Massachusetts Gubernatorial Election 2014

[jimrtex]

Massachusetts Cumulative Vote Share

This is similar to the analysis that I had done for Sedgwick County, Kansas. It is a work in progress. I am now on Essex County.

Some interesting items:

Barnstable

The largest precincts are in Chatham, Eastham, and Orleans, three small towns that consist of a single precinct. There is not a large amount of partisan variation, other than Provincetown and Truro are extremely Democratic.

Interesting is the low turnout in Barnstable 9, which is the central part of Hyannis. I did find something that said Hyannis had an unusually high percentage of rentals compared to the county (about 20% higher). So I'm wondering whether there were seasonal workers who registered in 2012 to vote in the presidential election, and were long gone by 2014, and there was not a comparable registration drive in 2014. Alternatively, it is just a low income area made up of people thought it would be fun to live on Cape Cod, but had to work 80 hours a week in the Summer, and discovered that there was no work in Winter and it was cold.

Berkshire

Berkshire has a lot of very small towns and so the smaller precincts tend to be more Republican. The two largest precincts are the single-precinct towns of Lenox and Lee, with almost identical population but quite different partisanship. Williamstown 2 is quite partisan Democratic, and it the location of Williams College. The precinct has about 3000 persons, and enrollment is just over 2000. Turnout tended to be somewhat lower in Democrat-leaning precincts, but it appears that this is primarily due to greater variation as precincts are more Democratic. Turnout in the old mill towns of Pittsfield, Adams, and North Adams was particularly low, while Democratic registration was higher.

Correlation between Coakley share and the Democratic registration share was not particularly high (0.607). This appears to be related to high rates on not-party enrollment. New Ashford, for example, had 20% Democratic enrollment, and was 71% for Coakley. But Republican enrollment was about 5%, and 75% unenrolled. Democrats outnumbered Republicans 4 to 1, but the unenrolled only broke about  2 to 1 for Coakley.

Bristol

This is the first county where the largest precincts were Republican, and the smallest precincts were in cities and more Democratic. The county is dominated by New Bedford and Fall River, though less so than in the past. In 1940, the two cities had more than 60% of the county population, but that has declined to just over 30%, as the population has boomed along I-495 in the northern part of the county. Both cities hit their peak population in 1920, and have been declining ever since. But it appears that their government has not declined. Cities in Massachusetts almost uniformly have the same number or precincts in each ward (Boston is the exception). Fall River has 9 wards of 3 precincts each, while New Bedford has 6 wards of 6 precincts each. As the population has declined and (if) the number of precincts remained constant, the population per precinct will decline. In any even Fall River and New Bedford have the smallest precincts and are strongly Democratic leaning. Meanwhile towns closer to I-495 have generally quadrupled in population since WWII. Division into precincts was restrained, and towns have larger precincts and are more Republican. Taunton appears to be a medium position. A mill town, but close enough to commute towards Boston, it has experienced some growth since WWII. It has medium-sized precincts and an intermediate partisanship. Somerset and Dartmouth, which are adjacent to Fall River and New Bedford, have had some growth probably from their neighboring cities. They are less Democratic than their neighbor cities, but more Democratic than the county as a whole, and precinct sizes are more typical.

Not only do the Democratic-leaning precincts have fewer registered voters, they had substantially lower turnout in 2014. Given the effect of the smallest precincts and the lowest turnout, it is not remarkable that the Baker cumulative vote share increased as larger, higher turnout, more Republican precincts were included. Correlation between Democratic registration and Coakley support was 0.890, indicating that enrollment was more conventionally partisan in Brisol County.

Dartmouth 3 had 6.1% turnout. It is identical to the campus of UMass Dartmouth. The campus is new (mid-1960s) and appears to have been built on an open oiecee of land, so there is no off-campus housing immediately adjacent. I also suspect that a university that touts itself as an affordable public college, has a large share of commuter students. The registration may reflect large numbers of freshmen in 2012, who are no longer living in dorms, and perhaps are no longer in school. The precinct is the 37th smallest in registration, but 3rd smallest in terms of votes cast. Turnout was also low (mid-teens) in precincts W3-A and W2-A in New Bedford, which are near the port. W3-A is dominated by a freeway interchange plunked down in the center.

The highest turnout was in Mansfield 1, at 56.7%. Mansfield is the wealthiest town in the county, and Baker defeated Coakely 2 to 1 in the precinct.

Dukes

Dukes has too few precincts to do much analysis on. For the county, registration is 107% of the voting age population. It is particularly high in the smaller towns. It is over 200% for Gosnold, which is comprised of the Elizabeth Islands off the southwestern tip of Cape Cod (the rest of the county is on Martha's Vineyard).

[Brittain33]

Hyannis has a large immigrant population. I imagine the drop-off in participation there is the familiar story of people only tuning in to the Presidential races and dropping out for off-years and especially in special elections (look at turnout in the "Gateway Cities" in Massachusetts in Brown-Coakley in 2010) rather than a population change.

