June 19, 2003
From the desk of Mindles H. Dreck:Games With National Accounts
There have been a number of lengthy posts in the blogosphere over the last few months that attempt to prove one or the other point of view about the effects of taxation and fiscal policy by graphically presenting national accounts data. Many of them have made elementary errors (some that surprised me, given the source) that I have wanted to describe, but for time or lack of esprit de blog combativeness, I haven't. Also, I suppose it's good to see people yanking out the data and making what they can from it as opposed to thinking you need some kind of macro-econ license before driving the information superhighway to bea/bls.gov.
However, you might look at this rant to see many of these mistakes in one convenient place. (as the title of the blog suggests, the author usually comments on other topics).
So here's a start on a checklist:
- Don't forget to use REAL growth figures so you don't wind up touting the inflationary Carter years as the sine qua non of economic growth. That one always makes me laugh.
- When comparing GDP across time, don't forget that real GDP is bounded by workforce growth and productivity growth and is therefore difficult to compare across periods where workforce growth varied dramatically. Consider how this rule, in combination with the first, might change your conclusions , or at least challenge your confirmation bias. I've seen several assertions about growth or productivity droughts that just aren't born out under proper measurement.
- When presenting dollars or price levels, or anything that compounds over time, use a logarithmic scale. Otherwise, the same percentage change looks larger in the more recent part of the time series (the graph linked immediately above violates that rule, I'm afraid).*
- Don't equate a change in revenues/expenditures as a percent of GDP with a decrease in absolute revenues/expenditures For those of you trying to disprove supply-siders, Laffer doesn't fall that easily. See Nordhaus for better arguments.
- Don't claim that giving money to the government stimulates the economy and yet, in the same breath, say letting rich people keep their money doesn't.
- For your surviving data series, please don't forget that correlation isn't causation. I ran a correlation the other day on individual income taxes as a % of GDP vs. Real GDP growth since 1941. It's negative 22%. Does that prove supply-siders right? Of course not. Apply the same statistical caveats to the data you are mining.
- Understand the more subtle implications of national accounts, and test your forecasts against them. The first thing you learn in Macro is that national output is the sum of private sector output, government and net exports (Y=C+I+G+X). This isn't a theory, its an accounting identity, how one computes GDP. It follows that the private sector (business and individual) surplus or deficit, plus government's deficit or surplus, plus net imported investment, all expressed as a percentage of GDP must equal zero. Up through 2000, the private sector ran enormous deficits - about triple the long-term average at -6% of GDP. Both business and consumers were dissavers. Government ran a surplus and we certainly imported foreign capital, so these two parts of the identity were positive. Businesses have made a sharp retrenchment, and consumers have followed suit, although much more slowly (thank goodness for that). One of the following two things HAD to accompany this: the government goes into deficit, and/or the FX situation reverses by way of massive dollar depreciation or a complete reversal in trade.
I was surprised at how many people thought after 2000, the private sector could re-save and the budget could stay balanced (with the right President, of course). This could only have happened if the trade deficit disappeared simultaneously. The fact of the matter is that the near term government deficit was practically a given when the private sector deficit bubble of the '90s burst. Trying to resist it would have created a depression. For a good primer on the implications of the national accounts identities, particularly the lesser-of-two-evils deterioration in the fiscal balance in the post '90s environment, see Wynne Godley's 2001 analysis. He and Woody Brock have been pointing this out since 1999.
I'm sure there's more, and I'm sure you'll feel free to stick it, and me, in the comments.
_____
*Actually, Robert Shiller violated scale principals big time in Irrational Exuberance, even as he set new standards for the proper use of inflation-adjusted data. See figure 1.1 in the linked spreadsheet for a potpourri of double-axis deception. He needlessly compresses the right scale and fails to use logarithmic scaling. I would argue the ratio of the left to right scale should be the long-term average P/E of about 15. The top of the right scale would be about 120. Go ahead, you can do it yourself. Doesn't look as extreme, does it? His point, however, survives.
So the effort to add a constitutional balanced budget amendment from 95-98, part of the Gingrich revolution, would have caused a depression within three years of its passage? Good thing we had Clinton in office, keeping us from this fate.
Don't claim that giving money to the government stimulates the economy and yet, in the same breath, say letting rich people keep their money doesn't.
