CMSC636: Data Visualization Critique
Author: Don Miner
Class: CMSC636 with
Dr. Rheingans
Project Description
This document is available at:
http://maple.cs.umbc.edu/~don/projects/636critique
Critiques of U.S. Election Maps
In this document, I will be presenting two different visualizations of United States
presidential elections in 2000 (Bush v. Gore) and 2004 (Bush v. Kerry).
Also, I briefly discuss two more visualizations just to provide more variety.
Some I believe do not
convey important information and are meaningless when it comes to determining who actually
won the election. This is not necessarily the visualization's fault, as the United State's
electoral college process makes conveying important information about the election results
difficult. However, the media should take responsibility in considering the importance of the
way the electoral process works and adjust their visualizations accordingly.
For each visualization I review, I will address the following questions:
- How effective do you judge this visualization to be? Why?
- What are the most noticeable elements of the visualization? Are they the most important?
- Is there essential information missing?
- Are there elements which carry no information?
- Are colors used effectively?
- How could this visualization be improved? Offer specific suggestions.
Simple results by county

"Source: The Associated Press,
ESRI Inc., USA TODAY analysis by Paul Overberg."
(Retrieved on the web 10/7/08 at
http://www.lagrange.lib.in.us/teens/Usamap.JPG)
This image is a figure presented in USA Today shortly after the 2000 presidential election between
Bush (R) and Gore (D). It shows which counties Bush won as red and which counties
Gore won as blue. Graphs are appended to the figure that show how much square area
and how much population was won from counties.
This visualization is some effective in conveying the fact that Gore won the urban vote
while Bush won the rural vote, which is what the USA Today article was about. This is evident
from the Bush winning five times the square miles but only getting approximately
half the U.S. population's votes.
This point is also conveyed by the most noticeable element of this visualization:
the stark contrast between the blue and red areas on the map.
In this respect, color was used effectively.
However, this graphic has many shortcomings in conveying
the information it wants to as well as any other kind of information that is of importance.
There are no labels whatsoever on the map.
If someone is not familiar with the location of U.S. Urban areas, the only information someone
could infer from this map is Gore supporters like living near water (East and West coast,
along Mississippi river).
Simple pinpoints showing where urban centers are located (New York, Chicago, Denver, etc.)
would help immensely in conveying the point that Democratic voters are mostly urban.
The other problem I see with this map is it fails
to present any further insights other than the race was close
in terms of votes but not area. From the amount of red seen at first glance,
one might assume that Bush won.
However, Gore was very close to winning the election because he won states with more electoral
college votes. The map fails to address any of the electoral college weights which actually
matter in who gets elected or, at the very least, the population of the counties. Basically,
this map is conveying who won the election in terms of area. It is not conveying who won by popular
vote or who won by electoral college votes. Fixing this problem is difficult, however it is
possible. Perhaps the colors of red and blue could be darker for more densely populated
areas. Also, cartograms could prove to be useful to convey necessary information (as seen
in later visualization critiques).
In addition, this visualization suffers from what many electoral
visualizations suffer from. Another
artifact of the electoral college process is winning counties is meaningless. Visualizations
like this are all too common and really all they show is the geographical positions of
the political parties. What really matters in an election is who won the states. Showing data
by county can be misleading. For example, in Maryland, Gore only won seven counties. Meanwhile,
Bush won 17. If you are not familiar with the population density distribution
of Maryland you may guess
that Bush won Maryland. However, in reality,
Gore beat Bush 56.57% to 40.18%.
Therefore, perhaps coloring entire states one color, representing that all electoral votes went to
that candidate, would be more effective. This still suffers from some issues since you are still
viewing the map in terms of state area. At least it would properly represents the states' votes in
the electoral college, which is what determines the winner of the election.
The most misleading piece of data from this figure is the "Population of counties won" graph.
Although this provides a rough estimate of the popular vote,
this piece of data, much like most parts
of this visualization, is completely meaningless.
Winning a county means nothing, so attributing this county's entire population to that
candidate is a distortion of the actual popular vote (at the state level and national level).
Imagine that a candidate won every
single county in the U.S. 50.001% to 49.999%.
The "Population of counties won" would be the entire U.S. population for
one candidate and zero for the other candidate. The reasoning behind this particular graph
is even more broken than the electoral college process. Being the most atrocious part of this
visualization, it is ironically the easiest to fix: use the actual popular vote. Note that
Bush won more "population of counties won," but
Gore won the popular vote.
Overall, I would like to say again that this visualization is effective
(with some shortcomings) in conveying some
demographic information
about likely Republican and Democratic voters. However, looking anywhere beyond
that is going to result in confusion.
3D view of results by county

