Kansas Elections Explained
       |           

Welcome, Guest. Please login or register.
Did you miss your activation email?
March 28, 2024, 07:31:47 PM
News: Election Simulator 2.0 Released. Senate/Gubernatorial maps, proportional electoral votes, and more - Read more

  Talk Elections
  General Politics
  Political Geography & Demographics (Moderators: muon2, 15 Down, 35 To Go)
  Kansas Elections Explained
« previous next »
Pages: [1]
Author Topic: Kansas Elections Explained  (Read 5088 times)
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« on: August 02, 2015, 08:59:21 PM »

There is a lawsuit in Kansas by a statistician who claims that results of Kansas elections show evidence of vote fraud.

WSU statistician sues seeking Kansas voting machine paper tapes

In particular, she showed as the number of votes cast in a precinct increases the results are more Republican.  She then jumps to an erroneous conclusion that because we "know" that big cities are more Democratic, big precincts are more Democratic, and therefore the only explanation is that there has been vote fraud.

Her analysis arranges election precinct in order of votes cast and shows that as the number of votes cast increases, the Republican vote share increases.

How trustworthy are electronic voting systems in the US?

Instead using the numbers of votes case, I used the number of registered voters in a precinct.  And instead of election results, I used the partisan registration.  The following data is for Sedgwick County (Wichita).

The cumulative Republican registration for the ith precinct is sum(j=1,i) GOPi; the cumulative registration for the ith precinct is sum(j=1,i) registrationi.

The cumulative Republican share is the cumulative Republican registration divided by the cumulative registration.  This is shown with the blue curve.



The initial minimum is at 37,959 cumulative registrants, with 38.9% Republican registration.  This includes the 104 precincts with 865 or fewer voters.  There are only 146 precincts (samples) to the right of this is likely do to sampling noise, due to runs of Republican or Democratic precincts.

In this case, there does not appear to be any relationship between the number of registrants and the Republican registration share.

In the red curve, the precincts have been reordered based on the votes cast in the 2014 election, but the x-axis is still based on registered voters.  For example, in the blue curve, precincts UN01 and 0119 are adjacent to each other, with 1384 and 1444 registrants, respectively at roughly the 93,000 cumulative registration (they are the 153rd and 154th smallest precincts).  UN01 is 62.2% Republican, while 0119 is only 17.6% Republican.

Turnout in UN01 was 60.3% for 835 votes cast, versus 43.9% for 615 votes cast in 0119.  In the red curve UN01 was shifted right to the 182th sample (about 140,000 cumulative registration), and 0119 was shifted left to the 136th sample (about 76,000 cumulative registration).

Turnout was strongly correlated with the Republican registration share (0.888).  Using a least squares fit, turnout increases about 2/3% for every 1% increase in Republican registration.  Being Republican greatly increases the likelihood of voting.  Precincts with more votes cast, have more votes cast, because they are more Republican.  As these more Republican precincts are added in, the cumulative Republican registration share increases.

Registration data may exaggerate the number of eligible voters, particularly Democrats. Democrats are younger, less married, less home owners, less economically stable, all factors that cause them to move more frequently. When they move they may not register.  And they are not likely to inform their old county that they are moving.  The registrations include "inactive voters". An inactive voter is not someone who doesn't vote. An inactive voter is someone whose mail was returned by the post office, particularly because of no forwarding address.  The federal government won't let such a person be removed from the voting rolls unless they miss two elections after the confirmation was returned unforwarded. If they do show up, they can vote, but I suspect turnout for inactive voters is in the single digits.

Independent voters may be caught up in voter registration drives.  For some
independent" is synonymous with "dunno".  Though independent registration continue to rise, independent registration tends to drop a greater percentage after post-election purges.
Logged
Oldiesfreak1854
Atlas Icon
*****
Posts: 13,674
United States


WWW Show only this user's posts in this thread
« Reply #1 on: August 03, 2015, 12:22:44 PM »

Certainly, I agree that electronic voting machines should be banned.  Other than that, I don't know what to make with this.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #2 on: August 03, 2015, 03:44:12 PM »

Certainly, I agree that electronic voting machines should be banned.

With whom do you agree. Why do you believe this.

Other than that, I don't know what to make with this.

I'll take it a step at a time.  The following chart shows the percentage of GOP registration (vertical axis) versus the number of registered voters (horizontal axis) for Sedgwick County, KS.



You should be able to identify black precincts. 

There does not appear to be a relationship between the size of the precinct and the GOP registration share.  For example, there is a precinct with 1466 registrants, that is 70.5% R and 10.0% D (Minneha MI01) on the eastern edge of the county, centered on the Crestview Country Club; and a precinct with 1476 registrants, that is 3.0% R and 65.2% D (Wichita 0114) in north Wichita.

There does appear to be some polarization.  The countywide Republican percentage in Sedgwick County is about 40.6%.  But there doesn't appear to be a lot of precincts around 40%, there is a gap between more Republican leaning and Democratic-leaning precincts.

Let me know if you understand this part.
Logged
Torie
Moderators
Atlas Legend
*****
Posts: 46,057
Ukraine


Political Matrix
E: -3.48, S: -4.70

Show only this user's posts in this thread
« Reply #3 on: August 03, 2015, 05:07:37 PM »

I can't understand any of this either. One thing that would be relevant is consistency. If precincts have consistent patterns, and there is fraud, then the fraud must be replicated with remarkable consistency, election after election, in just about the same degree. And all over Kansas. It's remarkable that if such widespread fraud, somebody, somewhere would not have gone rogue, and revealed at least some individual instances.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #4 on: August 03, 2015, 08:00:11 PM »

I can't understand any of this either. One thing that would be relevant is consistency. If precincts have consistent patterns, and there is fraud, then the fraud must be replicated with remarkable consistency, election after election, in just about the same degree. And all over Kansas. It's remarkable that if such widespread fraud, somebody, somewhere would not have gone rogue, and revealed at least some individual instances.
It might be be hard to understand.  That's what I'll take it a step at a time.

The following chart shows the percentage of GOP registration (vertical axis) versus the number of registered voters (horizontal axis) for Sedgwick County, KS.



This is the data being displayed in the chart.

Sedgwick Registration

You should be able to identify black precincts. 

There does not appear to be a relationship between the size of the precinct and the GOP registration share.  For example, there is a precinct with 1466 registrants, that is 70.5% R and 10.0% D (Minneha MI01) on the eastern edge of the county, centered on the Crestview Country Club; and a precinct with 1476 registrants, that is 3.0% R and 65.2% D (Wichita 0114) in north Wichita.

There does appear to be some polarization.  The countywide Republican percentage in Sedgwick County is about 40.6%.  But there doesn't appear to be a lot of precincts around 40%, there is a gap between more Republican leaning and Democratic-leaning precincts.

Let me know if you understand this part.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #5 on: August 04, 2015, 07:31:43 PM »

I can't understand any of this either. One thing that would be relevant is consistency. If precincts have consistent patterns, and there is fraud, then the fraud must be replicated with remarkable consistency, election after election, in just about the same degree. And all over Kansas. It's remarkable that if such widespread fraud, somebody, somewhere would not have gone rogue, and revealed at least some individual instances.
It might be be hard to understand.  That's what I'll take it a step at a time.
You are familiar with cumulative vote share, if you have watched election night returns.  As the night wears on, the share of the anticipated total votes cast will be reported, along with the share of the votes for the candidate.

"With 8% of precincts reporting, Senator Bugtussle leads with 52% of the vote".

Assuming you are familiar with Pennsylvania, you will know the early cumulative vote will be extremely Democratic.  That is because Philadelphia reports first.  As the night goes on the Republican share will continue to climb, because everywhere in the state is more Republican than Philadelphia.

In some states, the cumulative candidate share will increase or decrease, and we can trace that to a vote dump from a particularly liberal or conservative county.

