California Here I Come

[jimrtex]

I've decided to redistrict California based on a mechanical process that is intended to produce somewhat rectangular districts that as much as possible avoid crossing county boundaries.  Districts are based on census tracts.

The basic process is to divide the state in two parts of roughly equal area, apportioning districts to the two parts, and then to repeat the process.

The splitting into two parts is an iterative process.

1) the coordinates are rotated based on a least squares fit.  For example, in the initial split of the entire state, this rotation recogizes that San Francisco is more northwest than north of Los Angeles.

2) the area is divided into parts of equal area (in practice, I simply took the average of the extreme coordinates as the initial dividing line).

3) the population on either side of the dividing line is computed, and converted to the equivalent number of CDs.

4) the number of CDs on either side of the dividing line is rounded to a whole number, and a new dividing line determined such that the population on either side of the line corresponds to the whole number of CDs.

5) the dividing line between the two parts is converted to a dividing boundary that conforms to county boundaries.  When a county is split by a dividing line, the entire county is assigned to the side of the boundary where the majority of the people in the county live.

6) the population on either side of the boundary is computed, and converted to the equivalent number of CDs, and then rounded to a whole number of CDs.  If the whole number of CDs has changed from step 4, then step 4 is repeated.  This process is intended to shift the boundary such that most of more populous counties are on one side of the boundary by shifting one or more whole CDs.

7) reverting back to the dividing line of step 4, a boundary line is determined that splits only one county.  Initially, one or more counties are split by the dividing line into two parts. 

7a) The smallest county part is shifted to the other side of the dividing boundary.   

7b) The dividing line in the remaining split counties is adjusted to maintain the proper population on the other side of the boundary (corresponding to a whole number of CDs). 

Steps 7a and 7b are repeated until only one county is split.  This is the final boundary between the two parts.


The initial dividing line of the state was roughly through Fresno County on a dividing line running southwest to northeast.  The population to the southeast of the line was equivalent to 34.36 CDs and to the northwest 18.63 CDs. 

This was rounded to 34 and 19 CDs and the dividing line shifted to the southeast, so that a population equivalant to 34 CDs was southeast of the dividing line.  This dividing line split Monterey County in the extreme southeast corner, Fresno County (72% southeast of the line), and roughly the Mono-Inyo county boundary.

The small area in Monterey County was shifted to the northwestern part, and the dividing line in Fresno County shifted slightly northwestward to produce the final boundary splitting the state into two areas with 34 and 19 CDs.

The southeastern area of 34 CDs was then split.  Initially, Los Angeles and San Bernadino counties were split, producing a 24:10 CD configuration. 

But over 3 million people in Los Angeles county were to the northwest of the dividing line (and over 6 million to the southeast).  The split was changed to 29:5 so that most of Los Angelese County could be in the southeastern part.  The new dividing line now split Ventura, Los Angeles, and (extreme northwestern) San Bernadino counties. 

All of Ventura County was placed in the northwestern area that includes the coast northwest of Los Angeles as far as San Luis Obispo, as well as Bakersfield and most of Fresno County.  It also includes about 40,000 people in Los Angeles County.  This area will be divided into 5 CDs.

San Bernadino County and almost all of Los Angeles County were placed in the southeastern part along with Orange, Riverside, San Diego, and Imperial counties.  This area will have 29 CDs.

This area of 29 CDs was itself split.  The initial dividing line split San Bernadino, Riverside, and San Diego counties, with most of the former two to the northwest of the dividing line, and most of the latter to the southeast.  Placing all of San Diego County in the southeastern part results in a 24:5 CD split.  The final area in the southeastern part included San Diego and Imperial counties, and the southeastern part of Riverside County (about 240 thousand people).


The southeastern area consisting of San Diego, Imperial, and about 1/6 of Riverside was then subdivided into 5 CDs.  Around 85% of this area is in San Diego County.