[Figueira]

Berkshire County is solidly Democratic, at least in terms of how they actually vote.

Anyway, I look forward to your analysis of Franklin and Hampshire Counties.

[Okay, maybe Mike Johnson is a competent parliamentarian.]

Anyway, I look forward to your analysis of Franklin and Hampshire Counties.

Do you live in the Pioneer Valley by any chance?

[Figueira]

Anyway, I look forward to your analysis of Franklin and Hampshire Counties.

Do you live in the Pioneer Valley by any chance?

I PMed you the answer to this question.

[jimrtex]

Hyannis has a large immigrant population. I imagine the drop-off in participation there is the familiar story of people only tuning in to the Presidential races and dropping out for off-years and especially in special elections (look at turnout in the "Gateway Cities" in Massachusetts in Brown-Coakley in 2010) rather than a population change.

It appears that "Gateway City" has an economic definition₁, rather than a specific immigration-related definition, though they are probably related. Cities that grew as industrial centers in the late 19th and early 20th centuries attracted immigrants since they provided low skill-jobs. Even as low-skill industrialization moved South, there may have been continued chain immigration. Those who went to college would tend to move where they could get jobs, so they would now live in Nashua or Austin, San Jose or Bellevue.

In any event, Barnstable 9 (Hyannis) is an extreme outlier, at the bottom of this chart.



At 34% turnout, it is 4.01 standards deviations below the county average. It also appears that turnout was partisan differentiated. Other precincts with around 35% Democratic registration went for Coakely at about 60%. While Coakley carried Barnstable 9, she did not receive a majority.

Coakley received a majority in only 5 precincts: the single-precinct towns of Provincetown, Truro, and Wellfleet on the northern tip of Cape Cod, and two precincts in southwest Falmouth town, in Falmouth proper, including Woods Hole and the ferry landings to the islands.

Baker was over 57% in 22 precincts, mostly in the western base of the cape: 7 of 7 precincts in Bourne, 4 of 6 in Sandwich, 8 of 13 in Barnstable, 2 of 5 in Mashpee; and Chatham in the southeast corner of the cape.

The 8 Republican-dominant precincts in Barnstable town are in the west and north, with the five competitive precincts in the southeast in and around Hyannis. Even compared to other precincts in the town, Barnstable 9 stands out. Its registration is about 60% of the largest precinct (fewer persons register or perhaps are not eligible), and with the low turnout, cast only 30% as many votes as the largest precincts.

When we accumulate the Baker and Coakley vote in order of votes cast per precinct, we see very littlle variation in the cumulative vote.



The main reason is that most precincts in the county are of roughly the same size, with 2000 to 3000 registered voters. The step up in Coakley support past 60% is Provincetown, which at 82% for Coakley is 20% more than any other precinct, and almost 40% more for Coakley than she received countywide. The step at 10% corresponds to Truro, which at 62% for Coakley is her second strongest precinct in the county. The smaller precincts are slightly more Republican than the county. They are not markedly smaller, and it just happens that the precincts in Bourne and Mashpee are at the low end of the range of 2000 to 3000 registered voters, and at the more Republican part of the county.

This shows the consistency of precinct size and Democratic registration share.



If you were to pick a precinct at random, I could guess that it had between 2000 and 3000 registered voters, and between 20% and 30% Democratic registration, and I would be correct 69% of the time, even though I guessed before you told me the precinct, or even before you had read the name of the precinct.

The three largest precincts are Chatham, Eastham, and Orleans, three single-precinct towns on the eastern cape. Provincetown is the extremely Democratic precinct (57.9%).

¹Between 35,000 a
nd 250,000 population, and educational levels (BA or higher) and income less than the state average. Barnstable town qualifies under this definition.

[Brittain33]

Interesting. How does Barnstable 9 compare to immigrant-heavy precincts in Lowell and Lawrence?

[jimrtex]

Interesting. How does Barnstable 9 compare to immigrant-heavy precincts in Lowell and Lawrence?

I'm not sure it is specific to immigrant-heavy areas. I have started on Essex County, and the smallest precincts are in Lawrence and Lynn.

I did some looking into Massachusetts laws with regard to precincts. The law requires that wards (in cities) be reapportioned each census, and that precincts within wards and towns be reapportioned on the basis of equal population. I don't read the statute as requiring precincts in different wards of a city to have an equal population, but every city other than Boston has the same number of precincts in each ward of that city. Boston has a specific exception, and doesn't elect its city council by ward. There is a bill that would removed the exception after 2020.

Precincts must have less than 4,000 persons, but there is an exception for towns with fewer than 6,200 persons. That explains why Chatham, Orleans, and Eastham are the largest precincts in Barnstable County, since their population is in the 4,000 to 6,200 range.