I have updated the spreadsheet to include the cumulative vote, and a chart of the vote (2nd Sheet)

Sedgwick County spreadsheet

The precincts were accumulated in alphabetical order. In Sedgwick County, this introduces a geographical bias, because the numbered precincts are in the city of Wichita ('bias' is used in a statistical sense, not a sociological sense).

Within Wichita, the precincts are numbered by Wichita City Council district (PDF)

District 1 is in northeast Wichita, and includes the highest concentration of blacks.  As these precincts are added in, the Republican registration share drops to 24.0% at about 33,000 registered voters.

District 2 is in eastern Wichita, and is a newer area (you can tell on the map by which areas have quarter sections, and which have mostly full sections, and which have a lot of curving streets). By about 72,000 the cumulative GOP share has increased to 36.2%.

District 3 in southern Wichita.  Wichita appears to have developed on a north/south axis parallel to the Arkansas (and Little Arkansas) rivers. It has some quite large floodways which suggest flooding in some of these areas.  The highest concentration of Hispanics is to the northwest of downtown, but there is a secondary concentration towards the south. This suggests that the area was historically more of a working class white area. The cumulative GOP share drops to 33.6% at about 98,000.

District 4 is in southwest Wichita. It appears to be a mix of areas, and divided up by freeways and airbase.  The western extent is along US 34, so there may be areas lined with motor courts, etc. The cumulative GOP share inches up to 34.4% at about 129,000.

District 5 is in far west Wichita. It is clearly a newly developing area.  The GOP share increases to 38.2% at about 170,000.

District 6 is in northwestern Wichita. It causes the GOP share to drop a bit to 37.8%, at around 201,000. Regardless of the order in which we would add the Wichita precincts, we would always end up at this point.

Then finally we add in the area outside Wichita. Precinct IDs have a two-letter prefix indicating a city or township followed by a two digit number, so we begin with 'AF' for Afton and continue through 'WA' for Waco. As we would expect, the areas outside the city are more Republican and bring the final total up to 40.6%.

Districts 2 and 5, as well as the area outside Wichita are about 48% Republican (46% for District 2, 50% for District 5, and 49% for outside the city). As we would expect the areas that are added in last have less of an effect. This is the same thing that happens on election night, when 95% of the vote is in, and a candidate has 46.1% of the vote in a two-way race. We know he has lost, because the last five percent of votes can't change the result. If he is at 46.1%, with 5% counted, we know it is too early to know. If he is at 46.1% with 50% of the votes counted, he will assure his supporters that our best boxes are still out.

As it turned out the Sedgwick county election server crashed twice on election night in November 2014. If the precinct results had to be physically transported to a central location, we might expect that the results from close to the election commission office would be received first, and initial results would skew Democratic. If the server crashed, and then the Republican precincts from further out were added in, leftist hacks might claim conspiracy, particularly if they get a mathematician with a PhD in statistics to prove their suspicions.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #6 on: August 05, 2015, 05:04:18 PM »

I can't understand any of this either. One thing that would be relevant is consistency. If precincts have consistent patterns, and there is fraud, then the fraud must be replicated with remarkable consistency, election after election, in just about the same degree. And all over Kansas. It's remarkable that if such widespread fraud, somebody, somewhere would not have gone rogue, and revealed at least some individual instances.
It might be be hard to understand.  That's what I'll take it a step at a time.
You are familiar with cumulative vote share, ....

I have added a third sheet that accumulates the precincts in random order.

Sedgwick County spreadsheet

I used the RAND() function to generate a random number for each precinct, then fixed those values, and then sorted the precincts based on the random number. Thus I have used one of 250! (250 factorial) possible random orders.

As one should expect, the cumulative Republican percentage converges to the average value of 40.6% and stays there. The initial convergence takes a bit longer than I might have expected, but there are several factors that influence this.

There are only 250 precincts in Sedgwick County, and a relative large number are small. 41 have less than 200 registrants. The smaller precincts in Sedgwick County in general are not rural, but rather on the fringes of cities, where annexations have fragmented townships. Kansas requires that election precincts be contiguous, and conform to township and city boundaries, as well as legislative and other district boundaries.

With a small number of samples, the average for a sample is less likely to be near the mean of the total population. Flip a coin 10 times, and the chances that it will be heads between 45% and 55% of the time is about 1/4.  Flip a coin 1000 times, and the chances it will be heads between 45% and 55% of the time (between 450 and 550 heads) is almost a certainty. Note: ordering the precincts randomly is analogous to shuffling a decade of cards, since a precinct can only be counted once. The order of the remaining precincts is dependent on the order of the preceding precinct, since they can't be repeated. This is unlike rolling a die, where the same value can come up repeatedly.

While the precinct average is 40.6%, this is not a typical precinct. Precincts that are either more Republican or more Democratic are much more common. The histogram shows the percentage of total registrants who live in precincts with a particular GOP registration share. For example, 6.7% of the voters live in precincts that are 40% +/- 2.5% (37.5% to 42.5%).   

While 6.7% live in precincts that are about 40% Republican, 18.0% live in precincts that are roughly 50% Republican, and 13.9% live in precincts that are 30% Republican. The population is polarized with a Democratic maximum, and a Republican maximum on either side of the "average" precinct.

When accumulating results, it would be like rolling a die numbered: 1, 2, 3, 5, 6, 7. That is the four is missing. No matter how many times you rolled the die, it would never land on 4. But the average roll would be a 4.0.  Because of the polarized results, convergence to the mean is slower.

There are some extremely Democratic precincts in Wichita (because they are predominately black). There are more precincts that are Republican than are Democratic, but the Democratic precincts tend to be more Democratic.  When adding the precincts in random order, the small number of extreme Democratic precincts will tend to be more sporadically distributed, which will cause the cumulative GOP share to take longer to converge on the mean.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #7 on: August 07, 2015, 03:38:30 AM »

I can't understand any of this either. One thing that would be relevant is consistency. If precincts have consistent patterns, and there is fraud, then the fraud must be replicated with remarkable consistency, election after election, in just about the same degree. And all over Kansas. It's remarkable that if such widespread fraud, somebody, somewhere would not have gone rogue, and revealed at least some individual instances.
It might be be hard to understand.  That's what I'll take it a step at a time.
Beth Clarkson claims that as precinct sizes increases, that the Republican share of the vote increases.  She says that this is inexplicable other than by manipulation of the voting machines.

So we will test her claim, by ordering the precincts by number of registered voters.

Sedgwick County spreadsheet See Cumulative GOP, Size Order.

This doesn't appear to support Clarkson's claim at all. It looks similar to the curve for random order.  The right side might be somewhat more volatile, but now the precincts are uniformally becoming larger, so that there are fewer precincts needed for a given total registration precinct. The largest precincts are about 2,000 (with a singular precinct over 3,000), so that it only takes 10 precincts to add 20,000 to the cumulative total. With the polarization of the population, in a small sample there can easily be more or fewer Republican or Democratic precincts than typical.

For certain sized precincts, there is a geographic connection. The smallest precincts are predominately outside Wichita. Annexations can chop up townships forcing creation of small precincts. In Kansas, precincts must be contiguous, so an annexation that isolates a small part of township will create a new precinct. The newly annexed area may also be in a new precinct if a legislative district boundary followed the former city limits.

While small precincts in Wichita are uncommon, they do exist. For example precinct 0321 has 67 voters in a 4-square-block area, created where a senate district boundary zigged, while a house district boundary zagged. Being outside Wichita, the small precincts tend to be more Republican.

There is a minimum at about 38,000 voters. There is a string of 8 districts in a row that are more Democratic than average. There is unlikely to be a systemic cause of districts between 770 and 870, that causes them to be Democratic, so it was just a case of luck.

The increase in GOP registration share between 38,000 and 75,000 corresponds to precincts between about 850 and 1200 voters. A somewhat large share of these are outside Wichita. It is not that there are not Wichita precincts of this size, but they are on the lower range for Wichita. With more of these outside Wichita, they tend to be somewhat more Republican than the county as a whole.

