A Few Graphs on Real GDP Growth Rates versus Taxes and the Size of Government
by Mike Kimel
A Few Graphs on Real GDP Growth Rates versus Taxes and the Size of Government
Cross posted at the Presimetrics blog.
This post has a few simple graphs showing the relationship between the growth in real GDP and a few other variables: the top marginal tax rate, spending by the federal government as a share of GDP, and non defense spending by the federal government as a share of GDP. I also plan to link the three graphs together into a bit of a story, a story which extends from my recent posts, and which I intend to build on over the next few weeks.
One note before we begin – all of the data in this post comes from the National Income and Product Accounts (NIPA) tables which go back to 1929, or the IRS (tax rates go back to 1913), so the graphs will all begin in 1929.
Figure 1 shows the % growth in real GDP from one year to the next on the primary vertical axis and the top marginal income tax rate for that year on the secondary vertical axis.
A couple things to notice… to the naked eyeball, there doesn’t seem to be all that much of a relationship between marginal rates and growth, despite the commonly accepted story line about how higher marginal rates lead to slower economic growth. Making matters worse for conventional wisdom, the correlation between the top marginal tax rate in any given year and the growth rate from that year to the next has been positive for every time period in the little table that is pasted in the graph. In other words, for all the years for which official data exists, higher tax rates have been accompanied by faster economic growth and lower tax rates were accompanied by slower economic growth. (Regular readers know this also applies at the state and local level.) Anyone telling you that lower tax rates lead to faster economic growth in the U.S. either should begin by explaining why this time it’s different or has nothing useful to say about the subject. (And I guess it is becoming a tradition, but let me ask – please don’t bother me with Romer and Romer. It doesn’t say what you think it says. It doesn’t even say what Romer and Romer think it says.)
Now, the American mythos circa 2011 also tells you that that the bigger the government, the slower the economic growth. (Note – since the tax rates I showed are federal tax rates, I am going to use the federal government here rather than “total” government. If I went with the entire government it wouldn’t change all that much of the verbiage below.) Figure 2 shows the % growth in real GDP from one year to the next on the primary vertical axis and the size of the federal government relative to GDP for that year on the secondary vertical axis.
Well, for most of the sample, it seems the correlation between growth rates and the size of the Federal government is positive. That is to say, the larger the federal government’s role in the economy in any given year, the faster the growth rate from that year to the next. However… the correlation turned negative in the tail end of the sample. (See the note at the bottom of the graph – the federal government shrunk under Bill Clinton (in fact, it shrunk every single year of his term, which means you can’t credit Newt Gingrich, at least not if you’re honest), and then rose in six of the eight years GW was in office.)
Figure 3 is similar, but instead of the Federal government’s entire piece of GDP, it shows non-defense Federal spending as a share of GDP:
Now, the non-defense piece of government spending is the stuff Republicans and libertarians really hate – welfare, fighting epidemics, infrastructure, etc. And it turns out that for much of the sample, in years when that bad stuff has made up a bigger part of the economy, the economy went on to grow more quickly from that year to the next.
Now… let me repeat a few facts (in a slightly different order than they came up):
1. The top marginal tax rate is positively correlated with real GDP growth over the coming year. The correlation dips a bit in the ‘70s and ‘80s, though it never goes negative.
2. Non-defense federal spending as a share of GDP is positively correlated with real GDP growth over the coming year until the 1980s.
3. The size of the federal government is positively correlated with real GDP growth over the coming year until the 1990s.
So what’s a story that fits these and other known facts? I think it’s one where the economy was growing rapidly for many of the years from 1933 to about 1973. That year, due to a combination of various forms of incompetence* malaise and fat sideburns started setting in. A story line started making the rounds: the tax rate is too high and government is too big. The tax rate is too high thing caught on with the public first – after all, nobody likes to pay taxes. The historic relationship between tax rates and growth began to fray, but it never quite died. Nevertheless, it led to a political groundswell and tax rates were, in fact, cut tremendously in the 1980s. Growth rates subsequently fell.
