Some time in early 2004, I ran across a Web site maintained by the Department of the Treasury, listing the U.S. National Debt year by year since 1791. The numbers by themselves are too big to be meaningful, so I put them into a spreadsheet to see if I could extract any interesting trends and patterns.
In particular, I wanted to be able to compare the effects of government policies under different Presidents and Congresses on budget deficits. This sort of comparison will inevitably have political implications, so I resolved to make the page as transparent and fact-based as possible: if you don't like my conclusions, here are the data and you can replicate them yourself. I was concerned about (conscious or unconscious) political bias, so I wanted to avoid any sort of sophisticated analysis that required lots of judgment calls. But some judgment calls are unavoidable.
One issue is to which President or Congress a particular year's deficit should be attributed. U.S. Congressional terms start on Jan. 1, Presidential terms later in January, and the fiscal year on Oct. 1, so none of them exactly match up. Since most Presidential and Congressional actions take months to be implemented, much less to show a measurable effect on the debt, I decided that the fiscal year spanning a change of (Presidential and/or Congressional) terms would be counted as belonging to the outgoing President and Congress.
Another question was how to define the word “deficit”. In trying to keep things simple and transparent, I chose “one year's national debt minus the previous year's national debt”. But even that simple definition doesn't entirely resolve things.
The Treasury reports three numbers: "Public Debt", "Intragovernmental Holdings", and "Total Debt", which is the sum of the other two. The difficult question is whether to use "Public Debt" or "Total Debt" in the data analysis. The question is made easier by the fact that "Public Debt" and "Intragovernmental Holdings" weren't reported until 1997, so if I want to cover history farther back than 1997, I have no choice :-)
Intragovernmental debt, as I understand it, is mostly Treasury bonds held by the Social Security and Medicare trust funds, which are part of the Federal government in a sense, but there are legal walls between them and the Treasury. When you pay $1 of Social Security tax, your dollar goes to the Social Security administration, which promptly transfers it to the Treasury in exchange for bonds. The whole transaction has increased intragovernmental debt by $1, while giving the Treasury $1 and thus reducing the public debt by $1 from what it would otherwise be. When you draw a Social Security check, the process is reversed: the SSA sells bonds back to the Treasury in exchange for dollars, thus reducing both the intragovernmental debt and the Treasury's liquid assets by the same amount. In short, Social Security and Medicare taxes and payments affect the "public debt" and "intragovernmental debt" but have no effect on the "total debt". So if I want to ignore the effects of Social Security and Medicare, "total debt" is the right figure to use.
I wanted to adjust for inflation so one could meaningfully compare
deficits from more than a few years apart. Originally, I used the
(debt in year X) - (debt in year X-1) ------------------------------------- (CPI in year X)
Then an economics-professor friend of mine pointed out a problem with this. Suppose the national debt is a trillion dollars, and the inflation rate is 10%. If the government spends exactly as much as it takes in this year, the national debt will still be a trillion dollars, but those dollars will be 10% less valuable a year from now; the nation is less deeply in debt than before, because it's easier to pay off the debt with inflated dollars.
To look at it another way, part of any year's government expenditures is interest on the national debt, paid at a rate that must normally be higher than inflation (or nobody would buy the bonds). So suppose we want to compare two governments: each inherits a $1 trillion national debt, and each spends exactly as much as it takes in, aside from interest payments, but one of them has 0% inflation and the other 10% inflation. All else being equal, the latter will have to pay 10% higher interest rates to bond-holders. So of these two governments, the latter shows up as having a $100 billion larger deficit than the former, even though both are in some sense equally fiscally responsible and this $100 billion increase in nominal debt is cancelled by inflation.
So as of Oct. 2013, I'm calculating deficit as
(debt in year X) (debt in year X-1) ---------------- - ------------------ (CPI in year X) (CPI in year X-1)Note that if there was no inflation in year X, this produces the same answer as the previous formula; in the presence of inflation, it produces a smaller number, and when there's deflation, it produces a larger number. The old version of the page, before this revision, is here.
Tabulated the national debt by year, back to 1911. (I originally went only back to my birth, in 1964, but then expanded the chart to 1911, stopping there because I ran out of CPI data.)
Adjusted these debts by the Consumer Price Index as of the end of the fiscal year to get a debt in constant (1983) dollars. (There's nothing magical about 1983, except that CPI figures are often expressed in 1983 dollars. If I used, say, 2006 dollars, all the numbers would be about half as large, but the shape of the curve would be exactly the same.)
Subtracted each year's constant-dollar debt from the next year's, as a measure of one-year federal deficit (including interest paid).
In 1977, the reporting date shifted by 3 months, so that "year" was actually 15 months long; I multiplied the deficit in this year by 4/5 to give an annualized figure. I did something similar for 2012-2013, 2013-2014, 2014-2015, and 2015-2016, which were distorted by debt ceiling standoffs; see below.
Annotated each year by the party of the President and the majority party in each house of Congress (see House history and Senate history). Many years show a transition from one party to another.
Also annotated each year with the top-bracket individual marginal Federal income tax rate for returns filed that year (i.e. on the previous year's income), drawn from The Tax Foundation. This is an oversimplification, of course, since it doesn't say where that top tax bracket starts: for example, in 1992 the top tax bracket was 31% on income above $86,500, and in 1993 it was 39.6% on income above $250,000, but somebody who earned $125,000 in each of those two years would have seen almost no change. More dramatically, in 1941 it was 81% on income above $5 million, and in 1942 it was 88% on income above $200,000; somebody earning $250,000 in both years would have seen a marginal tax rate rise not from 81% to 88% but rather from 71% to 88%.
Note that it's hard to ascribe cause and effect here: for example, when a war starts, tax rates and deficits usually both go up, but it would be hard to claim that one is because of the other; in fact, both are because of the cost of the war.
I haven't done much analysis of the Congressional data yet: I'd like to see whether party control of the Senate makes a consistent difference, whether party control of the House makes a consistent difference, whether the margin of control (e.g. 51% as opposed to 69%) makes a consistent difference, whether having the House and Senate controlled by the same party makes a consistent difference, whether having one or both houses of Congress controlled by the same party as the President makes a consistent difference, etc. But the data are there: perhaps somebody else will do that analysis for me :-)