A large share of Wichita precincts are in the range 1000 to 1900 voters. Precincts that are somewhat larger could be split and fall in this range, precincts that are somewhat smaller could be merged into precincts in that range. This appears to be the "normal" size for a precinct that is a section or half-section in an urban area.

Why are our results different than Clarkson's? We used the actual precinct registration. She used the number of votes cast in a precinct. While larger precincts will tend to have more votes cast, the turnout (votes cast divided by registration) may vary. A smaller precinct with higher turnout may have more votes cast than in larger precinct with lower turnout.
Logged
muon2
Moderator
Atlas Icon
*****
Posts: 16,788


Show only this user's posts in this thread
« Reply #8 on: August 07, 2015, 08:44:35 AM »

Clarkson claims that the effects only arise with precincts having more than 500 votes. If I only take those precincts with over 500 votes the GOP% is 76.2%, and if restrict it to those precincts with over 1000 votes then the GOP% is 81.2%. Compared to the countywide average of 40.6% this could look suspicious.

However, I think that there may be two effects that contribute to Clarkson's claims, neither of which has to do with the official in charge of voting. One obvious contribution comes from turn out. The countywide turnout of registered voters is 53.4%. Among the large precincts with over 500 votes cast the turnout is 54.6% and for the precincts with over 1000 votes cast the turnout is 59.6%. Since the large precincts are also more Pub this creates a correlation between turnout and GOP%. That isn't a surprise and has been observed many time before. I don't know if Clarkson controlled for that variable.

The second effect is hard to ferret out of the data but is one I observe in IL. Suburban and exurban precincts tend to be the largest. Around me this is due to the growth patterns and the time lag to adjust precinct sizes to population. As urban areas decline and suburbs grow one tends to find urban precincts that are smaller than suburban precincts. I suspect KS is like IL in this regard. If it is the case, then given that suburban areas will be more Pub than urban areas this will result in a bias towards large precincts favoring Pubs.

Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #9 on: August 08, 2015, 04:28:02 AM »

Clarkson claims that the effects only arise with precincts having more than 500 votes. If I only take those precincts with over 500 votes the GOP% is 76.2%, and if restrict it to those precincts with over 1000 votes then the GOP% is 81.2%. Compared to the countywide average of 40.6% this could look suspicious.
I have added a chart "Cumulative GOP, Vote Order" that shows this in the form similar to what Clarkson has used. The precincts are ordered based on the votes cast in November 2014, while the values charted are cumulative Republican registration share vs. cumulative registration. Clarkson uses cumulative Republican vote share vs. cumulative votes cast.

I think you have compared the wrong columns, though the effect is the same.

For precincts with over 1000 votes cast, the GOP registration share is 48.7%. For precincts with over 500 votes cast, GOP share is 41.6%, but only 37.2% for precincts between 500 and 1000.  For precincts with less than 500 votes, the GOP share is 35.5%.

This concept of precincts with more than 500 votes cast being different, first came from Republican primaries, where Paul supporters claimed that votes were being flipped by Romney-supporting election official. They used votes cast because it could be computed directly from the election returns.

In a Republican primary, turnout measured against total registration can vary a large amount. The number of Republican voters in a primary tells you a lot more about the precinct than its size. I took a look at the 2012 Republican primary in Orange County, California, and not only did turnout correlate with the Republican registration share, turnout among Republicans correlated with the Republican registration share. That is, if you had two precincts with 1000 voters, one that was 25% Republican and one that was 50% Republican, there would be more than twice the number of votes cast in the primary in the more Republican precinct.

Areas that have high Republican registration, are likely to be wealthier, be more home-owning, be older, be married, and be more economically stable. These factors would increase the chance that a registered voter still lives at the address they are registered at, and in California that they are permanent vote by mail. Such voters would be more likely to vote, and to vote conventionally (eg for Romney).

However, I think that there may be two effects that contribute to Clarkson's claims, neither of which has to do with the official in charge of voting. One obvious contribution comes from turn out. The countywide turnout of registered voters is 53.4%. Among the large precincts with over 500 votes cast the turnout is 54.6% and for the precincts with over 1000 votes cast the turnout is 59.6%. Since the large precincts are also more Pub this creates a correlation between turnout and GOP%. That isn't a surprise and has been observed many time before. I don't know if Clarkson controlled for that variable.
Clarkson did not, since she used votes cast as her ordering value (and also calculating the Republican vote share). There is a correlation between GOP registration share and turnout.

The registration statistics include inactive voters, who are voters that election materials have been sent, but where there was no forwarding address. Election officials can't remove someone from the voting rolls unless they have a confirmation that they have moved, or died. If they move and re-register, they can be removed, but only if the election officials know that they are the same person. This is likely if the move was within a county, possible within a state, less likely between states. Even if someone who has moved between states is detected with a cross-check program, they have to be affirmatively contacted. An inactive voter can be removed after not voting for two general elections.

An inactive voter can vote the same as an active voter - they are just much less likely to, because they probably have moved. Democrat voters may tend to be less self-motivated to vote. This is also true of independent voters, since some probably were picked up in registration drives, and did not indicate a party. Not checking a party box may indicate independence, but it may also indicate indifference.

It is not universal that being more Republican leads to higher turnout. For example, in the Cook County suburbs (ie Cook County Board of Elections), turnout in Evanston and Oak Park was every bit as high as New Trier and Barrington, but at the same time turnout was much lower in Cicero.  That is, turnout may correlate with economic status, and GOP support may correlated with economic status (Evanston and Oak Park are counter-examples).

Quote
You must be logged in to read this quote.
This might be true in Kansas. 

I tried to discover a relationship in Sedgwick County. There is almost no correlation between GOP registration share and precinct size. If I exclude precincts with less than 500 voters, there might be a weak relationship. If I do a weighted least square fit, using registered voters as both the independent variable and the weight, it appears that there might be about a 0.45% increase in GOP share for every increase of 1000 voters. But most typical precincts are between 1000 and 2000 voters, so that would only be a small increase among the smallest and largest typical precincts.

About 3/4 of Sedgwick County voters are in Wichita. There aren't areas that are outside the city that are developed enough to have a regular precinct structure.

Precincts in Sedgwick County come in fairly regular sizes:

Townships.
Sections.
Half-Sections
Quarter Sections.

But there are some irregular precincts that are larger than a section, that tend to have the most voters. And the section precincts also tend to be larger.

This is a work in progress.  The darkest three shades indicate precincts greater than 2000, 1750-2000, and 1500-1750. There is a small core of smaller precincts, mostly half-sections, with a few quarter-sections, but it is only 14 precincts. The west side appears to be the newest developing area, with lots of curving residential streets. The precincts that are half-sections are smaller.



Sedgwick County uses single polling centers for multiple precincts. There are a few instance of one polling center for one precinct, but they are in the remote parts on the county, where voters would have to travel to a neighboring township which might be 10 or more miles away. These shared polling centers even combine precincts from different cities or townships. So long as a precinct is not divided by district boundaries of some sort, it might not be any reason to divide precincts for size or convenience reasons. But this might be a new policy based on a certain type of voting equipment.

In a state like Illinois, where not only is Cook County filled in, so are the neighboring counties, and you have to go out to Kane County to find new subdivisions, there is no room for population to grow, except infill with apartments. Precincts may lose population as they mature, couples are replaced by widows or widowers, new families may be single, or non-citizen.

It may be impractical to merge precincts. The population loss might not be large enough to justify doubling. There may be political resistance. If someone has to travel three blocks instead of one blocks to vote, it will be indicated as a 200% increase. But by the same token, it may be difficult to split a precinct, since it will require a new polling place. It may be easier to add another voting machine or election clerk to handle a small increase.

Communities are created on a larger scale that assumes access to automobile. At one time cities were expanded by adding to the street grid, and home builders adding a house at a time. Some houses might have been built by the owner. So perhaps there is a relationship.