However, the 1980s brought with it something else, another political change in public perceptions. The government is not solution, or even a help, but rather a problem. (Truth be told, the cachet of the government had been falling for a while, but St. Ronald the Reagan and his acolytes really opened the floodgates.) The military excluded, of course. The result: in many fields, competent people who in earlier years would have gone to work for the government instead went into such value-adding enterprises as market manipulation and financial arson. The correlation between non-defense government spending as a share of GDP and real growth went negative. The problem with maligning the non-defense portion of the government is that ill-will doesn’t stay contained. Eventually, the correlation between even defense spending and the economy flipped.
So where are we now? Some of the very things that created the very rapid growth from 1947 to 1973, who’s kidding who, make that from 1933 to 1973, were taken apart, first in the public perception and second on the ground. And we’ve been growing more slowly, and at times, not at all, ever since.
—
As always, anyone who wants my spreadsheet is welcome to it. Drop me a line at my first name (mike) period my last name (there’s only one “m” in my last name) period gmail.com. Because I’m getting requests for a lot of different files, be clear about which post for which you want the spreadsheet.
Data sources:
Real GDP from NIPA Table 1.1.6.
Federal gov’t spending, federal non-defense gov’t spending, and nominal GDP from NIPA Table 1.1.5.
Top marginal tax rates from IRS SOI historical table 23.
Incidentally – I mentioned last week I intended to cover the fact that a number of the graphs I’ve been putting up are bimodal. That’s still on my radar screen. I just realized I want to build my narrative in a slightly different order.
* Incompetence by the Nixon administration (too much spending, an adherence to the gold standard, and price controls), incompetence by the Arthur Burns led-Fed (inflation, which only got exacerbated by Burns’ attempts to ensure Nixon’s re-election in 1972), incompetence by the USSR (failure to keep the Arab countries from realizing that regardless of their advantage in numbers and materiel, they were too ridiculously incompetent to expect success from attacking a second rate local power yet again), and incompetence by Henry Kissinger (reigning in the Israeli military when it was 60 miles from Cairo and about the same distance from Damascus… which eventually led to the Oil Embargo, not to mention the Middle East we see today).
Nice data, but your story is too cute. Looking at the graph with my human capacity for seeing patterns in noise, I think I see that what happens in the 1980s is not that the non-defense-spending/GDP-growth correlation breaks, but that non-defense-spending becomes more of a leading indicator, anticipating growth by a year or two. Since I’m going out on a limb even by seeing that pattern, I’m not going to go further out by proposing a cute story.
This is a very in depth study, but I believe it suffers from a major flaw. The author ignores the fact that relatively very few people were in the top income bracket between 1929 through the mid 1970s. Also, the many tax code changes during the this era altered the way income is now taxed: from ‘net’ to ‘gross’.
One can look at past income tax returns for a great example of the changes I mention. Franklin Roosevelt, in 1936, paid 16.5% in Federal Income taxes with a gross take home pay equivalent to $1.23 million 2010 dollars. In short, if we created a top bracket of 99% today, what good does it do if nobody pays it?
http://www.taxhistory.org/thp/presreturns.nsf/Returns/2F4DCA60170248AE85256E430078AF15/$file/F_Roosevelt_1936.pdf
“Now, the non-defense piece of government spending is the stuff Republicans and libertarians really hate – welfare, fighting epidemics, infrastructure, etc. And it turns out that for much of the sample, in years when that bad stuff has made up a bigger part of the economy, the economy went on to grow more quickly from that year to the next. “
Can you spell STIMULUS, boys and girls? 🙂
Re item 3: “ The size of the federal government is positively correlated with real GDP growth over the coming year until the 1990s.” The problem with this “fact” as well as any other “conclusions” based on comparisons with GDP is, to trade on Clinton’s retorts – it depends on what you mean by GDP. When the government collects a dollar of its citizen’s AGI, and redistributes it to an infrastructure contractor or welfare recipient, etc. who in turn buys a box of Cornflakes, that sale is counted towards the GDP. In other words, the GDP includes both the results of labor that directly produces a product that can be resold for profit and labor supported solely from tax revenue. Only when you strip out from the GDP the so-called “productivity” of labor supported solely from tax revenue, and then redo the graphs, can you arrive at more meaningful conclusions.
