Tweet this
Share on Facebook
Post on Google+

Scaling Counties in a Checkerboard State

State delegate equivalents and the 2004 Iowa caucuses.
Iowa county map
Click to enlarge

A caucus is a closed meeting of equals, the word probably derived from the Algonquin cau-cau-asu, translated as “proponent” or “advisor.”

At 6:30 pm on January 19th, Iowa’s 1,993 precinct caucuses were called to order, beginning the first step of an elaborate six-month process:

  • In January, precinct caucuses choose precinct delegates to the county caucuses.
  • In March, county caucuses elect 3,000 state delegates to the state convention.
  • In June, the state delegates choose 16 delegates to the Democratic National Convention.
  • In July, Iowa’s 16 delegates vote for a presidential nominee.

How many precinct delegates?

To make things confusing, the Iowa Democratic Party does not release the number of precinct delegates chosen on the 19th. Instead, the party releases a tally of state delegate equivalents, an estimate of the results of the county caucuses that will be held in March.

Although the count of state delegate equivalents is at least two steps removed from the actual votes cast by caucus attendees, the count is about as close as we can get to actual voting data, and deserves some study.

A checkerboard state

Iowa caucus county results
County results
(Redrawn from a map appearing on the Des Moines Register’s web site)

Iowa caucus county results for John Kerry
County results for John Kerry
(Percentage won)

Iowa is an interesting state because its 99 counties are all roughly the same size and shape.

Published maps of the Iowa caucus results tend to use a county map marked with different colors for each candidate. For example, the map at right shows the first place finisher in each county, with split fields of color indicating two- and three-way ties. The overall visual impression is of a fairly even statewide split between Kerry (light gray) and Edwards (light red), with Dean (dark red) winning only a small number of counties.

County maps can also be drawn for individual candidates. At lower right, a map of county-by-county results for John Kerry shows the percentage of support he has received from each county. Kerry appears to enjoy broad support across the middle of the state, with strong support along the west and east borders, and weak support in the south.

Although both of these maps display accurate county data, and each county has roughly the same visual weight, there is a catch: the overall impression of each map is accurate only if each county elects an equal number of state delegates.

Scaling counties

Guthrie and Polk counties, Iowa
Guthrie and Polk counties

In other words, with 3,000 state delegate equivalents and 99 counties, displaying state results on a county map of Iowa only works well if every county chooses roughly 30 delegates, or 1% of the state total.

In reality, the number of state delegates elected by each county varies widely. Polk county elects 430 delegates (a full 14% of the state total), while Guthrie county elects only 12 delegates (0.4% of the state total). The two counties have similar surface areas, but a 35-fold difference in their share of state delegates.

The reason for the difference is that state delegates are allocated according to population. Iowa’s estimated population in 2001 was just over 2.9 million, and with a population of only 11,323, Guthrie county does indeed represent only 0.4% of the state. Polk county contains the state capital and has a population of 379,029, or 13% of the total state population.

Iowa scaled counties
Guthrie and Polk counties scaled in proportion to their number of state delegates

Scaling each county relative to the number of delegates it elects produces a more accurate visual weighting (at right), though Polk county has broken its boundaries relative to the rest of the state, and Guthrie county has shrunk to a fraction of its former size. Although my previous attempts at mapping votes by county in the 2003 California Statewide Special Election scaled counties while maintaining most internal borders, Iowa’s gridlike pattern of counties makes it difficult to maintain internal borders without massive distortion. Another method of scaling is needed.

A framework of counties

One technique to reduce distortion is to modify the presentation of the county borders. Plotting the midpoint of each county and then drawing connections through shared county borders creates a simple network diagram of counties:

Iowa county framework Iowa county framework Iowa county framework
1. Plot county midpoints 2. Connect counties along shared borders 3. Remove county borders
Iowa county framework Iowa county framework
4. Scale counties 5. Remove midpoints

Now when a county scales it has a flexible framework in which to move. Although county size is distorted, the relationships between adjacent counties are preserved.