I had looked at statewide precinct results for Kansas. What was very clear was that precinct sizes vary from county to county. Wyandotte (Kansas City) and Shawnee (Topeka) had small precincts (measured by votes cast). They are two of the three Democratic counties in the state, with the third being Douglas (Lawrence). Looking at the cumulative Republican share, there would be a few Republican precincts that would increase the GOP cumulative share, and then a Democratic precinct that would cause a notch downward (Democratic precincts are less numerous, but more extreme). After the Wyandotte and Shawnee precincts stopped showing up, there was Douglas and the Democratic areas of Wichita, which were not numerous enough to match the continuing Republican counties. Butler County, which might be considered exurban, but certainly suburban, has some very large precincts.  The precincts in Johnson County, had more votes cast as you go west and south, but that might be due to their having higher turnout.

So in Sedgwick County, at least, the key factor is differential turnout. I think after the shenanigans with the Democratic senate nominee, that Republican turnout was enhanced.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #10 on: August 08, 2015, 01:29:23 PM »

Clarkson claims that the effects only arise with precincts having more than 500 votes. If I only take those precincts with over 500 votes the GOP% is 76.2%, and if restrict it to those precincts with over 1000 votes then the GOP% is 81.2%. Compared to the countywide average of 40.6% this could look suspicious.

However, I think that there may be two effects that contribute to Clarkson's claims, neither of which has to do with the official in charge of voting. One obvious contribution comes from turn out. The countywide turnout of registered voters is 53.4%. Among the large precincts with over 500 votes cast the turnout is 54.6% and for the precincts with over 1000 votes cast the turnout is 59.6%. Since the large precincts are also more Pub this creates a correlation between turnout and GOP%. That isn't a surprise and has been observed many time before. I don't know if Clarkson controlled for that variable.

The second effect is hard to ferret out of the data but is one I observe in IL. Suburban and exurban precincts tend to be the largest. Around me this is due to the growth patterns and the time lag to adjust precinct sizes to population. As urban areas decline and suburbs grow one tends to find urban precincts that are smaller than suburban precincts. I suspect KS is like IL in this regard. If it is the case, then given that suburban areas will be more Pub than urban areas this will result in a bias towards large precincts favoring Pubs.
I have added two charts showing Turnout% v GOP Registration%.  The second one excludes precincts with less than 100 votes cast.

I also calculated a weighted least-squares linear fit.

The first uses the number of registered voters as the weight:

Turnout = 26.7% + 0.657% * GOP%

The second uses the number of cast votes as the weight.

Turnout = 27.6% + 0.647% * GOP%

The precincts with very low GOP% are above the trend line, suggesting that heavily black precincts did not have depressed turnout to the extent that working class white and Hispanic areas did.

If we project the results for GOP% = 0% and 100%, we come up with a remarkable estimate that turnout among non-Republicans was about 27%, while among Republicans it was over 90%.

Voter registration data is available, including party affiliation, and voting history, so it is possible to verify that individual voter records are consistent with aggregate data, without examining secret ballots.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #11 on: August 08, 2015, 09:55:16 PM »

The latest chart is fairly remarkable.

Sedgwick County spreadsheet

The chart 'Roberts% v Size Order' shows the cumulative Roberts percentage in the 2014 Senate election, with the precincts arranged by increasing registration.

The chart shows:

y = sum(Roberts) / sum(votes)
x = sum(registration)

So even though the GOP registration percentage is neutral with respect to precinct size, the Roberts vote decline with larger precincts.

Possible factors include that suburban precincts are somewhat smaller, and supported Roberts to a greater extent than Republicans within Wichita.  The correlation between the GOP registration share and the Roberts share is only 0.735, so it was far from simple straight-ticket voting.


Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #12 on: August 09, 2015, 09:26:53 AM »

The latest chart is fairly remarkable.

Sedgwick County spreadsheet

The chart 'Roberts% v Size Order' shows the cumulative Roberts percentage in the 2014 Senate election, with the precincts arranged by increasing registration.

The chart shows:

y = sum(Roberts) / sum(votes)
x = sum(registration)

So even though the GOP registration percentage is neutral with respect to precinct size, the Roberts vote declined with larger precincts.

Possible factors include that suburban precincts are somewhat smaller, and supported Roberts to a greater extent than Republicans within Wichita.  The correlation between the GOP registration share and the Roberts share is only 0.735, so it was far from simple straight-ticket voting.
The last two charts show the relationship between the Roberts vote share and the registration size.  The first chart masks precincts with less than 200 voters, while the second chart excludes them, thus the key difference is in the calculation of the trend line:

Roberts% = 54.6% -3.52% per 1000 voters.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #13 on: August 09, 2015, 11:26:39 AM »
« Edited: August 12, 2015, 11:39:31 AM by jimrtex »

Summary of Data

Sedgwick County spreadsheet

Sedgwick County Overview

Sedgwick County has 440 election precincts. 130 of these have no registered voters. Kansas law requires election precincts to be self-contiguous, and to not cross city and township boundaries, nor cross legislative and other district boundaries. Annexations can leave unincorporated enclaves, and some cities have detached exclaves. If an annexation occurs after district boundaries were defined, additional small precincts may be created.

In 2014, Sedgwick County consolidated 100 election precincts into 40 precincts for purposes of voting. This was done where the election precincts shared a polling location and a ballot style (ie they had the same races).

This consolidation resulted in 251 precincts that had election results reported for the 2014 election. One of these (DL02) had zero votes cast and I have excluded it. It is data from these 250 consolidated election precincts that I have analyzed.

Precincts in the city of Wichita have a 4-digit precinct number. The first digit is always 0, and the second (hundreds) digit indicates which of Wichita's six city council districts the precinct is in.

Other precincts in the county have a two letter prefix indicating the city or township (eg 'AT' is for Attica township), followed by a two digit number.  If there is only one precinct (or one main precinct) for the city or town, the number may be omitted (eg 'GO' is the precinct for the city of Goddard, and GO01 ... GO04 are additional precincts. 98% of Goddard voters are in GO).

Raw Data Sheet

Registration data is from May 22, 2015, and includes party registration (Democratic, Libertarian, Republican, and independent(unaffiliated). Sedgwick County maintains registration using the 440 base election precincts. I have consolidated this data to match the 250 consolidated precincts used for the 2014 general election.

The 2014 election data included total registration, but did not include partisan information. The 2014 total registration is consistent with the 2015 total registration for each precinct. I have used the 2015 data since it includes a partisan breakdown. Arguably, the 2014 data should be used for measuring turnout, but it should make negligible difference.

Registration includes both active and inactive voters. An "inactive" voter under federal law is a voter who has apparently moved from the county, but which has not been confirmed by the voter. It includes voters whose confirmation notice was returned as having no forwarding address. Inactive voters may vote, but most likely will not. They are removed from the voting rolls if they miss two federal elections after they have been sent a confirmation notice.

The number of ballots cast, and the senatorial and gubernatorial results are from the November 2014 elections. They are reported directly for the 250 consolidated precincts.

GOP v Registration Data Sheet GOP v Registration Chart, GOP v Registration (Wichita) Chart

This sheet calculates the GOP registration share for each precinct. The chart displays the GOP registration share vs the number of registered voters in each precinct. There does not appear to be a strong relationship between GOP registration and the size of the precinct. Using a weighted least-squares linear fit (weights are the number of voters in each precinct), yields a slope of less than 0.5% per 1000 voters).

Countywide, the GOP registration is 40.6%. There appears to be a gap in the data, with most precincts being more Republican or less Republican than the average. That is, the average is not typical.

The second chart is limited to precincts in Wichita with more than 200 registered voters. There may be a weak relationship between precinct size and partisanship (correlation is only 0.125 though). The precincts in city council districts 2 and 5 are on average larger (15% or 20%). These precincts are in the newest areas of the city on the east and west side, and are more Republican. The size difference may reflect larger eligible populations or higher registration rates.