It appears that in Barnstable County, at least, that the number of precincts is close to the minimum or one additional.

For example, if Yarmouth with a population of 23,793 had six precincts, they would average 3966 persons each. It would be difficult to keep all below 4,000. With seven precincts, the average is a comfortable 3,399.

Bourne is the only town with two extra precincts. With 19,754 persons it could have five precincts with an average of 3,951 or six with an average of 3,292. Instead it has seven precincts with an average of 2,822. The town is divided by the Cape Cod Canal, and has three precincts north and four precinct south of the canal. If there were six precincts, one might have to be a cross-canal precinct.

Overall, Massachusetts law will result in broad uniformity of precinct sizes, with most being in the range of 2800 to 3800 persons or so. Pushing the 4,000 limit makes for difficult redistricting. Perhaps a town that was increasing in population might not want to change the number of precincts, and would try to keep precincts at just below the maximum. A town in decline may not want to decrease the number of precincts, since it requires a lot more than just eliminating one precinct, but also requires adjustment of all or most of the other precincts.

Some exceptions to the 2800 to 3800 rule:

(1) Small towns, mostly in the western part of the state with fewer than 2800 persons.
(2) Modest size towns in the 4000 to 6200 exceptional range.
(3) Cities that have declined in population without reducing the number of wards. In addition, the requirement to have equal-sized precincts in equal-sized wards may require smaller precincts. A city with seven wards may have to choose between seven wards of three precincts each (21 total) or seven wards of four precincts each (28 total).

Registration will be somewhat less than the population, since minors can't vote, and some adults may not be eligible to vote because they are not citizens. Others eligible to vote won't be registered. Persons who move frequently may not be registered at their new residence, and they may feel less connected to their community of residence, particularly if they are renting. Younger persons and single persons may also not have as strong an attachment to their community of residence.

College students are typically enumerated by the census at their college, but may be registered to vote at their parent's residence. Bourne 3, location of the Massachusetts Maritime Academy has about 2/3 of the registration of the the other precincts in Bourne. Because precincts are sized based on population, and not number of voters, even a fairly small college with a few thousand students will represent a full precinct.

In Barnstable County, precincts with lower registration typically have lower turnout. Someone who moves frequently may not bother to register to vote, and if they do, they may not vote in all elections. Or they may have moved since they registered. Voters typically are not removed from the rolls when they move, until they register in another location, and that might not happen either, if the move is to a different state, county, or town.

In most towns in Barnstable County, the precincts with higher registration (and presumably a higher  share of the population, since precincts are apportioned on the basis of population), typically had higher turnout. Not only did they have more registered voters, a higher share of them actually voted.

Based on partisanship (Democratic registration share or Coakley vote share) most of the towns appear to be politically homogeneous. The precincts with higher registration and higher turnout are not necessarily more Democratic or Republican - but in Massachusetts in general, and in Cape Cod in particular, wealthier persons may not be more Republican. Residence on Cape Cod may be a lifestyle choice that outweighs other factors (ie someone chooses a place to live, and then finds a job to pay for it; vs. finding a place to work, and choosing a place to live nearby).

Barnstable town is exceptional. It is the largest town, and there is a large partisan variation. 10% of the population in Barnstable is foreign-born, half naturalized and half alien. Yarmouth is also 10% foreign born, while the remainder of the county averages 5% foreign-born, with a slight preponderance of naturalized citizens among the foreign born.

Among the census tracts in Barnstable town, there is large variation in the percentage of foreign born.

8%, 7%, 7%, 7%, 6%, and 4% is just above the average for the county.

24%, 16%, 16%, 12%, and 11% are much higher.

The foreign-born population in the town is 31% Brazilian, and they are present in most of the census tracts, but particularly within Census Tract 153, which is 24% foreign born, and overlaps Barnstable 9. Brazilian-born persons represent 11% of the total population in that census tract. There are all kinds of clusters of fairly exotic groups such as Bulgarians, Moldovans, Ecuadorians, Laotians, Kenyans, Costa Ricans, Ukrainians, and absence of other groups, such as Mexicans, Salvadorans, etc. There are large numbers of Dominicans (from the Dominican Republic) and Jamaicans, who of course are not so exotic.

Barnstable County reached a peak population in 1860, before having a small decline through WWII. It has increased 6-fold since then. The Brazilians originally were employed as landscapers, painters, builders, etc., but have now begun opening restaurants. So in this case, "Gateway City" does not refer to immigration back into the 19th century.

Incidentally, Truro is 19% Portuguese ancestry (I don't think the census bureau distinguishes Azorean ancestry; among place of birth, it counts Azores as a subcategory of Portugal).

[jimrtex]

In Massachusetts, a Gateway City has between 35,000 and 250,000 persons, and income and share of college graduates below the statewide average. 250,000 is well above the population of Worcester and Springfield, and well below the population of Boston.