I'm not sure what you're claiming here. If you are claiming that when considering a particular policy, we should note that the stimulative effects of government spending will be offset by the "de-stimulative" (?) effects of taking money away from rich people, agreed. If you are asking me to accept as an axiom that that calculus will always produce a particular result, I can't agree.
hi mhd,
here is another game for you: the "is gdp really a good measure of the extent of economic activity of a country" game. what do you think about using using an alternative measure (that takes into account externalities or non-monetary transactions into account), such as the genuine progress indicator or the index of sustainable economic welfare. according to the gpi, "real" gdp has been falling in recent years.
http://www.redefiningprogress.org/projects/gpi/updates/gpi1998_execsum.html#GPI_S2
What do you use for your statistical analysis, plain old Excel? If so, know of any books/websites where someone can learn how to do it. I haven't done correlations since college and then it was using SPSS.
Here's a basic point.
When comparing GDP growth rates make sure you choose two periods that are in the same point in the business cycle. Don't, for example, try to measure GDP growth from the bottom to the top of the cycle and pretend that is the sustainable growth rate.
Which is what most Reaganites/supply siders do when they measure Reagan's record from 1982-1989.
Think Bartley's (sp?) attrocious book as an example of this mistake.
BTW, not sure what your point #4 means.
Are you saying that you look at nominal numbers?
Roger - yes, if it were not waived or circumvented in some Gramm-Rudmanesque way. Blunt policy instrument meet fragile bubble!
Alkali - this goes to the general tendency to think that rich people keep their savings in a sock and that these funds will be liberated by taxing and redistribution. To be fair, the linked author attempts to justify this through the dubious logic of the liquidity trap - that banks and the money markets are basically mattresses and that government redistribution produces a higher velocity of money even after the frictional costs of redistribution.
cas - a bit off topic, but I'm happy to say I have already written on a similar suggestion. In the case you cite, the usual list of desired social correctives (inequality, sustainability, etc.) bastardize the measure. the definition of "sustainable" would have fluctuated..er..unsustainably over the decades, no?
Sean- Excel is pretty powerful now and all that is necessary for most statistical measures on small datasets. I also use an Excel add-in called @risk and, for large datasets, SAS.
GT - true re. business cycle, althought you may have problems identifying equivalent points other than peak and trough. Also a problem for folks who compare 2000 to 2002, right? No shortage of this silliness on either side.
#4 means that people slip back and forth from measures relative to GDP and dollar measures (inflation-adjusted or not). In the linked post, the author is trying to show that the Reagan tax cuts were not followed by an increase in revenues by using revenues as a % of GDP. Revenue growth was thereby masked in GDP growth.
Excellent post. Fans of Ed Tufte's work will recognize some of the errors you point out.
The failure to correct for inflation is pervasive, and ruins many arguments. I don't think that some of the louder voices today really experienced the Carter years. You had to be there; there's been nothing quite like it since, thank God. Although I wish my family had locked in some of those 18% Treasuries. . .of course, it took an especially self-confident person to jump into fixed-income investments during a period of galloping inflation!
>
You've already addressed this, but I'll comment anyway in slightly different terms. Stealing money from rich people so that the government can spend it is certainly not Pareto optimal, but there may be cases in which an argument can be made that satisfies the Kaldor-Hicks criterion for optimality (the position from which economists make the judgment that perfect competition is more efficient than monopolistic market forms).
The topic of spurious correlation is an interesting topic. Jevons is always the first person that comes to mind, but there is a plethora of examples. I like the one below because of its mention of sunspots, as in Jevons' theory that sunspots caused business cycles. "Ole-Jørgen Skog reports that the correlation between the quarterly index of intravenous drug abuse in Stockholm and Wölfer's index of sunspot activity is 0.91" (www.geocities.com/hmelberg/papers/pd_causal2.pdf)
I'm new here... I didn't realize the characters I used for quotation would be treated as HTML (though I should have). The first part was re: #5.
Mindles,
I agree that there is silliness on both sides. But i was referring to something the right claims that, AFAIK, has no counterpart on the left. And that is that a favored policy of theirs raised the permanent level of economic growth. They use as proof the 82-89 period which is economically bogus. And there are ways to check that you are comparing like with like. Blinder suggests using the unemployment rate as a good proxy.
On the point # 4. Yes, you can use income tax revenues to GDP to test the supply side theory. It's not the best way but it's a simple way to do so. Unless you think that the taxable personal income to GDP ratio changed dramatically one year to the next. So, for example, when personal income taxes drop from 9.2% of GDP in 1982 to 7.8% in 1984 it's telling you something. Something that directly contradicts the “the tax cuts paid for themselves” argument.