Alphabetical Order Sheet Cumulative GOP, Alphabetical Order Chart

This sheet calculates the cumulative GOP registration share. To calculate the cumulative GOP registration, the total registration and the GOP registration are added one precinct at a time. For the ith precinct,

GOP_share(i) = sum(GOP1 ... GOPi) / sum(registration1 ... registrationi)

The chart shows the cumulative results as precincts are added in alphabetical order. Since election precincts are numbered by Wichita city council district, we can observe segments of the cumulative curve corresponding to each district.

DistrictAreaGOP%Regis.GOP%Regis.
1Northeast24.0%3288424.0%32884
2East46.3%3941236.2%72296
3Southeast26.3%2581933.6%98115
4Southwest37.1%3064634.4%128761
5West50.4%4075938.2%169520
6Northwest35.2%3109437.8%200614
Non-Wichita48.6%6967440.6%270288

The two most Republican districts, Districts 2 and 5, have the highest registration, even though presumably all six districts have similar populations.

Random Order Sheet, Cumulative GOP, Random Order Chart

This sheet accumulates the precincts in random order, or more correctly, a random order. A random number was generated for each precinct, and then the precincts were sorted based on the random numbers.

The chart shows the cumulative GOP% registration for randomly ordered precincts. As we would expect, the cumulative GOP% converges on the countywide share of 40.6% and stays fairly close to that value. The convergence is somewhat slower than might be expected. There are three contributory factors.

(1) There are relatively few precincts, 250 in total. Moreover, many are quite small. 40 have less than 200 voters, and contribute little to the overall result.
(2) The precincts are polarized. While the county is 40.6% GOP registration, most precincts are more Republican (50% or so GOP), or less Republican (30% or so GOP). A precinct that has about 40% GOP registration is atypical.
(3) There are a few extremely Democratic precincts (eg 10% GOP ) in areas that are predominately black. 56% of voters are in precincts that are more Republican than the countywide 40.6% GOP Registration. This preponderance of voters in more Republican areas is counterbalanced by 10 or so extremely Democratic precincts. With so few of these precincts, it is unlikely that they would be evenly distributed among a random sequence.

GOP% Distribution Chart

This chart shows the percentage of voters who reside in precincts with a particular range of GOP registration. For example, 18.0% of voters reside in precincts that are approximately 50% Republican registration. Each bar corresponds to a band 5% wide. So more precisely, 18.0% of voters reside in precincts that are between 47.5% and 52.5% Republican.

While overall the county is 40.6% Republican, relatively few voters reside in precincts that are 40% Republican, compared to the number who live in precincts that are 50% Republican or 30% Republican. This polarization of the electorate, may contribute to a misperception of election results.

Size Order Sheet, Cumulative GOP, Size Order Chart

This sheet and chart order the precincts by number of registered voters. The cumulative GOP% is initially more Republican, due to the smallest precincts being outside Wichita. The small size is not due to the area being rural or agricultural, but due to precinct fragmentation caused by annexation and irregular city boundaries. Within Wichita, the most common precinct size is in the range of 1000 to 1900. Precincts that are somewhat large can be split, and the two halves fall within the range, while precincts that are somewhat smaller can be merged with the resulting whole being within the range.

There appears to be a sensibility that precincts should either be full sections or half sections. This provides recognizable boundaries, since arterial streets are usually along section lines, or half section lines, and may also correspond to sense of neighborhood or community.

The increase in Republican percentage between a cumulative registration of 38,000 and 76,000 reflects a larger share of precincts from outside Wichita in the range of 850 to 1250 registered voters. Within Wichita, precincts of this size are less typical (not extraordinary, but merely slightly less ordinary). Within that range of precinct sizes, there is a larger contribution to the total from outside Wichita, than there is for larger-sized precincts.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #14 on: August 10, 2015, 02:26:59 PM »

The second effect is hard to ferret out of the data but is one I observe in IL. Suburban and exurban precincts tend to be the largest. Around me this is due to the growth patterns and the time lag to adjust precinct sizes to population. As urban areas decline and suburbs grow one tends to find urban precincts that are smaller than suburban precincts. I suspect KS is like IL in this regard. If it is the case, then given that suburban areas will be more Pub than urban areas this will result in a bias towards large precincts favoring Pubs.
I did find a small effect if I limited the data to Wichita precincts with more than 200 registered voters. Precincts less than 200 voters are generally caused by district boundaries slicing off a small area that has its own unique ballot style (combination of legislative and other races). They may also be caused by city annexations, where legislative district boundaries coincide with the former city limits. The newly annexed area must then be place in a small precinct, since it is now within the city, but still in its legislative district for the remainder of the decade, at least. Because of their size they may be considered noise, since there may be no demographic association with their size.

Within Wichita, precincts in city council districts 2 and 5, tend to be a bit larger than precincts in the other districts.  Districts 2 and 5, are the most Republican districts, and the newest developing areas on the east and west side. The size difference may reflect that there has been less division and have higher CVAP, or it may be related to higher registration rates.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #15 on: August 11, 2015, 08:42:13 PM »

The second effect is hard to ferret out of the data but is one I observe in IL. Suburban and exurban precincts tend to be the largest. Around me this is due to the growth patterns and the time lag to adjust precinct sizes to population. As urban areas decline and suburbs grow one tends to find urban precincts that are smaller than suburban precincts. I suspect KS is like IL in this regard. If it is the case, then given that suburban areas will be more Pub than urban areas this will result in a bias towards large precincts favoring Pubs.
Though the effects within Sedgwick County appear to be quite mild, slightly higher registration in Republican council districts of Wichita balanced by by somewhat smaller precincts outside the city, there are significant intracounty differences.

For example, Shawnee County (Topeka) has no precincts with so many as 1000 voters, while in Wichita, a precinct with less that 1000 voters is definitely a small precinct.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #16 on: August 12, 2015, 01:02:54 PM »
« Edited: August 22, 2015, 04:24:12 PM by jimrtex »

Summary of Data (Part 2)

Sedgwick County spreadsheet

Vote Order Sheet, Cumulative GOP, Vote Order Chart, Cumulative GOP (est), Vote Order Chart

In this sheet and chart, the precincts are ordered by number of ballots cast in the November 2014 general election. This is the order that Beth Clarkson used. While she claimed that the precincts were ordered by size, they were ordered by ballots cast. The difference in this chart and the previous chart is that precincts that were more Republican had higher turnout. This does not necessarily mean that Republicans were more likely to vote. It could mean that voters in more affluent areas were more likely to vote, and that voters in more affluent areas are more likely to be Republican. In either case, this will have the effect of making the electorate more Republican, and shifting the more Republican districts to the right in the chart.

Who is registered, where they live, which precinct they live in, their party affiliation, and whether they voted in 2014, is public information. That is, if one were so inclined, the data in this chart could be audited, included directly determining if there is any relationship between party affiliation and whether someone voted or not.

The first chart does not show the composition of those who voted, but does show that as the number of votes cast increases, the precincts have a larger share of registered Republicans. The second chart shows an estimate of the number of Republican voters. It uses a simple model. If 60% of registered voters in a precinct vote, then 60% of registered Republicans, 60% of registered Democrats, etc. voted. That is, it assumes that there are demographic factors such as income, education, age, race, marital status, race, etc. that increase the likelihood of voting; and that these same factors increase the likelihood that a voter will register as a Republican. That is, in this simple model, being a Republican does not directly increase the likelihood of voting - Republicanism correlates with increased turnout, but does not cause it.

For the second chart, we estimate the Republican voters for a precinct as: GOP% registration multiplied by ballots cast in the precinct. We calculate the cumulative Republican percentage using these estimates and display it as a function of the cumulative number of ballots cast (red curve). In addition, we display the cumulative registration share from the previous chart. This blue curve underestimates the GOP% share since it in effect assumes uniform turnout.

