Spending Under Congresses and Presidents, Part II

In my last post on Spending under Congresses and Presidents, I mentioned that the data on budget spending would be more revealing if done in real (i.e. inflation-adjusted) terms. Below I have done just that, looking at the total increase in spending per 2-year Congress in constant 2005 dollars. Each years’ spending has been adjusted using the GDP deflator (GDPDEF) series available from the St. Louis Federal Reserve’s FRED database.

As before, the lines represent total spending under each president, indexed to 100 at the spending level in the final of the preceding administration. Bars represent percent increases in spending from one Congress to the next. Red bars represent Republican control of both houses of Congress, Blue bars represent Democratic control, and Violet bars represent mixed control, which, in practice, has usually meant Republican control of the Senate and Democratic control of the House (although in 2000, Senator Jeffords left the Republican party, declared himself an independent, and caucused with the Democrats, meaning one bar in the Bush years should be violet too).

For comparison, I show both the original and the inflation adjusted chart:

The original (nominal spending increases) first:


And now inflation adjusted budget increases:


When budgets are shown in inflation-adjusted terms, there really does not seem to be clear differences between Republicans and Democrats, either in Congress or the White House. The 1960s appear to be the times of massive federal budget expansion, starting in the late 1950s and tapering off in the early 1970s. The early Reagan years controlled expansion the best, and the Clinton administration was also relatively modest under both Democrat and Republican Congresses.

The statistical geek in me couldn’t resist seeing if there was any statistical evidence for the claims that Democrats or Republicans keep federal spending under control in any office, given this data. Coding variables for House, Senate and President as 0=Republican and 1=Democrat, and regressing Real_%_increase on these variables, one would expect – if the Republicans as budget cutters hypothesis holds – positive coefficients for House, Senate and President, indicating higher real budget increases under Democrats.

Basically, none of the statistical tests I ran offer any statistically significant support suggesting that Republcans or Democrats are any different from each other in terms of controlling federal spending. All F tests had p-values suggesting that none of the variables in the model explained any statistically significant differences in the rate of budget increases. There may be differences in budget deficit numbers under one group or the other, but that would require different data to test.

Because percentage increases jumped around substantially in the early part of this data series (causing substantial heteroskedasticity), I ran the main regression analysis starting in 1965 (fist year of Johnson’s elected term). I also ran the regression on the full data set, correcting for heteroskedasticity using weighted least squares, and came up with substantively similar results. The F-statistic of 1.64 (p-value = 0.20) suggests that which party controls of the House, Senate, or Presidency has no statistically significant effect on the real rate of federal spending increases over the long term.

Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.143 1.093 2.875 0.00633 **
Pres -0.971 1.020 -0.952 0.34677
House -2.111 1.425 -1.482 0.14589
Senate 2.738 1.302 2.103 0.04153 *

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3.318 on 42 degrees of freedom
Multiple R-squared: 0.1046,        Adjusted R-squared: 0.04061
F-statistic: 1.635 on 3 and 42 DF, p-value: 0.1957

However, the T-statistic on the Senate is barely significant at the 0.05 level. In general, if a t-statistic is significant, the F-statistic should be significant too, but the two tests can occasionally lead to different if the t-statistic is on the border of statistical significance and there is multicollinearity (House and Senate scores are moderately correlated here).

If we relax our statistical standards from “beyond a reasonable doubt” and simply look at what the regression suggests as “the preponderance of evidence,” this suggests that Republican control of all three bodies has resulted in an average real increase in spending of 3.1%. Democratic control of the Senate tends to add about 2.7% to federal spending. However, Democratic control of the House would arguably take away 2.1% so that democratic control of the entire Congress only adds about 0.6% to federal nondefense spending. Technically, this number is not statistically significant (p=0.15), but it is arguably “close,” particularly for a small-n, somewhat collinear model. Less statistically defensible but still more likely than not, Democratic presidents have overseen budgets that grow by about 1.0% less than their Republican counterparts.

The main conclusion should be that there is no statistical difference between the two parties in terms of who keeps federal spending under control, which is substantively significant, given that Republicans are claiming budget-cutting expertise in this election. But to the extent that one does consider the evidence, nondefense federal budgets have tended to grow slower under Democratic Presidents and Houses, though this is partly offset by the fact that the largest budget increases have taken place under Democratically controlled Senates. Indeed, the most likely reason that the Senate score is significant is that it is skewed positively to balance out the “budget cutting” influence that Democrats have when they control the House.

1 Comment

  1. Spending Under Congresses and Presidents Said,

    October 27, 2010 @ 23:46

    [...] (Added 27 Oct 2010: Part II of this post takes a look at the same data in inflation-adjusted terms) [...]