Before you can draw conclusions on a comparison of GDP growth and taxation, isn’t it necessary to elucidate the causes of GDP growth? You are making the same mistake everyone makes in presuming causation in the face of correlation. There are myriad factors which determine the growth of GDP, including population demographics, weather, political events such as wars, etc. Growth of GDP is dependent in part on labor costs, and labor productivity. If you have a cohort of working-age people who are young and well educated, you have a recipe for cheap, productive labor, and big GDP growth. If you have a labor force “X” and you add 30% to that (women entering the work force in the 1980s), you will have reduced labor costs due to increased supply.
It seems to me that these factors probably influence GDP much more than do rates of taxation.
So your correlation could conceivably be masking the opposite effect. In other words, in those years where GDP growth correlated with high marginal rates, isn’t it possible that GDP growth could have been higher _in that particular year_ if rates were lower?
I think you also completely discount timing, or “lag.” It takes some time before the effects of a change in taxation rates are felt. In any given year X, perhaps marginal rates are raised because it is politically safe to do so, because the economy is already humming, and everyone is happy. Suppression of growth is not felt until a few years later, when, of course, it is decided that rates should be lowered.
There are other flaws in your analysis. You correlate with the highest marginal rate, rather than looking at all taxation, etc.
Interesting data, I agree. But I don’t think you do a very good job of substantiating your opinions, beyond pointing out a single, narrowly focused narrative about a correlation. Sheesh….you don’t even calculate the correlation to see if it actually holds up to your eyeball’s impression.
Amazing this was linked to RealClearMarkets.
JPB,
Would you find it more convincing, more “substntiated” if Kimel were to, say, provide a book-length exposition of some of his ideas regarding government policy and policy outcomes? “Cause he has. There is a link around here somewhere that would allow you to read his fairly detailed look at actual outcomes – with real live correlations.
Even if a book seems like too much reading, there’s this:
“I also plan to link the three graphs together into a bit of a story, a story which extends from my recent posts, and which I intend to build on over the next few weeks.”
So the business about not substantiating opinion, single, narrowly focused narrative, failure to calculate correlation – that might just be premature. In fact, I don’t think you do a very good job of substantiating your opinions, given the apparent narrowness of the sample of Kimel’s work you have taken into account.
Maybe I haven’t….but is the burden of proof on me? My main point is that he is drawing a conclusion on the basis of a correlation. The strength of the correlation – statistically speaking – is not specified. He does not indicate that he even realizes that he is talking about a correlation, and that correlation is not the same thing as causation. And, if we assume that the correlation is due to a causal effect, he does not explain, in this blog post, why the relevant correlation is between GDP and taxation rates _in the same year_. Maybe it’s more appropriate to correlate GDP X number of years following a change in taxation rates, etc. I didn’t write this article, he did; I think it’s incumbent upon him to substantiate his opinions better.
John Paul Boffard,
OK. I didn’t link to a lot of my past posts in this one. Sorry.
That said, I also know you can’t prove causality. The best you can do is granger causality, and that only says one thing leads the other.
Here’s a post that shows the correlation between tax rates in one period and a number of other economic variables in later years: http://www.presimetrics.com/blog/?p=253. As I said, one can’t say tax rates cause these other things, but you can say they lead. If there is causation, the thing that leads is the cause.
All that said, the main point of this post is simple: I’m pointing out three easy graphs that completely contradict what most people believe. Higher taxes have not been associated with slower growth. Bigger government has not been associated with slower growth.
That was the main point. I just decided to tack on a story that is consistent with the graphs. Maybe I didn’t prove my story, and at no point do I claim I did prove it, but the point is, the story is consistent with the graphs. What I was taught in my macro classes isn’t. You don’t think that’s worth a post and I do.
denis haty,
In general, when top rates go up, so do the other marginal rates. When top rates go down, so do other marginal rates.
JPB –
Mike has been posting about this stuff here for years, and has looked extensively at time lags. You are criticizing this post with no sense of context.
So – yes, the burdon of proof is on you, since you seem not to have any knowledge that this is one of many, many posts that have covered this information in a variety of ways, and cut across the data from many diffferent directions.
Also, Mike has validated that the top marginal rate is more meaningful in this analysis than is the effective rate, and has posted extensive statistics on his analyses, many times.
There are tons of prior posts and data and analysis that Mike has put together – even a book as kharris pointed out.
Maybe the good people at RealClearMarkets know something you don’t . . .
Cheers!