Turnout v GOP% Chart, Turnout v GOP% (precincts more than 100 votes) Chart

These two charts show the relationship between turnout and the GOP registration share. The second chart excludes precincts with fewer than 100 votes cast. This should make the trend line more reliable. As we have noted before, there are relatively few precincts with about 40% GOP registration, while there are more precincts around 50% and 30% registration. The 50% GOP precincts had around 60% turnout, and the 30% GOP precincts had around 45% turnout, which is consistent with the trend line.

If we assumed that whether someone voted was entirely dependent on their party affiliation, we could use the trend line to predict the probability that a registered Republican voted, as well as the probability that a registered non-Republican (Democrat, Libertarian, or independent) voted. If there were no other predictive factors, we could draw a precinct around each voter, which would be either 100% GOP or 0% GOP.

The GOP-registered voters would have a 93.3% probability of having voted, and non-Republicans would have a 25.5% probability of having voted. This is so implausible that we must conclude that there are other factors besides, or in addition to, partisan registration that drive voter participation.

Trend line:

turnout% = 25.5% + 0.68%*GOP%  Unweighted, precincts less than 100 votes excluded.
turnout% = 26.7% + 0.66%*GOP%  Weighted by precinct registration
turnout% = 27.6% + 0.65%*GOP%  Weighted by precinct turnout.

Cumulative Roberts, Size Order Sheet, Roberts%, Size Order Chart

This sheet and chart show the cumulative vote share for Pat Roberts in the US Senate race, with the precincts ordered by registration. This is what Beth Clarkson purported to show. But she did not order the precincts by size (number of registered voters), but rather by the number of votes cast.

A surprise is that there was a decline in the Roberts vote share as larger precincts were added in. Looking carefully at the cumulative GOP share, it appears that there was a small decline (about 2%), which he had characterized as being negligible. The cumulative Roberts vote share declines about 4%. This indicates that the Roberts% minus GOP% difference does decline as larger precincts are added in.

Roberts% v Registration Chart, Roberts% v Registration (small precincts excluded) Chart

These two charts show the relationship between Roberts support and the number of registered voters. The first chart hides values from precincts with fewer than 200 voters, while the second chart excludes them. Thus they are also excluded from calculation of the trend line.  The trend line line shows a decline of about 3.5% in Roberts support per increase of 1000 voters.  This contrasts with about 0.5% increase per 1000 voters for GOP registration.

It is important to remember, that just because a precinct had 1500 registered voters, who voted 40% for Roberts, it does not mean that Roberts got 600 votes to 900 for Orman and Batson. The 40% is measured among those who actually voted.

Roberts% v Senate Votes Chart, Roberts% v Senate Votes (small precincts excluded) Chart

These two charts present the Roberts share of the vote as a function of the votes cast in the senate race. The second chart excluded precincts with fewer than 100 votes from the trend line calculation.

The trend of larger precincts tending to support Roberts, is reversed. This does not mean that larger precincts tended to support Roberts, but rather that the larger precincts that supported Roberts tended to have higher turnout. A large precinct with high turnout will have a large number of votes cast. A large precinct with low turnout will have a moderate number of votes cast.

Countywide, Roberts had 50.9% support. His support exceeded the countywide average in 54.3% of precincts. The median is higher than the mean. There are some extremely Democratic, predominately black precincts, in the northeast part of the city. Three precincts between Washington and Hillside had Roberts support below 7%. They also had turnout below 35%, which is far below the countywide turnout of 52.7%. There is a cluster of precincts below 30% Robert support, with between 400 and 800 votes cast, all of which are in the northern part of the city, and all but three in city council district one.

There are two precincts with around 1200 votes cast, and about 30% Roberts support. They are located in the Riverside neighborhood between the Arkansas and Little Arkansas rivers, northwest of downtown. They had average turnout, but were supportive of Orman. In precinct 0607, the GOP had 36.6% registration, but Roberts only received. 29.3% of the vote.

The 10 precincts with the most votes all supported Roberts at a rate greater than the county wide average. The three precincts with the highest Roberts support are GD (Grand River township), UN01 (Union township), and GA (Garden Plain township). These are west of the city. While they are not urban, the population density (51 in Garden Plain, 70 in Union) is too high to be supported by wheat farming. They are within commuting distance of Wichita, so that people can live in the country, while earning a living in the city. These precincts were heavily Republican to start with (above 60% GOP registration, about 20% higher than the county average), had high turnout (around 60%, about 8% to 10% higher than the county average), and very high Roberts support (about 75%, about 25% higher than county support). Rather than Orman peeling off a few Republican votes, even independents were voting for Roberts.

Cumulative Roberts, Vote Order Sheet, Roberts%, Vote Order Chart

This sheet and chart show the cumulative Roberts share, with the precincts ordered by votes cast in the Senate race. This is typical of the results produced by Beth Clarkson.

The minimum occurs at 49,612 cumulative votes cast, 47.4% Roberts support. This corresponds to a precinct with 698 votes cast. This follow a downward dip beginning at 32,806 votes cast and 49.8% support. The dip begins at a precinct with 596 votes cast. If we look back at the previous chart, we will see that there is a cluster of precincts with extremely low Roberts support, between about 600 and 700 votes cast. When these precincts are added in, the cumulative support drops. But once we are past these precincts, the cumulative vote begins to climb back towards the countywide average.

The final part of the cumulative curve (past 127,481 50.0%), which appears to be a straight line,  corresponds to the 10 precincts with the most votes cast, which were all more Roberts supporting than the average.

Unlike Clarkson's finding that something special happens at 500 votes cast, the minimum occurs at 700 votes cast. This is the point at which the extremely Democratic precincts are largely exhausted. From a statistical point of view, they are outliers.

On a statewide basis, the minimum does occur with smaller precincts. The precincts in Wyandotte (Kansas City, Kansas) and Shawnee (Topeka) are particularly small, considering that these two counties are the third and fourth most populous counties in the state. These two counties, along with Douglas (Lawrence) are the only three large Democratic leaning counties. Once the votes from those two counties were added in, only Douglas and the Democratic areas of Sedgwick county remained to counter the Republican leaning precincts in the remainder of the state.
Logged
VPH
vivaportugalhabs
YaBB God
*****
Posts: 4,682
United States


Political Matrix
E: -4.13, S: -0.17

Show only this user's posts in this thread
« Reply #17 on: August 12, 2015, 08:05:11 PM »

As a Sedgwick County resident, this is INCREDIBLY in depth and well done.
Logged
Vega
Junior Chimp
*****
Posts: 5,253
United States


Show only this user's posts in this thread
« Reply #18 on: August 12, 2015, 11:02:48 PM »

Great thread.

If we were to standardize elections and have an official agency/department for their conduct like nearly every first world country does, issues like these wouldn't be so prevalent.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #19 on: August 22, 2015, 04:28:36 PM »
« Edited: August 30, 2015, 06:46:48 PM by jimrtex »

Sedgwick County Spreadsheet

Turnout%/GOP%/Registered Sheet, Turnout%/GOP%/Registered Wichita Chart, Turnout%/GOP%/Registered cities Chart, Turnout%/GOP%/Registered townships Chart

This sheet and these three charts were an attempt to display the relationship between three variables: (1) the number of registered voters, x axis; (2) the percentage of Republican registered voters, y axis; and (3) the turnout percentage, bubble size.

City of Wichita

The first chart shows precincts within the city of Wichita. The color coding indicates a general trend from more Democratic to more Republican. The more Republican a precinct is, the higher the turnout. If two precincts were of similar size, say 1500 voters, the more Republican precinct might have 60% turnout (900 voters), while the more Democratic precinct might have 40% turnout (600 voters), and more of the voters in the Republican precinct will vote for Republican candidates.

The city council districts are numbered in clockwise order. District 1 (red) is about 1 o'clock, and includes the largest concentration of black voters. These precincts have very low GOP registration, but also very low turnout. There are a few precincts that are more Republican, and also higher turnout. These are generally on the eastern edge of the district, near downtown.