Source: ESRI staff in New York City and Redlands in collaboration with CBS News
(Retrieved on the web 10/7/08 at
http://www.esri.com/industries/elections/business/uspres_election2004.html)
This election map displays the results of the 2004 U.S. presidential election between
Bush (R) and Kerry (D). This visualization is very similar to the one critiqued in the
previous section but has a major advantage: the population of the counties are represented
as pillars coming out of the map. Therefore, higher pillars convey more population.
I believe this map to be much more effective than the previous map but it still has
similar problems. The first thing that stands out when viewing this image is the
large blue pillars coming out of the U.S.'s urban areas. This immediately shows the viewer
than although the Democrats lose in rural areas, they win in urban areas. An uninformed
viewer may understand why the popular vote was close since the blue counties typically have more
weight. A less qualitative advantage to this image over the other one is that it is pleasant
to look at. There are a lot of interesting correlations between population and votes
on the map. For example, Texas and Oklahoma have a urban centers that voted Republican.
I find it very interesting that urban areas frequently dominate the vote in coastal states,
making Democrats to win most of these states. Although this visualization has positive qualities,
it has one major misleading property and one minor problem.
Similar to the previous figure, county winners are awarded the entire "tower" through the
color label. For
example, the tower in the Chicago area is very high. However, this makes it appear that
the volume of that tower is representative to the number of people who voted Democrat. This
is not the case. If Kerry won 50.001% of the vote and Bush won 49.999% of the vote in Chicago,
this tower would be the same height and same color as if Kerry won 100% and Bush won 0%. I believe
that since Democrats are getting most of their votes from these towers, it makes it appear
Democrats are getting more votes than they really are. Therefore, this visualization is an
estimation, at best, of the popular vote for each state. Although this may make the graphic more
cluttered, the towers could be split up into individual pie-graphs to show a percentage. Another
way would be to split up each county into two towers: one for each candidate.
Although not as much a issue as the previous image, this figure also suffers from the fact
that it does not convey electoral college information. This one does a better job because
it shows the population of the states (which is the basis for the number of electoral votes)
and guessing which candidate a state voted for in the
electoral college is possible. Unfortunately, it is not entirely clear for some states
which candidate they voted for. For example, California has some very Democrat learning
urban areas but the
majority of the counties vote Republican. Also, Wisconsin appears to be split half and half.
Perhaps to fix this, this image could be showed along with a visualization that uses the
same technique except shows states as pillars of their electoral college vote. For example,
California would be a tall blue column (because they voted Democrat) and Texas would
be a tall, but not as tall, red column (because they voted Republican).
A possible minor problem this visualization has is that it
lacks labels. However since this graphic was shown on
television while accompanied by verbal commentary, labels would just clutter
the image. Therefore, I am not sure if this is a good thing or a bad thing. I would have to see
a comparison of both to decide.
In summary, I believe this visualization to be successful. It catches the eye with its
3D perspective. Although it has its shortcomings when it comes to the details of the election,
at first glance it provides a lot of useful information. I believe this visualization would
be made more useful if viewed side-by-side with an electoral college map.
Other election visualizations

Source:
http://www-personal.umich.edu/~mejn/election/
This image is a cartogram of results of the 2004 election
by electoral college value. States with more electoral college votes
appear to be bigger. Therefore, theoretically, if 51% of the area was red, Republicans
won the election. This visually answers the question "who won the election?"
Unfortunately, this map conveys very little other information and some people may not like
the distortion. Also, it would be interesting to see how one would convey states like Maine
and Nebraska in a cartogram that
sometimes split electoral votes.

Source:
http://www.princeton.edu/~rvdb/JAVA/election2004/
This figure shows who won each county based on percentage with a color gradient.
Therefore, purple areas are 50%/50% areas. In my opinion, this image better answers the question
the first critiqued visualization was trying to answer: "Where are the Republican parts of
country and where are the Democrat parts of the country?" However, if fails in many of the same
ways.