Most cities that have above 35,000 are Gateway Cities. The exceptions are some suburban cities around Boston.

In Essex County, Beverly is not; while Haverhill, Lawrence, Lynn, Peabody, Salem, and Metheun are Gateway Cities.

In Middlesex County, Cambridge, Marlborough, Medford, Newton, Somerville, and Woburn are not. Everett, Lowell, and Malden are Gateway Cities. It is not clear whether large towns (over 35,000) are considered eligible or not. Arlington, Billerica, and Framingham are towns (many cities in Massachusetts have an official name of Town of _____, but are classified as cities).

In Norfolk County, Braintree and Weymouth are not Gateway Cities, but Quincy is. Brookline is a large town.

In Suffolk County, Chelsea and Revere are Gateway Cities. Boston is not by virtue of its size.

Elsewhere in the state, all cities that have more than 35,000 are Gateway Cities:

Barnstable: Barnstable.
Plymouth: Brockton. Plymouth is a town.
Bristol: Attleboro, Fall River, New Bedford, and Taunton are Gateway Cities.
Worcester: Fitchburg, Leominster, and Worcester are Gateway Cities. Shrewsbury is a town.
Hampden: Chicopee, Holyoke, Springfield, and Westfield are Gateway Cities.
Hampshire: Amherst is a town.
Berkshire: Pittsfield is a Gateway City.

[jimrtex]

Berkshire County CVS curve is fairly flat.



Berkshire County has 30 towns and two cities, Pittsfield and North Adams. Of these 26 are comprised of a a single precinct. Many of these are quite small, but others are relative large. Lee and Lenox are by far the largest precincts, because their population is in the exceptional range of 4000 to 6200 persons. These small precincts are not necessarily more Republican, particularly around Great Barrington in the southern part of the county. The least Democratic precincts are generally along the eastern side of the county, closer to the Connecticut River valley. It is also possible that these are the areas that are most remote from Great Barrington and the industrial areas of Pittsfield, North Adams, and Adams. There are some less Democratic areas along the Connecticut line (ie more remoter from Great Barrington, and in a belt between Pittsfield and Great Barrington.

Berkshire County hit its peak population in 1970. Since WWII, many of the smallest towns have had significant growth, but this was from a very small base, perhaps from 200 to 400. Meanwhile the industrial areas have been hammered, with North Adams and Adams losing about 40% of their population, and Pittsfield about 10%. Because of the decline in population, the precincts in these cities and town are smaller than necessary, perhaps because of an unwillingness to reduce the number of precincts. Great Barrington has four precincts rather than the two that are required by state law.

One precinct in Williamstown is dominated by Williams College, with a consequent smaller registration, and in 2014, lower turnout. College students, particularly from out-of-state, may not be interested in the gubernatorial politics of a temporary home. The precinct in Great Barrington that contains Bard College at Simon's Rock also has a smaller registration. This is an early entry college, for students who would normally be juniors or seniors in high school, and thus many are too young to vote.

This chart shows the relationship between precinct size, total town registration, and Democratic Registration share.



The version in Google Sheets Mass Governor 2014 (Berkshire Precinct Size v Dem%) is interactive.

This shows that the smallest towns may be slightly less Democratic than the county as a whole, but there is a wide partisan range. In Berkshire County, basing partisanship only on Democratic registration may be misleading. In many areas there are huge numbers of enrolled voters. New Ashford is 19.9% Democratic, 4.8% Republican, and a whopping 72.6% unenrolled, and was 73.1% for Coakley. In small towns, voters may be less willing to reveal or display their partisanship to neighbors. In Pittsfield, Democratic registration is as much as twice as high, but the Coakley performance was no better.

Correlation between the Democratic registration share and Coakley share was 0.619, while correlation between the Democratic/Republican registration and the Coakley/Baker ratio was 0.838, suggesting that the unenrolled to some extent vote like their more overtly partisan neighbors.

Pittsfield precincts are larger than the median for the county. This is not so much that they are large,  but that many of the single-precinct towns are small, and that cities and towns such as North Adams, Adams, Great Barrington, and Williamstown have been subdivided into smaller precincts than required by statute.

When calculating the cumulative vote share 41 precincts are needed to accumulate 50% of the total vote, while only 18 are needed for the final 50%. Coupled with lower turnout, this results in Pittsfield precincts being well distributed over the cumulative vote share curve, with half (7 of 14) being added in during the first 50% of the accumulated vote, and the other half after 50%.

This chart shows the Democratic registration as a function of the number of registered voters.



There is a weak relationship between number of registered voters and Democratic registration, but even if it were significant, it would only mean that a precinct with 500 voters would be expected to have 31.6% Democratic registration, while one with 1500 voters (1000 more) would be expected to have a Democratic registration of 35.4%.