District 3 (orange) is about 3 o'clock, and is consistently Democratic and low turnout. District 4 (yellow) is about 7:30 o'clock, and contains a mix of precincts, with those south of downtown along the Arkansas River performing like their neighbors in District 3 to the east, and those in the far west being similar to their neighbors in District 5 to the north.

District 6 (green) is about 10:30 o'clock. Precincts 0607 and 0609 in the Riverside neighborhood between the Arkansas and Little Arkansas rivers have the highest turnout among districts that don't have majority Republican registration. Generally, the district is less Republican than districts 2 and 5, likely because it is closer to downtown.

District 5 (aqua) is at 9 o'clock in far west Wichita, with an eastern boundary along I-235, a western loop bypassing downtown. The residential streets are generally curved, indicating they are more recently developed. Precincts 503 and 504, are not quite Republican majority, and are along the inward boundary of the district. District 2 (blue) is at 3 o'clock in the eastern part of the city. The less Republican, and lower turnout precincts are generally in the southern part of the district near I-35 (Kansas Turnpike). I-35 enters the southern part of the city from Oklahoma City and veers more east towards Kansas City leaving Wichita in the eastern part of the city.

Other Cities

The second chart shows the election precincts for the seven 2nd Class cities in Sedgwick County. Kansas has (had) quite liberal annexation laws which permitted Wichita to sweep up territory before it can be formed into suburban cities. Wichita had a major population surge during WWII and thereafter, as airplane manufacturers located away from the coasts. Previously, there was not the population base to support suburban cities. The largest of these cities is Derby, with 22K population (2010), compared to Wichita's 382K.

Only two of the cities, Bel Aire and Park City, are results of suburban development that later became incorporated. The five others are long-established towns that have grown as Wichita has grown. Most new growth is likely to occur away from their original core, and be more suburban in nature. Several of the cities are divided into wards, which result in there being additional election precincts, making the precincts in general smaller than in Wichita.

Bel Aire, northeast of Wichita has a population of 6.8K (8.6% black).  Not having wards results in having relatively large precincts. Park City is along I-135 north of Wichita, with a population of 7.3K. It is divided into wards, resulting in smaller precincts. It's alignment along the interstate appears to make it a bit less Republican than the other suburban cities, and lower turnout. Valley Center, with a population of 6.8K, north-northwest of Wichita, has annexed towards Wichita as it tries to protect territory for growth. It has the highest Republican vote share of the cities, and high turnout. It has additional precincts because of its four wards. The cities north of Wichita, including the small city of Kechi, have annexed extensively in an effort to provide for growing room, and block Wichita. The cities gained from 25% to 40% in population during the 2000s.

Goddard abuts the western-most edge of Wichita. Despite the smallest population at 4.3K, it is essentially single precinct is the largest among the cities. Goddard more than doubled during the previous decade.

Haysville, south-southwest of Wichita, has grown to a population of 10.8K. Four wards have divided the city among four precincts. The city is slightly less Republican than the county as a whole. Derby, with a population of 22.2K, is south-southeast of Wichita. Derby is divided into four wards, which are stacked north to south, with generally the areas further east being more Republican in each ward. Derby experience explosive growth in the 1950s, growing from 432 to 6,458, after McConnell Air Force Base was established north of Derby. Mulvane, with a population of 6.1K is on the county boundary between Sedgwick and Sumner counties, with Main Street running along the county line. Most of the population (5/6) is in Sedgwick County. Mulvane has created a fence line west of the city, to block off Wichita, Derby, and Haysville.

Townships

In Kansas, political townships are not separate from survey townships, though state law strongly encouraged them to be so (eg requiring a new political township to have 36 square miles, later 30 square miles). First and second class cities are, under Kansas law, supposedly independent of any township, but in Sedgwick County, this appears to be totally true only for Wichita and Bel Aire. On the other hand, they are separate for electoral purposes. The other six second class cities have separate election precincts, and do not vote for township officers. It appears that in Kansas that townships have very limited powers, which might explain why cities have annexed so much territory - the townships are legally equipped to provide municipal services.

Sedgwick County has 27 townships, which do largely correspond to the survey townships. Wichita Township has been abolished, after almost all its territory was incorporated into the city of Wichita, with small remnants assigned to adjacent townships.

Afton township is west of Wichita. The entire 36 square miles are in a single precinct with 1531 persons. This is too dense for purely agricultural settlement, so can be considered exurban. Attica township is on the western fringe of Wichita and has begun to be decimated by Wichita and Goddard into small precincts. It has 19 precincts, though only six have actual registered voters, with only three with enough registered voters (more than 200) to appear on the chart. AT02, south of the western tip of Wichita is among the less-Republican township precincts, though with 41% GOP registration it is near the county average. Attica has 2431 person, excluding the population of Goddard.

Delano township has nearly been obliterated by western Wichita, with its government taken over by the county. The largest fragment includes the zoo and some gravel pits which appear to be in the process of conversion to a regional park by the county. It has 11 persons and four registered voters. In 2014, all four voted: Roberts(R) 3, Batson(L), Orman(I) 0. Eagle township is northwest of Wichita. The 1205 residents, include 530 in the city of Bentley. In Kansas, 3rd Class cities remain part of their township(s), though they also have a separate municipal government. A single precinct covers the entire township. Erie township in the far southwestern corner of the county has 100 residents. With no direct highway links to Wichita, it is truly rural. Even school children must travel to attend school - there is absolutely no relationship between school districts and townships or cities in Sedgwick County. One section of the survey township has been annexed to the city of Wichita. Turnout in Erie Township was 75.5% (40  of 53 registered voters), with Roberts receiving 24 (60%), Orman 13 (32%), Batson 3 (8%).

Garden Plain township is west of Wichita, with the city of Goddard just beginning to nibble on it. The population is 1838, with 849 persons in the city of Garden Plain. The entire township is in a single township. Grand River township is the next further west township from Garden Plane. Grand River has 603 persons, and a single precinct. Grant township is north of Wichita, and is beginning to be fragmented by annexations by the cities of Valley Center, Park City, and Kechi. The population is 4973, with 3223 in Valley Center, 418 in Park City, 363 in Kechi, and 969 in unincorporated areas. The area in Kechi is included with unincorporated area for township elections, and election precincts. When Kechi becomes a 2nd class city, it will be treated separately. There are 12 physical election precincts, but GN01 has all but two of the registered voters, and all votes are tallied in GN01, which contains about 4/5 of the original township. It is very Republican, with Roberts gaining 67% of the vote.

Greeley township is in the northwestern corner of the county. Of the 1035 residents, 813 live in the city of Mount Hope. On the chart, the single precinct of GR is nearly covered up (630 registered, 50% GOP). It is slightly less Republican than many of the other townships. The dominance might Mount Hope may account for this. Too far away to be suburban, there will be people who work at low paying jobs such as waitress at the cafe. Gypsum township is southeast of Wichita. There are 7379 residents, including 2175 in Derby, 1777 at McConnell AFB (census designated place), and 3427 in unincorporated areas. McConnell AFB is on the western edge. Precinct GYO2 appears so be base housing, and had 8.9% turnout in 2014. Registration is 23% Republican, 16% Democrat, and 61% independent. Roberts carried the precinct 10 to 8. But most of the township is in precinct GY01, which with 2222 voters is the largest precinct outside the cities.
Logged
muon2
Moderator
Atlas Icon
*****
Posts: 16,788


Show only this user's posts in this thread
« Reply #20 on: August 22, 2015, 04:46:34 PM »

Nice graphs. Based on them, I don't see how anyone could draw any significant correlation other than higher turnout correlates to higher Pub%, especially in cities.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #21 on: August 29, 2015, 12:08:51 PM »

Cities in Kansas may annex exclaves.  Wichita has annexed almost an entire section in quite rural Erie Township about 25 miles southwest of the city.  I though that it might be a water well field, but this image from the center of the area suggest some other use.