JzB
Frank –
Mike linked his sources and offered his sread sheet. Why don’t you take on what you suggest, instead of giving a homework assignment.
Cheers!
JzB
Actually, non-defense spending does break, while defense spending does not. I just happened to look at that.
http://jazzbumpa.blogspot.com/2011/01/government-spending-and-great.html
My impetus is an examination of evidence that the post WW II years exist in two quite different economic realms, before and after about 1980.
http://jazzbumpa.blogspot.com/2010/10/us-economy-is-dying.html
Part of the reason is the diversion of credit into financial tail chasing, and away from actual productive investment – facilitated by increasing wealth disparity since the 80’s, in turn facilitated by the tax rate being not only too low, but paid by people who’s incomes are too low. (Lack of progressivity.)
http://www.asymptosis.com/who-prints-money-and-who-gets-to-have-it-its-up-to-the-banks.html
You see, the more you did into the data, the more you realize that Mike’s narrative is internally coherent, consistent with any data you’re going to find, and not a bit cute.
Cheers!
JzB
So you’re saying, basically, that it was a self-fulfilling ideology? I like the notion, but I’m having trouble understanding the (interrelated) arrows of causation.
Your narrative has too many causatives (especially if the asterisked footnote is included), not really linked together coherently, for me to find it satisfying.
But I think it’s going somewhere…
Steve – Yes, I think there’s a bit of a self-fulfilling prophecy there. I think there’s more to it too – wanting low taxes doesn’t make them effective. Trumpeting that the gov’t sucks often enough though might make gov’t less effective… and thus make taxes less effective at benefiting the economy too.
The always fabulous Steve Randy Waldman:
“the fabulously successful strategy of governing incompetently while using each failure as evidence that government action cannot help but be corrupt and inept. Heckuva job, Brownie!”
Something’s not right. According to the NIPA tables you cite, the federal government hasn’t spent over 15% of GDP since about 1956, and hasn’t spent over 10% of GDP since around 1975. I’m not sure what that’s measuring, but it’s not the size of the federal government as commonly understood. Can you explain the difference between these numbers and the OMB’s?
If you superimpose the NIPA spending numbers over the OBM’s (Table 1.3 from http://www.whitehouse.gov/omb/budget/Historicals), you’ll see the linear trendlines practically mirror each other across the 15% line. Using the OBM numbers, the correlations betweeen spending and T+1 growth for the 80, 70, …20 years through 2009 are quite different. The average of those looks to be a little under 3%, and the average of the 8 individual decades is about 3½%, with samples ranging from 90% to -63% if I did things right. Might be worth a separate analysis.
There’s also another interesting trend if you look longer-term at the OMB numbers from 1931 to 2011. If you correlate the surplus/deficit and the margin of outlays over/under 20% of GDP with GDP growth for 1, 2 and 5 years out, they follow a similar pattern to each other.
Over the entire period, deficit spending exhibits a positive correlation of 20.5% for growth at T+1, but a -12.4% for the average growth in the subsequent 2 years, and -23.2% for the subsequent 5.
Federal outlays over 20% of GDP have even stronger correlations with GDP growth in the further out periods. At T+1 the correlation is +18.2%, but it’s -43.8% for the average of the next 2 years’ growth, and -58.9% for the average of the next 5.
…and federal spending over 20% of GDP exhibits an even stronger negative correlation on subsquent 10-year average growth, at about -78%. There’s definitely something interesting about the tendency for that correlation to strengthen as the out-periods get lengthier.
Adding graph of the correlation between federal spending above/below 20% of GDP and average economic growth for the next 1, 2, 3… 10 years. The clear directionality of the relationship is interesting, no?
Just eyeballing the data, it’s apparent that growth rates have been slowing since shortly after WWII. Thus, comparing the recent lower-growth era (with lower marginal tax rates) with the earlier era (with higher marginal tax rates) you at least need to back out the trend. Second, it’s apparent that higher marginal tax rates are associated with a MUCH higher level of growth volatility, and volatility hurts long-term growth (ask any investor how that works). Third, any measure like GDP that goes up during a catastrophic war, and then down when that war ends, is a highly suspect measure of national welfare – however “conventional” its use may be.