Many of the very small precincts have low Democratic registration. This does not necessarily indicate a high Republican registration, since some towns have very high unenrolled registration. The trend line is in a sense trying to reach closer to these smaller precincts.

If a weighted least square fit is calculated, using the precinct registration as the sample weight, the line is much flatter:

(unweighted): Dem% = 29.74% + 3.81% * (number_registered / 1000)

(weighted): Dem% = 34.47% + 1.29% * (number_registered / 1000)

The slope is expressed as a percentage change in Democratic registration per 1000 registered voters because it gives a better sense of the change as registration changes. An increase of one in registration is insignificant.

The larger intercept changes the projected Democratic registration share for our hypothetical precinct sizes:

(unweighted):   500: 31.6% Democratic; 1500: 35.1% Democratic.
(weighted):  500: 35.1% Democratic; 1500: 36.4% Democratic.

While reducing the influence of the smaller towns, the influence of the two largest precincts, in Lenox and Lee is increased. But it is impossible to fit two the two towns which are neighbors alphabetically (Lee v Lenox, geographically (4 miles apart), and registration (3936 voters vs 3906 voters), but 14% different in the share of Democratic registrants (43% for Lenox, 29% for Lee). But Lenox is not only more Democratic, it has 3% higher Republican registration (11% for Lenox, 8% for Lee), while Lee has 17% higher unenrolled registration (62% for Lee, 45% for Lenox). Despite these apparent differences in partisanship:

Lenox 43% D, 11% R, 45% U
Lee: 29% D, 8% R, 62% U

Yet, Lenox had only 4% more support for Coakley (66% vs 62%).

Compare to Barnstable County, Berkshire County is much more diverse. In Bartnstable, there was a neat box between 2000 and 3000 registrants, and 20% and 30% Democratic registration that contained 73% of all registration. In Berkshire County, few precincts have over 2000 registrants, and those that do are mostly in the 2000 to 2500 range, and the partisanship is much more diverse.

Another notable precincts are Williamstown 2, with 54% Democratic registration (just over 1000 registrants). Williamstown 2 corresponds to Williams College. Williamstown is named for Col. Ephraim Williams, who died in the French and Indian War, who left a sizeable bequest to the town on the condition that it be renamed from West Hoosac to Williamstown, and that a free school be established. The free school later became Williams College. Because Massachusetts apportions precincts based on population, even a small school (2000 enrollment) may constitute a precinct.

Registration is lower for the campus precinct compared to the other two town precincts, and among those that were registered, turnout was lower. Students may be eligible to register at their parent's home, and may have less interest in a gubernatorial election than a presidential election. This may be particularly true among out-of-state students. Some out-of-state students may strategically choose their voter registration residence. A student from Virginia or Ohio might decide they were more influential in their home state than in Massachusetts. Some students from 2012, will have graduated, but their registration in Williamstown not deleted. These non-voters will reduce turnout.

Williamstown 2: 1068 registered, 431 votes, 54% Dem, 40% Turnout, 83% Coakley, +299 margin
Williamstown 1: 1721 registered, 1068 votes, 44% Dem, 62% Turnout, 71% Coakley, +483 margin
Williamstown 3: 1813 registered, 1025 votes, 42% Dem, 56% Turnout, 75% Coakley, +557 margin

The other two precincts produce almost 2-and-1/2 times as many votes, and a much larger margin for Coakley, despite not being quite so Democratic.

In the lower left corner of the chart is New Ashford with 20% Democratic registration and 200 registered voters. One might expect the town to favor Baker. But it only has 5% Republican registration, and has 75% unenrolled and gave Coakley 73% support.

[cinyc]

Did we see significantly lower turnout in Berkshire County than the rest of the state, either as a percentage of registered voters or total residents?  I seem to recall that was the pattern during the Scott Brown Senate Special Election - perhaps in part because Berkshire County is in the Albany TV market and doesn't get as many political TV ads or as much Massachusetts political news as the rest of the state.  That might not translate to a regularly scheduled election when everyone already knows the date, though.

[jimrtex]

Did we see significantly lower turnout in Berkshire County than the rest of the state, either as a percentage of registered voters or total residents?  I seem to recall that was the pattern during the Scott Brown Senate Special Election - perhaps in part because Berkshire County is in the Albany TV market and doesn't get as many political TV ads or as much Massachusetts political news as the rest of the state.  That might not translate to a regularly scheduled election when everyone already knows the date, though.

I have a couple more charts that are pretty interesting.

There is a big difference in turnout between the mill towns (Pittsfield, North Adams, and Adams) and the rest of the county.

21 of 24 precincts in those cities and town were under 45% turnout. Only the one Hyannis precinct in Barnstable was that low. But it is quite isolated, and I was surprised that it even existed. The three mill towns in Berkshire are 50% of the population (down from 70% in 1940). Berkshire hit its peak population in 1970, while Barnstable has 6-tupled in population since 1940. I guess my image of the two is of areas dominated by 2nd homes and B&B, but that may only be true of the southern part of Berkshire.

