A2-DavidPoll
From CS294-10 Visualization Fa08
Contents |
Data Sets
For this assignment, I used two datasets from Swivel.com (and manually merged them for the relevant years). The first gave the composition of congress (Democrat, Republican, Independent) from the 1850s on (Congressional Party Division). The second gives a breakdown of the budget by category from 1976 onward (United States Budget).
Progression
Initial Question
How does the composition of congress (Democrat vs. Republican) impact the amount of Defense and Education spending in the federal budget?
Hypothesis
When Democrats control congress, they will spend more on Education and less on Defense (based on stereotypes). Vice versa for Republicans.
Initial Visualization
This initial graph shows a simple plot of defense and education spending in dollars (non-normalized) as well as a plot of the quantity of Democrats, Republicans, and Independents in congress. At the outset, it's not clear that there are any relationships at play here. For more relevant data, I'll derive the ratios of each of these values as part of the whole (# of Democrats / # of Congresspeople, Defense Spending / Total Spending, etc.)
Question Change
Does the proportion of Democrats or Republicans in congress correlate with the proportion of Defense and/or Education spending as a part of the total budget?
New Visualization
This new visualization shows that there may in fact be some correlation between the proportion of Democrats in congress and the proportion of the budget that defense spending contributes. Since the relationship between the percentage of education and defense spending is unimportant compared to their overall trends, using separate scales for each will make the trends clearer.
New Visualization
While this visualization does help make the trends in education spending clearer, it doesn't show much correlation. Perhaps a more pertinent question is the defense/education spending ratio as it compares to the democrat/republican congressional composition ratio.
New Question
How does the ratio of federal spending on Defense versus Education vary with the composition of the Congress (Democrats versus Republicans)?
New Visualization
I'm actually shocked by this graph, since I expected the results to be exactly the opposite. Here, I've plotted the ratio of Democrats/Republicans alongside the ratio of Education/Defense spending (using different scales). Surprisingly, there seems to be a very high inverse correlation -- as though having fewer Democrats is correlated with greater spending on Education (as it compares to defense/military).
New Visualization
In fact, inverting the ratio of Democrats/Republicans shows just how correlated these two values are. The question that then arises is... why? We've already seen that Democrats tend to budget more (proportionally) for defense than Republicans, so the correlation appears to be valid... the missing piece may simply be the President, though it's unclear how the President's party affects all of this. With such coarse data (D vs. R) for a single individual, Democrat/Republican don't fully describe the policies of that president.
Conclusion
My hypothesis was demonstrably false. While there exists some correlation between the party in power and education/defense spending, the results were nearly opposite my hypothesis. It can be seen most clearly in the following visualization:
It is noteworthy, however, that correlation here does not indicate causation. From the Republican/Education graph, it's clear that it took the election of a Republican president in 2001 to finally impact the Education proportion. Similarly for the Democrat/Defense graph, spending changes are offset from the election of Democrats by a few years.
Nonetheless, there seems to be a closer correlation between the ratio of Defense to Education spending and the ratio of Democrats to Republicans, as is shown in the following visualization:
Here, it's clear that in spite of presidential selection, for the last 30 years, the ratio of defense to education spending has mostly followed the ratio of Republicans to Democrats in congress, which still seems to contradict the stereotype used in my original hypothesis.