I think this data is unsufficient to form a conclusion. First of all, state and local spending is not included, which is a part of the government. Also, more countries should be included. See, that is a common thing about Americans, making general conclusions, when looking only in their own yard. This is almost as if you want decide, if copper conducts electricity, when making your experiments with only one sample of copper… or estimating how common should be life in the universe, when knowing only about one case of its emergence. I can go on…
The author also talks about what is the value of correlation, but never defines the type of correlation.
Frank,
You may be on to something, but if those parts of GDP are unproductive, they should not actually lead to long term growth. This study does not show us what is the link there, which is really what is important at the end.
Figure 2 shows that the Fed. Spending incl. defence is about 8% of gdp in 2008. I’m not sure that that is correct at all.
If you go to http://www.usdebtclock.org/index.html as of 12/31/2011
– gdp = $15.1T
– Fed. govt. spending = $3.6T … this is 23.8% … certainly not 8% that you show in figure 2
– Total govt. spending (fed,state,local) = $7.0T … this is 46%… i.e. about half the economy!!
Nobody would be complaining about of the size of fed. govt. if it were 8% as your show … something is not right with your data? Please clarify. I will need to look at your sources.
Figure 2 shows that the Fed. Spending incl. defence is about 8% of gdp in 2008. I’m not sure that that is correct at all.
If you go to http://www.usdebtclock.org/index.html as of 12/31/2011
– gdp = $15.1T
– Fed. govt. spending = $3.6T … this is 23.8% … certainly not 8% that you show in figure 2
– Total govt. spending (fed,state,local) = $7.0T … this is 46%… i.e. about half the economy!!
Nobody would be complaining about of the size of fed. govt. if it were 8% as your show … something is not right with your data? Please clarify. I will need to look at your sources.
Figure 2 shows that the Fed. Spending incl. defence is about 8% of gdp in 2008. I’m not sure that that is correct at all.
If you go to http://www.usdebtclock.org/index.html as of 12/31/2011
– gdp = $15.1T
– Fed. govt. spending = $3.6T … this is 23.8% … certainly not 8% that you show in figure 2
– Total govt. spending (fed,state,local) = $7.0T … this is 46%… i.e. about half the economy!!
Nobody would be complaining about of the size of fed. govt. if it were 8% as your show … something is not right with your data? Please clarify. I will need to look at your sources.
Figure 2 shows that the Fed. Spending incl. defence is about 8% of gdp in 2008. I’m not sure that that is correct at all.
If you go to http://www.usdebtclock.org/index.html as of 12/31/2011
– gdp = $15.1T
– Fed. govt. spending = $3.6T … this is 23.8% … certainly not 8% that you show in figure 2
– Total govt. spending (fed,state,local) = $7.0T … this is 46%… i.e. about half the economy!!
Nobody would be complaining about of the size of fed. govt. if it were 8% as your show … something is not right with your data? Please clarify. I will need to look at your sources.
Figure 2 shows that the Fed. Spending incl. defence is about 8% of gdp in 2008. I’m not sure that that is correct at all.
If you go to http://www.usdebtclock.org/index.html as of 12/31/2011
– gdp = $15.1T
– Fed. govt. spending = $3.6T … this is 23.8% … certainly not 8% that you show in figure 2
– Total govt. spending (fed,state,local) = $7.0T … this is 46%… i.e. about half the economy!!
Nobody would be complaining about of the size of fed. govt. if it were 8% as your show … something is not right with your data? Please clarify. I will need to look at your sources.
Figure 2 shows that the Fed. Spending incl. defence is about 8% of gdp in 2008. I’m not sure that that is correct at all.
If you go to http://www.usdebtclock.org/index.html as of 12/31/2011
– gdp = $15.1T
– Fed. govt. spending = $3.6T … this is 23.8% … certainly not 8% that you show in figure 2
– Total govt. spending (fed,state,local) = $7.0T … this is 46%… i.e. about half the economy!!
Nobody would be complaining about of the size of fed. govt. if it were 8% as your show … something is not right with your data? Please clarify. I will need to look at your sources.
The reason the government share of GDP decreased during the Clinton administration is because of the polices put into place by Newt and the contract with America. Why don’t you use a comparision of the real tax dollars collected in each year and compare that number to the federal income tax rate? You won’t do it because it will prove your theory incorrect.