profit	$21,795
Jason Lewis	D	Winchester	State Representative	$146,908
Monica Medeiros	R	Melrose	Alderman	$6,618

President Obama carried the 5th Middlesex District by 20 points and Elizabeth Warren defeated Scott Brown by 2 points in the district, making this seat unlikely to fall into Republican hands.

Rep. Lewis has a strong fundraising advantage over Rep. Fallon. In addition, the Democratic primary voting electorate will likely be more favorable to Lewis's ideology. Fallon had the dubious distinction of supporting Scott Brown over Elizabeth Warren for Senate and Fallon has a much more conservative voting record than Lewis. The one disadvantage for Lewis is the small overlap of his hometown of Winchester—only about 7% of the district votes—while Fallon's home town of Malden has about 27% of the 5th Middlesex votes.

Looking at all of the factors, Jason Lewis is the favorite. The winner of the Democratic primary should not have a problem winning the general election unless there is some strange and unexpected revelation.

13 Suffolk State Representative

There are five Democratic candidates and no Republican candidates for the 13 Suffolk primary to replace newly-elected Boston Mayor Marty Walsh, virtually guaranteeing that Tuesday's winner will be the general election winner.

Candidate	Pty	City/Town	Position	Fundraising
Liam Curran	D	Boston	City of Boston Attorney	$22,387
Gene Gorman	D	Boston	Emerson College Professor	$9,795
Daniel Hunt	D	Boston	MSBA Project Analyst	$104,568
Paul McCann Jr.	D	Boston	Boston Public Health Attorney	$15,070
John O'Toole	D	Boston	City Council Candidate, Plumber	$25,245

While this is a particularly strong field of candidates, Daniel Hunt has a big enough fundraising advantage to be considered the favorite. The Dorchester district, like much of Boston, is incredibly Democratic with Obama winning the district by 53 points, and Elizabeth Warren by 40 points.

4th Hampden State Representative

There is only one candidate in each of the Democratic and Republican primaries to fill Republican Donald Humason's seat who was elected to the State Senate in November. Democratic candidate John Velis is a recently retired Army Captain and attorney who will face off against Republican City Councilor Dan Allie. President Obama won the 4th Hampden District by 6 points, but Elizabeth Warren lost it by 13 points, making this seat winnable by either party.

Candidate	Pty	City/Town	Position	Raised
John Velis	D	Westfield	Army Captain (Ret.), Attorney	$9,740
Dan Allie	R	Westfield	City Councilor	$10,960

16th Suffolk State Representative

Revere State Representative and Assistant Majority Leader Kathi-Anne Reinstein resigned in January to take a position with the Boston Beer Company. There are three Democratic candidates running in the primary and Reinstein Chief-of-Staff Roselee Vincent has shown fundraising power and has received endorsements from groups like the Massachusetts Women's Political Caucus—she can be considered the favorite.

Candidate	Pty	City/Town	Position	Fundraising
Joshua Monahan	D	Chelsea	Attorney	$6,718
Linda Rosa	D	Revere	Former Bob Travaglini staffer	$5,100
Roselee Vincent	D	Revere	Reinstein staffer	$42,598
Todd Taylor	R	Chelsea	Staffing company owner	$7,115

The 16th Suffolk District went for Obama by 28 points and Warren by 14 points, so the Democratic winner should prevail in the general election on April 1.

2nd Suffolk State Representative

Chelsea State Representative and Judiciary Committee Chair Eugene O'Flaherty left the State House for City Hall, becoming the chief lawyer for Boston Mayor Marty Walsh. There are three strong candidates to replace O'Flaherty.

Candidate	Pty	City/Town	Position	Raised
Roy Avellaneda	D	Chelsea	Former City Councilor	$28,460
Christopher Remmes	D	Boston	Realtor	$49,897
Daniel Ryan	D	Boston	Capuano District Aide	$47,360

It is not at all clear who will win this primary and become the next State Representative for the 2nd Suffolk District. While Avellaneda had the smallest fundraising haul, he received a ringing endorsement from the Boston Globe.

Who is way ahead in MA Governor polling? The candidate whose name voters know

Party schisms are not the reason for difference between Coakley's lead in polls and Grossman's lead in caucus delegates
by BRENT BENSON

The race to replace Massachusetts Governor Deval Patrick is heating up as the Massachusetts Democratic Party conducts local caucuses to determine delegates to state convention. Candidates must win the first votes of 15 percent of the delegates to get on the primary ballot and the candidates have been working to line up pledged delegates at the caucuses.

Polls of the general electorate have shown Attorney General Martha Coakley with a commanding lead, while Treasurer Steve Grossman is the leader in pledged delegates. A National Journal article suggests that "discrepancy between Coakley's lead in the polls and Grossman's strength among the party faithful represents a schism between activists still wary of Coakley after her 2010 loss and a general electorate that's more forgiving."

In fact, Coakley's lead in the polls is due to the fact that she is the only candidate known to the majority of the voting electorate. The latest Suffolk/Boston Herald poll shows that a full 82% of the voters have an opinion of Coakley, while only 35% have an opinion of Grossman.

