Admistrative:

Please indicate

Question 1:

Choropleth Map of US Voter Behavior in 2000

Download the results of the 2000 election from the School of Public Affairs at American University in DC and create a map involving only the lower 48 states that show voter behavior at a county level. To keep things simple let’s only consider George W. Bush, Al Gore, and Ralph Nader. Your write-up must include:

  1. A choropleth map where
    • Counties are filled with red when they tend to favor Bush
    • Counties are filled with white when they tend to be split
    • Counties are filled with blue when they tend to favor Gore and Nader

The above choropleth map of the United States shows the proportion of votes in favor of Gore or Nader in a given county. The larger the percentage, the more blue the color, whereas the lower the percentage the more red the county becomes. White indicates a 50:50 split or values close to it. Overall, we see that the rooky mountains region is considerably dominated by votes to Bush, whereas the extremes of the westcoast and most of New England favored Al Gore or Ralph Nader. The Midwest appears to be split, so does Southern California and Arizona.

  1. An answer to the following question: which states exhibit the greatest within state heterogeneity in voting? Come up with a mathematical justification.

To measure heterogeneity in voting within states, we borrow a concept engineered by economists and adapted by political scientists: The Ethno-Linguistic Fractionalization. Long story short, how ethnically diverse (i.e. heterogenous) a country is matters (civil wars, difficulties in reaching consensus …etc). What the ELF index does is quantify the likelihood that two people chosen at random will have different ethnic backgrounds. How does it do that? Through Herfindahl’s concentration index.

For our purposes, we want to measure the likelihood that two people chosen at random will vote for a different presidential candidate in a given state. The higher this index, the more heterogenous the state is likely to be. Let’s call our index PADI: Political Affiliation Diversity Index.

It will be calculated as: 1-\(sum(S^2_i, i=1, n)\), where \(S_i\) is the proportion of voters who belong to group \(i\), which takes the values of Bush, Gore, and Nader in our simplified calculation. Notice that in political science, there is a tendency to exponentiate the index to \(-1\), thereby creating an index that takes on values \(1, ..., n\). However, I find that this complicates the interpretation unnecessarily. It is worth mentioning that squaring the terms allows us to end with a range that extends from \(1/N\) to \(1\), which makes the expected range intuitive.

STATE PADI
vermont 0.5578957
maine 0.5513469
minnesota 0.5482274
oregon 0.5472464
colorado 0.5454856
washington 0.5376561
newhampshire 0.5370693
hawaii 0.5369167
alaska 0.5365009
wisconsin 0.5345361

According to our measure, Vermont is the the state with most diversity in political voting in the 2000 presidential elections. While this may come as a surprise, it is consistent with the facts. Bush and Gore’s results were very similar (120,000 vs. 150,000) and Nader had substantially support (20,000) given the State’s overall voters who were 290,000. Since the split is more equal compared to other states, the probability of selecting two individuals at random with different opinions (i.e. the value of our index) increases. This becomes clearer when contrasted with the least politically heterogenous state: Wyoming. The breakdown in this state between the candidates was approximately 148,000 for Bush, 4,600 for Nader, and 60,000 for Gore. Thus, the likelihood of selecting two people at random with different presidential candidate preferences is very low, since the results are dominated by votes for Bush. That is why Wyoming scores very low on PADI.

Having Hawaii on the top ten heterogenous states highlights a draw back of this tool: It does not weigh the total population of the state in question. This would not be a problem, except that we only have information on 80 voters from Hawaii. Whether this information is useful in determining the political heterogeneity of Hawaii or not is left for the reader to decide. While the PADI is far from perfect, its advantages are two fold: First, it is relatively simple to compute from both micro and macro data. Second, it has a rather intuitive interpretation. Another way of displaying the political heterogenity of states is on a map:

Question 2:

In this question, you must make an interactive “Single File” Shiny app that uses Leaflet. For all 184 census tracts in VT in the 2010 census, present information on the proportion of the population that is either

Use Social Explorer to get census data. I did a demo of this in class. If you don’t remember how or are stuck, please speak to me or get help from your peers. Do not submit copies of the same file.

There should be some mechanism in your Shiny app that allows one the user to toggle between the different ethnic groups.

Loading Shapefile Data

Here is some starter code:

Write-Up

Upload your shiny app to the Middlebury Shiny Server (see Lecture 16) and post the url to the app here, replacing the nhl.com link with the link to your app.

Comment on general ethnic demographic trends that’s more substantive than just “Vermont is really white.”

The first Vermont map shows the proportion of white inidviduals by US Census tract. While it is tempting to think of the beige as tracts with minority white, a cursory glance at the lengend shows that these tracts have upwards of 70% of their inhabitants as white.

The second map shows the proportion of blacks by census tract. Visually, the pattern seem to be switched. That is, tracts that contain lower proportion of whites seem to contain higher proportion of blacks. This pattern is repeated by all other minorities, suggesting that minorities tend to live in the vacinity of one another, as opposed to desperse between different neighbourhoods.