But I know (so far) that New Bedford and Fall River also have low turnout, so any perception of low turnout in Berkshire might be based on not recognizing Pittsfield, North Adams, and Adams as having half the population. They don't appear to be Birkenstock Belt

What you said about New York influence might be true. The Berkshires are accessible from New York and Connecticut, and some people might not be tuned into Massachusetts politics. Cape Cod is probably very Boston-centric.

There were also three precincts in Berkshire not in the mill cities, Lanesborough, Cheshire, and one of two in Dalton. These are larger precincts, and there may be less of a community tie that you would find in a town of a few 100 persons.

College precincts may also have low registration and low turnout. Williamstown 2 which includes Williams College was right at 40%. In Bristol, UMass-Dartmouth was really low. It will be interesting to see if this holds up at least where the schools is somewhat isolated. Tufts maybe?

I don't know whether schools in places like Boston and Cambridge can be isolated residentially. The precinct in Bourne that contains Massachusetts Maritime Academy is just over 45%. Its student body might be a bit different than liberal arts colleges. Perhaps some older students who had worked on boats and realized that it might not be possible to advance through experience alone. It may have a more Massachusetts-oriented student body.

[Brittain33]

I don't know whether schools in places like Boston and Cambridge can be isolated residentially.

You're right, you can't. Probably the only place it is possible is in Allston-Brighton and the Fenway area where students living off-campus have priced out and scared off everyone else. It certainly wouldn't work with Harvard in Cambridge, and I don't think so in MIT (but it might be possible.)

[jimrtex]

This charts show the relationship between turnout and Democratic registration.



It seems to indicate that as Democratic registration increases, turnout decreases. But it is an extremely weak. In addition, as the Democratic registration reaches around 45%, it is higher for some precincts and lower for other precincts.

If we segregate the precincts by town type, a different picture emerges.



The mill towns are Pittsfield, North Adams, and Adams, which comprise about half the county and 24 precincts. The other 29 towns have a total of 35 precincts. Only Great Barrington (4), Williamstown (3), and Dalton (2) have more than one precinct.

In general the other towns had higher turnout than the mill towns, and it appears that the turnout increases as the Democratic registration increases. The Williams College precinct of 54% Democratic registration but only 40% turnout is an outlier. The relationship is weak. Perhaps the GOTV effort focuses on registered Democrats. If you can get them to vote, it is quite likely that they would vote for Coakley, and the fact that they indicated a party, may indicate more of a willingness to also vote. Some unenrolled voters may be indifferent about politics - and are only registered because someone at the DMV handed them a form. If it is were preprinted with their name and address, all they would have to do is check the citizenship and 18 YO boxes and sign it. It will be harder to get them to vote, and you won't know who they will vote for.

If you can convince 80% of Democrats to vote, and 90% will vote for Coakley you will yield a margin of 0.64 votes per contact. If you cnan convince 60% of unenrolled to vote, and 60% will vote for Coakley, you will yield a margin of 0.12 votes per contact. Calling Democratic voters is much more efficient.

Turnout in the mill towns appears to have an economic relationship - and this is only weakly related to partisanship. Consider Pittsfield. The number of registered voters among the 14 precincts ranges from 1494 to 2499, even though they nominally have the same population. The precincts toward the inner part of the town have lower registration. The seven wards in Pittsfield are radially arranged, and are divided into two precincts A and B. In six of seven wards, Precinct A, which is typically closer to the center of the city, has fewer registered voters.

Lower registration does not make them significantly more Democratic. The range of Democratic registration is from 40.6% to 46.4%, with a standard deviation of 1.8%. Correlation between registration and Democratic registration share is -0.516. Lower registration does correlate with higher Democratic registration, but not much higher and there are exceptions. The most Democratic precinct is the fourth largest, and the fourth smallest is the fourth least Democratic precinct. A least squares fit of Democratic registration share against the number of registered voters indicates a 3.3% decrease in Democratic registration for every 1000 voters, which happens to be the difference between smallest and largest precincts (1005 more voters, and 3.4% smaller registration share).

The range of Coakley support is a bit wider, ranging from 60.0% to 72.4%, with a standard deviation.of 2.9%. Correlation between registration and Coakley share is -0.486,  with a trend of 5.2% less Coakley support per 1000 voters.

The big difference is in turnout, which ranges from 21.8% to 52.7%, and a standard deviation of 8.1%. Surprisingly this is strongly correlated with the number of registered voters (0.918). One would expect the number of actual voters to correlate with the number of registered voters, but not necessarily the turnout rate.

If we assume that all precincts have the same population of 3196, then the registration rate ranges from 48% to 78%. Some of the difference would be due those ineligible to register (children and aliens), but some may be among those who are eligible but don't register. But then of those 48% who did register, only 22% turned out. In the highest performing precinct, of the 78% who registered, 53% turned out.