While there are plenty of factions and strong opinions about the best candidate within the party activists and elected officials who attend the Democratic convention, it is clear from the polling the larger set of people who vote in Democratic primaries do not know the candidates well enough to provide much insight into the outcome of the primary on September 9, 2014.

The democratization of data journalism

A report on the MassINC Big Data and the Future of Journalism panel discussion
by BRENT BENSON

The proliferation of interesting and easy-to-access internet-based data combined with fast computers and free or cheap analysis and visualization tools are democratizing and popularizing the field of data journalism. But what are the rules, parameters, and guidelines for this work which, like its name, bridges the worlds of data and statistics on one hand, and traditional journalism on the other?

This was the subject of a fascinating panel discussion at Boston's District Hall sponsored by MassINC, Google, WBUR, and the Knight Foundation. The introduction was given by the President of the MassINC Polling Group, Steve Koczela, who set the stage by giving an overview of the data journalism field and the work being done at the MassINC Polling Group. The debate was thoughtfully moderated by Northeastern professor and media commentator Dan Kennedy, who kept the discussion moving and asked important questions that allowed each of the panelists to provide their particular insight into this hard-to-nail-down subject area.

Laura Amico of Homicide Watch and WBUR's Learning Lab recommended data journalism driven by classic journalistic inquiry, combined with thoughtful collection of journalistic data, but augmented by categorization and analysis of the data with simple tools like spreadsheets in order to shed quantitative light on the driving questions. The idea is to turn regular reporting into structured data that can enable more powerful reporting.

Paul McMorrow of CommonWealth Magazine also comes to data journalism from a traditional journalistic angle, but Paul has augmented his journalistic chops by embracing tools like Google Fusion Tables to synthesize data from publicly available sources to produce maps and tell stories. His message was that traditional journalists and their organizations should dive head first into learning these new tools. He emphasized that data journalism does not have to be a major investment, given the availability of the tools.

John Bracken is Director of Media Innovation for the Knight Foundation and is primarily concerned with supporting and nurturing innovative journalists that can make use of data journalism tools to improve the world. He is fostering innovation to increase the impact, scope, and efficacy of data journalism.

Charlie Kravetz is the General Manager of WBUR. He is trying to figure out if data journalism could be the savior of the journalism and deserves significant investment, or whether it is just a passing thing. Charlie also wondered whether journalism and writing skills are mutually exclusive with mathematical and statistical skills (answer: No, evidence: Nate Silver and many others). He was unimpressed with a visualization of Manhattan income inequality shown in the opening slide deck.

My take on the panel discussion:

• Not all data journalism is good data journalism—it can use flawed statistical methods or analysis, or be uninteresting or unimportant from a journalistic point of view. The presence of a bad article or uninteresting data visualization does not mean data journalism is a bad thing or unimportant.
• While it is good for journalists to branch out and get comfortable with analytical tools, at some point journalists—at least those without strong mathematics backgrounds—will come up against a wall where an understanding of statistics is important to do analysis or decide whether a particular model or argument is valid. Once this wall is reached it is important to seek additional training, or work with a subject area expert (similar to the function Steve Koczela serves for MassINC and CommonWealth Magazine). 
• While the overlap between good writers/journalists with good statisticians/programmers might be small, they exist and will probably be valuable to news organizations.

I hope these interesting discussions continue in Boston and throughout the journalism community. 

MA Governor polling takeaway: low name recognition makes head-to-head numbers meaningless

Martha Coakley is only candidate known to a majority of voters
by BRENT BENSON

The race to replace Deval Patrick as Massachusetts Governor has begun to take shape with five Democrats, two Republicans, and three independent candidates indicating their a candidacy. The press release and news coverage of the latest poll from Suffolk University and the Boston Herald plays up the head-to-head match-ups between Democratic Attorney General Martha Coakley, Republican former Massachusetts Secretary of Administration and Finance Charlie Baker, and other candidates. However, the polling shows that the vast majority of the Massachusetts voters have never heard of most of the candidates and only Coakley is known well enough that a majority of voters have a favorable or unfavorable opinion.

The Suffolk poll asked respondents to judge the candidates favorably or unfavorably, allowing them to opt out as having never heard of the candidate, or as not yet having an opinion of the candidate. I have divided the respondents into Has Opinion (for favorable or unfavorable) and No Opinion (for never-heard-of or undecided), and sorted the table by Has Opinion.

A full 82% of those surveyed had an opinion of Martha Coakley, the sitting Attorney General of the Commonwealth and former U.S. Senate candidate. Only 49% of respondents had an opinion of Charlie Baker, a fairly surprising number given his Republican nomination and run as a challenger to Governor Deval Patrick in 2010. While Baker may find it concerning to be known by so few, he might hope it works to his advantage by allowing him to rebrand himself in a way that is more appealing to women and Democrats, voting blocks that shunned him in 2010.

Treasurer Steve Grossman was known to 35% of the respondents, former Homeland Security official Juliette Kayyem to 11% of respondents, and the remaining candidates were all below 10%.

There is plenty of useful information in these early polls, including information about the makeup of electorate, issues that are deemed important by the voters, and opinions about topics that may play a roll like casino gambling, but it makes sense to hold off on predicting likely election outcomes until the voters have a chance to get to know the candidates.