Visualizing Presidential Election Results (1992-2000)

using Spotfire

by Evren Sirin

 

 

 

 

Since there is no way to map the votes to actual people, all the analysis done depends on average values but it still should give an idea about the interesting features of the data.

 

 

Figure 1 Distribution of 2000 election results by states. Finding the states where the winner has changed is looking for red squares or blue trinages

 

This is a good example that shows the benefits of Spotfire, ability to encode many information in one simple easy-to-understand format. The blue color indicates that the winner is Democrat. We can easily see that Democrat party won the election in West Coast and in North East states. We can easily spot the outliers like New Hampshire where Republic party won the election, a different result than all its neighbors. The second thing we can see is how the winners changed in a state. The shape of a point tells the winner of the previous election in that state, squares for Democrats and triangles for Republicans. If we want to see in which state did Republicans win 2000 election but lost the 1996 is simply to spot red squares. We see that there are many states such as Nevada, Arizona, Ohio, etc. It is interesting to see that there is no blue triangle which means Democrat lost in all the states they lost in previous election. The last thing to note in the graph is the size of each shape which tells you the electoral votes for that state. Although the number of blue shapes is small their sizes are not, which implies that the overall result is very close.

 

The next figure plots the Democratic Party votes and the income levels of the states. The size of a shape increases as the income increase and the color approaches to blue as the percentage of votes for Democrats in the state gets higher. Looking at the figure we can easily spot the states with low income which are generally located in the Central and Middle South regions. The Democrat party votes in these seem low but we cannot make a conclusion since there are states like Utah or Alaska where Republicans win with high difference and the income level is above average.

 

Figure 2 The relationship between vote percentages and income level. It is hard to say that there is a correlation between two.

 

 

Figure 3 shows the relationships between income level, voting percentage and election results in the states located in north east part of United States.

 

Figure 3 Relationship between income level, voting percentage and election results

 

In this figure, sizes show the income level of the state, the voting percentage is represented with color starting from lowest as red to highest as blue. The triangle shapes say that winner in that state is Republican. It is seen that the triangles are smaller than squares and more to close red than the squares. This gives a hint about a behavior such as if the income level is high people tend to vote more and they tend to choose Democrat party. However, the differences between the states is not significantly high and also it may be the case that the income level in a state is high but people who vote in that state are from the lower income level which invalidates any conclusion based on this data. One problem here is trying to show the correlations using colors and sizes. Instead it will be better to change the axis as in Figure 4.

 

Figure 4 Correlation between income level and percentage of college graduates and their implication on election results

 

This figure clearly shows that the percentage of college graduates in a state increase as the average income level in that state increases. The colors show the winner party in the election and the points used are over the results of last three elections. Looking at the overall view, we see that there is not a strong relationship between the choices in the election and the income level or education level.

 

Figure 5 How the high education states are spread across the country

 

Figure 5 filters out the states where the percentage of college graduates are low. As we have seen in the previous visualization, we once again realize that there is not a correlation between education level and the election results. We also see that states with high education level are not strictly concentrated on one region of the country.

 

Figure 6 Visualizing different parts at the same time a) Upper left chart shows the change in percentage of voters through the years b) The upper right
graph displays the distribution of votes c) Lower chart shows the change in income level and education level in different regions

 

The last figure is a good demonstration for the ease of visualizing different parts of the data in the same screen. We see that voting percentage dropped from 70%’s to 60%’s in the last two elections. The lower chart gives information about the change in income level and college graduates for different regions of the country. As time increases, the colors become bluer, which indicates income levels increase by time in roughly all regions. We also see that west and northeast regions always have a higher average income level than the other parts but the difference between states is diminishing. There is also an increase in the sizes of the shapes which implies the increase in the college graduates as time increases.

 


Some of the queries were also applied with TreeMap to see if it is easier to extract the information. A similar query to the last example is to visualize the change in income level and percentage of college graduates as in Figure 7.

 

Figure 7 Investigating change in income level and percentage of college graduates using TreeMap

 

There are three columns to show the years and in each year we see the seven different regions. The color of each state shows the income level and gets a brighter value as the level increases. We can clearly see that the values get a brighter value as the time increases. It is also possible see west coast and north‑east states have a brighter value than others. However, there is a small problem with the visualization of college graduate percentages. As in Spotfire example, size of a state shows the education level. To compare the overall average of northeast states for different years we should look at the total size occupied by the region at that year. Suppose we are just considering northeast region for years 1996 and 2000. We first think the size is bigger for 1996 since the height is bigger but the widths for two years are different which makes the total area in year 2000 slightly bigger (by clicking at the regions we see that year 1996 occupies 300 units whereas year 2000 occupies 315 units). We reach the same conclusion as before with a little more effort but not mainly because of the application but because the values in the data are close. One may also argue that the confusion here is caused by slice and dice method but it turns out this method is much more better than stripping or squarifying for this kind of analysis.

 

Another query done in TreeMap is to find the states where the winner of the election changed between years.

 

Figure 8 Finding states where winners changed requires a new kind of hierarchy in the data

 

If we change the hierarchy in the data and define the top most level as the Winner field and the next level as the LastWinner field finding the states where the winner changed becomes a trivial operation. If we look at the year 1992, we see that many states who voted Republicans in 1988 chose Democrats this time. In 1996, two more Republican states turned Democrat but three states changed to Republicans. As we have seen in the first query of Spotfire there is no state where Democrats won the 2000 election but lost 1996 election. This query turns out to be quite easy in TreeMap as in Spotfire but TreeMap also has an advantage to show all years at once which was not possible with Spotfire.

 

·        Critique

 

The Spotfire is a powerful visualization tool with some specific advantages

 

§         Very easy to use and understand

§         Encode many information at the same time using color, size, shape and position

§         Various different visualization tools to apply different techniques

§         Extracts the properties of data automatically (e.g. categories) without any need for manual configuration

 

However, there are some disadvantages of Spotfire which we can list as follows

 

§         Very hard to manipulate maps (need to supply coordinates for each state)

§         No way of grouping data (e.g. showing average value over different years would be very helpful)

§         Labels are ill-positioned when data points are near to each other