Does Wichita have rather unique inauguration ceremonies for the mayor and city council?
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #22 on: August 30, 2015, 06:50:39 PM »
« Edited: September 01, 2015, 05:59:10 PM by jimrtex »

Sedgwick County Spreadsheet

Turnout%/GOP%/Registered Sheet, Turnout%/GOP%/Registered Wichita Chart, Turnout%/GOP%/Registered cities Chart, Turnout%/GOP%/Registered townships Chart

Townships continued

Illinois township is west of Wichita, just south of Goddard. It has 1860 residents in a single precinct that covers most of the original township. Kechi township includes the remnants of the survey township immediately north of the former Wichita townhip, and roughly 40% has been annexed into the city of Wichita. It has 9027 residents, but 6879 of those are in the 2nd Class cities of Park City (6879) and Valley Center (388). Of the remaining 1769 residents, 1456 are in city of Kechi which is a 3rd Class city, and who participate in Kechi township elections. Once Kechi becomes a 2nd Class city, the remainder of Kechi township would have 304 persons. The township is chopped into 29 election precincts, only five of which have votes tallied. About 90% of the voters are in GN03, which includes about 1/6 of the original township, and also most of the city of Kechi. At 66.6% turnout, it was one of the highest turnout precincts in the township area.

Lincoln township is in the northeastern corner of the county. Home to 523 residents, it forms a single precinct of 348 voters, which is very Republican (66.3%) and also high turnout (72.7%). Roberts carried the precinct 66% to 29% and 5% for Orman and Batson. Minneha township is immediately east of Wichita, and most of it has been annexed to Wichita. 3417 persons reside in the township, including 733 persons in the city of Eastborough, which has been engulfed by Wichita. Eastborough was one of Wichita's first suburbs, and was incorporated in 1937. When Eastborough sought annexation to Wichita, they were turned down. Wichita later attempted to annex Eastborough, but was put off. During the dispute, the legislature supposedly gave Eastborough the right to annex Wichita. Minneha has 23 election precincts, but only eight with votes cast. The three largest, with most of the vote are shown on the chart. Precinct MI01, the largest, contains the Crestview Country Club; while precinct MI02 is the city of Eastborough. They are among the most Republican precincts in the county, 70.5% and 66.3% respectively, with high turnout, 71.9% and 75.%, respectively. MI02 (Eastborough) is notable in that Roberts received 9% smaller share of the vote, than the GOP registration share (57% for Roberts vs. 66% for GOP registration).

Morton township is on the western edge of the county. Its population of 2667, includes 2094 persons in the 3rd Class city of Cheney, and 573 in unincorporated areas. The combination in a single election precinct covering the entire township, produces one of the largest precincts that are not in Wichita or the built-up area around it. Ninnescah township is southwest of Wichita. The population of 3231 includes 2481 persons in the 3rd Class city of Clearwater. Ninnescah township is divided into two election precincts, NI covers the eastern third of the township, and includes Clearwater and about 5/6 of the voters. NI01 is the western two-thirds of the township. Both precincts were similarly Republican in registration, 56.8% and 56.5%, and support for Roberts, 63% and 60%, but the more rural NI01 had higher turnout, 58.6% vs 49.7%.

Ohio township is southwest of Haysville, which has annexed about a half a square mile of the township. The township has 1980 persons, including 548 in Haysville, and 1432 in unincorporated areas. The area not in Haysville is in two precincts: OH03 is a strip 1/4-mile long by 50 feet wide which has been isolated by Haysville, and OH01 is the remaining 35-1/2 square miles of the township outside Haysville. Park township is northwest of Wichita, which has annexed about 1/5 of southern part of the township. The township population is 4613, including 3273 in the 3rd Class city of Maize, 172 in Valley Center, and 1170 in unincorporated areas. The original Park City was in Park township. The modern incarnation of Park City is in Kechi Township to the east. Park Township is fragmented into 12 precincts, 5 with votes tallied. The largest, PA05, includes much of Maize, and is the second-largest township precinct in the county. There are two small precincts on the chart, PA01 and PA02.

Payne township is immediately northeast of Wichita, and Wichita and Bel Aire have annexed about one-fourth of the township. The population of the township is 937, 90 within the 3rd Class city of Kechi, and 847 in unincorporated areas, most of which is formed into precinct PY01. Riverside township is immediately south of Wichita, and Wichita has annexed about 40% of the original township. The population is 13615, but that includes 4986 in Haysville and 3784 in Derby, both 2nd Class cities. of the 4786 persons in unincorporated areas, 3276 live in Oaklawn-Sunview CDP (census designated place). Oaklawn consists of small houses built during the Korean War to house workers at aircraft manufacturing plants and McConnell Air Force Base. They were intended to be rentals, and at one time, were limited to workers in the defense industry or air force. After major cutbacks in product, about 50% of the houses were vacant. In 1964, a California company purchased 700 for about $1500 each. They are not victorian cottages, but small 1950s-era mass produced housing. The area remains unincorporated, because Wichita and Derby don't want to annex it. A quite interesting fact is that 10.8% of Oaklawn-Sunview CDP residents are Laotian. On the chart, two precincts, RI03 and RI07 correspond to Oaklawn-Sunview. They are barely 20% Republican and had extremely low turnout. Despite the low Republican registration, Roberts did farely well. In RI03, which had 21% GOP registration, he received a 48%-44%-8% plurality.

Rockford township is in the southeastern corner of the county. The township extends west to the Arkansas River, rather than the range line for the survey township. The township has a population of 22,784, but 16,200 is in Derby and 5119 in Mulvane. Excluding these 2nd Class cities, there is 1465 persons in unincorporated areas mostly to the east. There are two precincts with votes tallied. Salem township is south of Wichita, west of the Arkansas River from Rockford Township. The population is 9486, including 5287 in Haysville, and 2 in Mulvane. Wichita has nibbled a bit from the northern edge, and Mulvane has outlined a claim in the south. Derby could annex across the river. But for now, there are 4197 persons in unincorporated areas, forming two election precincts. The two precincts are among the least Republican precincts in the townships, but they are about the average for the county, and are similar to Haysville which they are adjacent to.

Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #23 on: September 07, 2015, 06:13:50 PM »



This is a false color image of precinct size. I wanted to make sure I colored everything, and was experimenting with including images (maps) within a Google sheet:

Sedgwick Registration

See sheet "precinct size".

In general, the precincts with the most voters are on the periphery of Wichita in precincts that are larger than a section (square mile). In a given area, precincts that consist of a whole section usually have more voters than nearby precincts that are half sections. While the full section precincts might be less dense, they are not half as dense. The precincts near the downtown area of Wichita are fairly small.

There be a changing sensibility of how large a precinct should be. Wichita now uses voting centers shared by multiple election precincts. If voters will be driving to a voting center, the only reason for dividing precincts would be to reflect various district boundaries. In areas developed earlier, each precinct would have its own polling place, and the size of precincts may reflect the spacing of schools and churches. Once a precinct has been divided, there may be resistance to merging it back into a larger precinct. Voters are used to voting at a certain location, and if districts are built from election precincts, a district line may be drawn along the edge of the precinct. Kansas requires most election precincts to be frozen before and after the census to permit tabulation of population for election precincts so that they may be used for redistricting.
Logged
jimrtex
Atlas Icon
*****
Posts: 11,828
Marshall Islands


Show only this user's posts in this thread
« Reply #24 on: September 12, 2015, 09:05:11 PM »

This map shows GOP registration share by precinct.



The highest Republican shares are to the east and west of the city. In Wichita city council district 5, there is a a quite clear gradient, with the GOP% dropping about 5% for each section further east. There may also be a weaker north to south trend across the county.
Logged
Pages: [1]  
« previous next »
Jump to:  


Login with username, password and session length

Terms of Service - DMCA Agent and Policy - Privacy Policy and Cookies

Powered by SMF 1.1.21 | SMF © 2015, Simple Machines

Page created in 0.103 seconds with 11 queries.