The initial east-west division would have produced a 5:0 split due to the light population in the eastern portion of the area.  So instead an area equivalent to 1 CD was partitioned off, including Imperial County, the southeastern portion of Riverside County, and eastern San Diego County.  The population is roughly split 1:2:2 among the 3 counties, respectively.

The coastal strip of San Diego County with 4 CDs, was split in half on a north south basis, with both halves in turn split north and south.

The final split shows some deficiencies with the overall approach.  While the districts are compact, they tend to split communties.  For example, Coronado, El Cajon, and Escondido are all split by district boundaries.  Some of this could be avoided if cities are sub-county areas were used within counties, in a manner similar to how splitting of counties is minimized.

[minionofmidas]

You're moving to California? Change your username immediately.

Oh, you're not.

Interesting approach there; show me more.

[jimrtex]

The 24 CD area in the LA area was the next to be split.

Because this area included the portions of San Bernadino and Riverside counties that extend to Arizona and Nevada, there was relatively little population to the east of the initial dividing line.

The initial split has 204 thousand to the east of the dividing line.  The dividing line was changed so that 1 CD could be formed to the east of the line.  Most of the population (469 vs. 169 thousand) was in Riverside County rather than San Bernadino County, so the entire CD was formed out of the eastern and central Riverside County.

This left an area to the west with 23 CDs that still included eastern San Bernadino County.  This time the initial split only had 87 thousand to the east of the dividing line.  So once again the dividing line was changed so that 1 CD could be formed to the east of the line.  This time more of the population was in San Bernadino County (527 vs. 107 thousand) so the entire CD was formed out of eastern and central San Bernadino County.

This left an area to the west with 22 CDs, but with the extreme eastern areas of Riverside and San Bernadino County shed.  The initial dividing line was in Los Angeles County.  The initial split had the equivalent of 8.69 CDs to the east and 13.31 CDs to the west.  This was adjusted to produce a 9:13 split.

Almost 1.2 million people in Los Angeles County were to the southeast of the split, so the CD split was changed to 7:15 so that Los Angeles County would be mostly whole.  This split left about 289 thousand Los Angeles County residents to the southeast, and 360 thousand Orange County and 38 thousand San Bernadino County residents to the northwest of the dividing line.

All of San Bernadino County was placed in the southeastern area, and all of the split portion of Los Angeles was placed in the northeastern area.  Most of the Orange County portion was moved to the southeastern area, but 109 thousand remain in the northeastern area.

The two areas have the following composition:

Southeast (Eastern LA Metro), 7 CDs.

Orange 2.7 million
San Bernadino 1.1 million
Riverside 0.7 million.

Los Angeles County 15 CDs

Los Angeles 9.5 million
Orange  0.1 million.

[socaldem]

Your San Diego/Imperial/Riverside district, it seems to me, is probably a bit over the population requirements... in a map i did, Eastern Riversided, combined with Imperial is sufficient for a district.  Meanwhile, San diego currently has four self-contained cd plus Issa's district...

[minionofmidas]

Your San Diego/Imperial/Riverside district, it seems to me, is probably a bit over the population requirements... in a map i did, Eastern Riversided, combined with Imperial is sufficient for a district.  Meanwhile, San diego currently has four self-contained cd plus Issa's district...

You just went a little further west in Riverside...central Riverside is far more densely populated than eastern San Diego.

[socaldem]

Your San Diego/Imperial/Riverside district, it seems to me, is probably a bit over the population requirements... in a map i did, Eastern Riversided, combined with Imperial is sufficient for a district.  Meanwhile, San diego currently has four self-contained cd plus Issa's district...

You just went a little further west in Riverside...central Riverside is far more densely populated than eastern San Diego.

Now that I look at it, his North San Diego county district is entirely contained within the county so it all makes sense to me...

I'd just rather see Eastern san diego county not grouped with Riverside/Imperial in what would otherwise be a toss-up to a dem-leaning district...

[jimrtex]

Your San Diego/Imperial/Riverside district, it seems to me, is probably a bit over the population requirements... in a map i did, Eastern Riversided, combined with Imperial is sufficient for a district.  Meanwhile, San diego currently has four self-contained cd plus Issa's district...