The combination of the two factors results in Precinct 4-B having casting four times as many votes as 2-A (1316 to 326). Despite 2-A going 72% for Coakley, and 4-B only being 60%, 4-B ran up a margin for Coakley 1.77 times greater (+319 to +180).

Pittsfield is only 5% foreign born (it along with the Attleboro, among the Gateway Cities, have the smallest share of foreign or Puerto Rican born.

But the census tract that corresponds to 2-A is 15% foreign born, with about 1/5 naturalized. As you would expect, the largest groups are Brazilian, Ecuadoran, Salvadoran, Honduran, Ghanaian, Chinese, and other West Asians (given the countries that are itemized, this includes Palestinians, Azerbaijanis, Georgians, Omanis, and UAEns). From the sources, these are likely fairly recent arrivals. I don't think the low electoral performance of 2-A is necessarily due to the alien population, but the conditions that attract immigrants (low cost housing - probably rental and low-skill jobs).

The census tract that corresponds to 4-B has the second largest foreign-born proportion, 7%, but it is much more likely to be naturalized, and is from countries like Italy and Poland. Persons who are in their 70s and 80s who have been citizens for decades may even vote at greater rates than natives.

This connection of turnout to economic conditions is not that dissimilar to that in Cook County, where wealthy Democratic-leaning areas like Evanston and Oak Park have turnout every bit as high as wealthy Republican-leaning areas like New Trier and Barrington, but both have much higher turnout than much poorer Cicero.

The other census tract with a high foreign population is Great Barrington, also at 15% and with large numbers of Brazilians and Ecuadorans. They may be able to live in the largest town in the southern part of the county, and work in the smaller surrounding towns in jobs like builders, gardeners, painters, etc.

[jimrtex]

Bristol County is the first county (alphabetically) to exhibit the property of increasing Republican vote share as larger precincts are added in. To the unthinking, this proves vote fraud, since it is a well-known fact that larger precincts are in cities and therefore more Democratic leaning.



But let's first look at the results when added in order of number of registered voters.



They are similar patterns, but if you look closely, the curve for registered voters converges a bit faster. This indicates that Republican-leaning precincts had higher turnout, which shifted them toward the right when measured by votes cast.

In Bristol County, Democratic-leaning precincts are smaller.



The trend line hardly fits the data, even though it shows a sharp drop of 12.7% in Democratic registration for every 1000 extra voters. The trend line projects a Democratic registration percentage of 43.7% for a precinct with 1500 voters. Most precincts of that size are more Democratic. At 2500 registered voters, the trend line projects 30.9% Democratic registration, but most precincts of that size are less Democratic.

A step function would better represent the data:

Precincts between 1000 and 2000: 52% or so Democratic.
Precincts between 2000 and 3000: 24% or so Democratic.

It is not so much a stair step, but a cliff.  There is an area between 1800 and 2200 where precincts of both levels of Democratic-support exist, but it is not a blending area with intermediate levels of Democratic support.

There are four extremely small precincts in the county. Fall River W6-C1 and W6-B1; and Raynham 1A and 2A. Fall River and Raynham are the only city and town in Bristol County divided by congressional districts. Fall River is divided between MA-4 and MA-9, while Raynham is divided between MA-4 and MA-8.

State law requires each town and city to reapportion wards and precincts to be within 5% of the average. This process must be completed by June of the year following the census (2011, etc.). The city or town must submit the maps to the state for review. If they fail to do so, or the state rejects the map, then the city or town must use what the state tells them to do.

The legislature then uses the precincts when dividing towns and cities when draw legislative maps. Massachusetts House districts have around 40,000 persons. Since most precincts will have less than 4000 persons, a precinct will have 10% or less of the ideal size. If the house district being built is between 5% and 15% too small, then adding a precinct with 10% will bring it within -5% to +5% of the ideal. If precincts are smaller, it makes it even easier. The legislature thus avoids having to find dividing lines, etc. in towns.

Congressional districts are also built using precincts to divide towns. But because the legislature thought that congressional districts had to be exactly equal in population, they divided precincts. The resulting subprecincts are actually defined in the statute that defines the congressional districts. Had the legislature known about Tennant, they would have not had to define these microprecincts.

For example Fall River W6-C1 is a single block with 40 registered voters. The remainder of precinct W6-C has 1703 registered voters.

Since precincts are apportioned by population, we can often recognize college precincts by the number of registered voters. Most college students are over 18, so college precincts will have more persons over 18 (children make up 20 to 30 percent of the population), but many will not be registered at their college residence. Some will be registered at their parent's residence, and others may not be registered, particularly in non-presidential years.