A revised county map

Iowa caucus scaled county map
Counties scaled in proportion to their number of state delegates

Scaling all 99 counties in a similar way produces the revised county map of Iowa at right, shown at a scale of roughly 4 green pixels per state delegate (the enlarged map is more accurate, at roughly 25 pixels per delegate). For easier comparison, all counties are drawn as squares.

Although this map is not intended as a display of population, state delegates are distributed in proportion to population, and there is a correlation of about 974 people for every state delegate, or roughly 250 people for every green pixel (about 40 people per pixel in the enlarged map).

One advantage of this presentation is that the relative voting power of each county is made immediately apparent: the bulk of the state delegates are elected by the more populous central and eastern counties.

Iowa county framework Iowa county framework distortion due to scaling
No scaling, no distortion Distortion due to county scaling

Another interesting feature of this type of map is that any underlying graphical distortion of the state is made visible by comparing the before and after framework maps.

Scaled county results

Iowa caucus county results Iowa caucus scaled county results
County results Scaled county results

Redrawn maps for county-by-county results and for John Kerry’s individual results are shown at right.

Iowa caucus county results for John Kerry Iowa caucus scaled county results for John Kerry
County results for John Kerry Scaled county results for John Kerry

Although less colorful, the redrawn maps may present a more accurate overall impression of statewide trends.

Scaled individual results

Similarly, generating maps for individual candidates can reveal patterns of support not visible in traditional county-by-county maps.

Iowa caucus scaled county results for John Kerry Iowa caucus scaled county results for John Edwards Iowa caucus scaled county results for Howard Dean
John Kerry: 38% John Edwards: 32% Howard Dean: 18%
(1,128 delegates) (957 delegates) (540 delegates)
Iowa caucus scaled county results for Richard Gephardt Iowa caucus scaled county results for Dennis Kucinich Iowa caucus scaled county results for Wesley Clark
Richard Gephardt: 11% Dennis Kucinich: 1% Wesley Clark: 0.1%
(318 delegates) (39 delegates) (3 delegates)



  1. Des Moines Register
    Iowa Caucus Results (January 20, 2003)
  2. Iowa Democratic Party
    Iowa Delegate Selection and Affirmative Action Plan for the 2004 Democratic National Convention (August 2003)
  3. U.S. Census Bureau
    State and County QuickFacts for Iowa
    Census of Population and Housing (2000)

A note on the data

Official county data for the Iowa caucuses was surprisingly difficult to find, and data from published sources varied in quality. Below is a scorecard of potential sources:

Des Moines Register: B+

The Des Moines Register’s county-by-county data is comprehensive, and provided all the percentage data used in the above maps. The only drawback to their coverage is that they provide individual results for each candidate as a percentage of each county’s state delegates, rather than an actual count. CNN provides conflicting and apparently incorrect counts (see below), which limited this article to a discussion of individual percentages instead of individual delegate counts.


Published CNN county-by-county data is misleading, poorly explained, and unusable. Although their web site does report the correct state delegate equivalent totals for each candidate (which correctly add up to a state total of 3,000 delegates), CNN’s supporting results by county only add up to 2,728 delegates, undercounting the total number of delegates by 272. The CNN web site also misrepresents percentage support for each candidate by deriving percentages from the incorrect delegate count, instead of displaying actual percentage support.

Iowa Democratic Party: F

The party’s web site does not provide caucus results. Instead, a note directs visitors to “the most up to date information regarding the caucuses,” the party’s official result site, which was live for two days before exceeding its bandwidth limit and shutting down. At last viewing on January 22, the Caucus2004 site provided no 2004 caucus data at all, prompting one visitor to post the message “i need the iowa causus results for homework!! please please help me out someone!

County data was accurate as of January 22, with 98% of Iowa precincts reporting.
All maps and graphics were made with Adobe Illustrator and Adobe Photoshop.