The division line in Riverside County is stylized.  The districts were created with a spreadsheet and then mapped.  I haven't drawn the detailed map for Riverside County yet.

If eastern San Diego County were not included with the Imperial district, then part of western San Diego County would have to be included in a district that extended into Orange or Riverside Counties.

[jimrtex]

An area containing 7 CDs in Orange, Riverside, and San Bernadino Counties was the next to be split.  An area of Riverside County had been split off to form a CD with Imperial County and eastern San Diego County, and a CD in each of Riverside and San Bernadino counties had been split off before this area was split from Los Angeles County, so the area in the two northern counties is very much their western ends, which gives the region very much a north-south orientation.  There is also a sliver of western Orange County attached to the 15 CD area of Los Angeles County.

Since northwestern San Bernadino County is included, only 93 thousand people were north of the initial dividing line.  When a northern area with a population equivalent to 1 CD was determined, it included a majority of the San Bernadino County part, which was entitled to 1.66 CDs.  Rounding two 2 CDs, a new line was determined, which now included a majority of the Riverside County part.  Together, the San Bernadino and Riverside county parts were entitled to 2.71 CDs.  Rounding to 3 CD, a final 4:3 split was determined, with a southern area in Orange County entitled to 4 CDs, and a northern area consisting of parts of Orange (180 thousand), Riverside (666 thousand), and San Bernadino (1,068 thousand) counties entitle to 3 CD.


With two small areas along the western and northern borders detached, Orange County was entitled to 4 CDs.  An initial northwest/southeast split detached one CD in the southeast, including Irvine and areas further to the southeast.

The area with the remaining 3 CDs was roughly a square, that was divided along a north/south diagnonal, 2 CDs to the south and one to the north.  The southern area was then split east and west.  This produced three triangular districts: one centered on Santa Ana toward the center of the county; one centered on Huntington Beach and Garden Grove in the south west, and one centered on Anaheim and Fullerton in the northwest.



Riverside County was completed by splitting off the southern portion of a 3 CD area including parts of San Diego, Riverside, and San Bernadino counties.

This southern CD includes about 180 thousand in Orange County, and 360 thousand in Riverside County, including the southwestern part of the city of Riverside and extending south to the Lake Elsinore area.

The northern area with 2 CDs has about 1,070 thousand in San Bernadino County, and 210 thousand in Riverside County, including downtown Riverside and other areas along the San Bernadino County line.

The only whole CD in Riverside County is in the west central part of the county, including Moreno Valley, Temecula, Hemet, Banning, and Palm Springs.

The eastern portion of the county, roughly from Indio to the east is in a CD with Imperial County and eastern San Diego County.

[muon2]

I like the process so far. BTW, what is your source for maps of the census tracts.

[minionofmidas]

So San Bernardino County is next?

[jimrtex]

I like the process so far. BTW, what is your source for maps of the census tracts.

Census Bureau.  I used the reference maps under American Factfinder, and then screen capture to convert to .bmp files.  The advantage is that I can turn most everything off and get clean outline maps.  The disadvantage is that the maps are fairly small (about 600 pixels/22 miles across), and I have to piece the maps together.  The zoom factor is about 3:1, so the next scale up is a bit too large, though I did use it for the full county map of Riverside County.  It would be ideal if I could set a scale and get larger images.

After I created the maps, I could turn on the census tract numbers when I colored the maps.

The Census Bureau also has census tract maps in .pdf format organized by county on their FTP site, but they also include city and other boundaries, and aren't as useful for coloring.

I'm rather pleased with the maps, since they were created solely using coordinate information, and I didn't see what they looked like until I drew them in on the maps.

I have started over, trying an adjustment to the procedure.  When dividing an area, I first find a minimum bounding rectangle, choosing an orientation that gives a rectangle with the minimum area.   For example, for the entire state, the bounding rectangle is rotated 34 degrees from north-south.

The initial dividing line is based on having equal areas on either side, rather than dividing the bounding rectangle in half.