Turnout will also likely be low. Many of those registered will still be on the rolls, even though they graduated or left school years earlier. In general, voters can not be purged unless the registrar has affirmative knowledge of their death, re-registration elsewhere, or having moved. If the registrar sends mail to a voter and it is returned, they can put them on an inactive voter list, and then purge them if they miss the next two general elections. A few years ago, Orange County, California used commercial mailing lists to track down voters. It then sent a notice, asking the voter to confirm that they had moved.

Three identifiable college precincts in Bristol County are:

Norton 5: Wheaton College
Easton 1: Stonehill College
Dartmouth 3: UMass Dartmouth

The UMass Dartmouth (and its campus) are fairly new, dating from the the 1960s. It appears that the campus was built on a piece of undeveloped land, such that there are no immediately adjacent off-campus housing. UMass Dartmouth touts its affordability, and may attract older students from New Bedford and Fall River who realize they are going nowhere fast without a college education. The election precinct is identical to the campus, so that all the potential voters live in residence halls. Besides having a lower number of registered voters than other Dartmouth precincts, it had 6% turnout in 2014.

There were a couple of low registration precincts elsewhere which I could not associate with a college. Attleboro 5A appears to be the most central precinct in the city. Bridgewater State University has a branch in the city, but I don't think it is residential. Fairhaven 3 is directly opposite across the bridge from the harbor area of New Bedford. Two precincts in New Bedford just opposite are the lowest registration precincts in the city.

[jimrtex]

This chart explains why in Bristol County, that the smallest precincts are more Democratic.



The simple answer is that the smallest precincts are in New Bedford and Fall River, the two largest and by far the most Democratic places in the county. Massachusetts law requires wards and precincts within wards to be reapportioned following the census, and that precincts have less than 4000 persons (towns with less than 6200 persons do not need to have precincts, but there are no such towns in Bristol County, all have at least two precincts.

Many towns have the minimum number of precincts. If they had one less precinct, some precincts would have more than 4000 persons. The minimum number of precincts is:

minimum_precincts = ceil( population / 4000),

where ceil(x) is the largest whole number greater than or equal to x

The following towns have the minimum number of precincts for their population: Acushnet, Berkley, Dartmouth, Dighton, Easton, Freetown, Mansfield, Norton, Raynham, Rehoboth, Seekonk, and Somerset. Some are quite efficient. Mansfield with 23,184 residents has six precincts with an average population of 3864. Some are forced to have somewhat smaller precincts. Freetown with 8870 residents, has to have three precincts that average 2957 residents.

Attleboro, North Attleborough, Swansea, and Westport have one precinct extra. Taunton and Fairhaven have two extra, Fall River has four extra, and New Bedford twelve extra precincts.

Attleboro, Taunton, Fall River, and New Bedford are cities, and must have equal-population wards, and equal-population precincts within each ward. They may have to change charters or the size of their city council to change the number of wards.

Attleboro has six wards of two members each. If there were no wards, Attleboro could have 11 equal-population precincts. Instead it has six wards with two precincts each. Taunton has eight wards with two precincts each, rather than a minimum number of 14 precincts.

Fall River has nine wards with 3 precincts each, rather than the minimum number of 23 precincts. Fall River would have to reduce the number of wards to fit this size. New Bedford has six wards of six precincts each, rather than a minimum of twenty-four precincts. New Bedford could reduce the number of precincts per ward to four or five, but this might not match existing polling places or sensibilities about neighborhoods or whether polling places should be within walking distance of residences. In newer suburban areas, use of an automobile to go anywhere away from home may be presumed.

New Bedford and Fall River reached their peak population in 1920. There may be an inertia in reducing the size of the government.

Fairhaven and North Attleborough have a representative town meeting form of government. In (most) Massachusetts towns the town meeting is the legislative body for the town. Any eligible resident may vote in the town meeting. An annual meeting is required, and there may be special town meetings. The town also elects selectmen to serve as the executive body. Typically the board of selectmen has three members, so they act as multi-person mayoral body.

An alternative to an open town meeting in which 1000s of persons could show and debate and vote, is the representative town meeting. Members of the representative town meeting are elected from precincts, and are quite large (as compared to a city council). In Fairhaven, each of the six precincts elects about 72 town meeting members. Elections are by thirds to three year terms, so the ballot will say something like, "Vote for not more than 24". There may be vacancies that are also filled. If Fairhaven had fewer precincts, each would have to elect even more town meeting members (Fairhaven apportions the members such that about 5% of registered voters will be part of the representative town meeting.

North Attleborough has a somewhat smaller body, with each of the 9 precincts electing 15 members (5 each year for a 3-year term). It appears that precincts nearer the center of town have a harder time maintaining their representation.

The only towns in the county where the number of precincts is not the minimum, and where precincts are not tied to the election of the local government are Swansea and Westport. Both have just below 16,000 population, such that if they had four precincts, they would average around 3900 to 3950. In Swansea, a precinct with a 1% deviation from the ideal would exceed the 4,000 population maximum.