Finally you linked to graph showing GDP per capita labor force. Not sure what you meant it to prove but that graph is a good example of the limits and misuses of charts.
First there is selection bias. Why start with 1960? What is it you are trying to show? Why not 1945, or 1975?
Second, and more importantly, there is the selection of scales. Note the y axis scale? Why does it start at 30,000? By doing so it exaggerates the differences over time.
GT you needn't look beyond the 'rant' I linked to find equivalent arguments, but you and I aren't going to get anything but insomnia arguing about partisan myopia.
So, for example, when personal income taxes drop from 9.2% of GDP in 1982 to 7.8% in 1984 it's telling you something. Something that directly contradicts the “the tax cuts paid for themselves” argument.Last time I checked, bills were paid in dollars, not percentages of GDP. When I bought my house, it cost a multiple of my income. Today it is worth a smaller multiple of my income. Oh no! did my house depreciate? No, in dollars it is actually worth twice what I paid for it (in real dollars too, thankfully). The calculation totally misses the most important point in paying for things - growth in income.
Supply-side logic does not require that revenues remain a constant percentage of income. Just the opposite, in fact, it suggests that growth in income will make up for the lower taxation percentage. If 7.8% of a later year's GDP is greater than 9.2% of a prior year's GDP that's all that is necessary to "pay for the tax cut". No matter how you slice it, the government's revenue grew over the years after the Reagan tax cut. Whether the tax cut caused that growth is, of course, another matter, and has to be viewed in terms of workforce growth, the end of a horrid recession and declining oil prices and, the greatest unknown, the multi-tiered effects of technological innovation.
The linked graph you criticize has the virtue of controlling for changes in the workforce, per point #2, and is also responsive to the linked 'rant' which fails to consider labor force change. I would argue (and did in point #3) that the Y-axis should have a logarithmic scale. The minimum isn't a bad flaw because it affects all points on the line equally. Linear scale affects the observations unequally across the X-axis.
Incidentally, I didn't make that graph, I googled it up looking for an analysis that controlled for change in the workforce.
"... the inflationary Carter years as the sine qua non of economic growth."
Nitpick: that should be ne plus ultra, or some other such phrase.
Mindles,
I didn't mean to say that you had created that graph. I was just pointing out some of the 'problems' the graph had.
I agree that the graph controls for the size of the labor force. But I'm still not sure what the point was. What exactly is proved or indicated by that graph? In any case the use of 30,000 as the starting point is a bad one because it exaggerates the differences. If I hadn’t just packed my Tufte books I would check because I seem to recall it’s one of the examples he used.
I don't agree that a log scale would be useful in such a graph. The differences are not that big. A log graph makes sense if some of the observations are orders of magnitude greater than others. But that's not the case here.
Finally on the supply side and GDP. Maybe you and I are talking about different things. What I was referring to is the argument, which some have made, that Reagan's tax cuts so increased tax collections that the result was that taxes collected after the tax cut were the same (or more) than what they would have been had no tax cut been implemented. So if you collect $100 after the tax cut that means that you would have not collected more without it. It's a free lunch, if you will. You cut tax rates AND you still get the same amount of tax collections you would have gotten otherwise. This is what is normally understood to mean by tax cuts paying for themselves. I am not sure in what sense you use the term. The fact that over time nominal tax collections rose under Reagan does not in any way prove that they were as high as they would have been without the tax cuts
In fact, this idea that the tax cuts paid for themselves (as I define it above) is completely false. And it's not just me saying it or Krugman or some other economist conservatives disapprove of. It's your very own co-blogger, as well as Stephen Moore of the National Review, among other conservatives. In fact a quick Google search will show many instances of conservative economists very emphatically denying that anybody ever claimed that tax cuts would pay for themselves. Moore estimates around 30% of return, meaning that for every $100 of a tax cut you get some $30 back in taxes due to higher growth. Boskin (yes, the same one of the SS study you cited) also agrees that nobody in the Reagan administration claimed that tax cuts paid for themselves and estimates the return between 40% and 50%. This means that the size of the tax cut is smaller but it is still real and the tax cuts do not pay for themselves. I would note that both the CBO and the Joint Committee on Taxation have recently looked into this and come up with much lower estimates of return, and in some scenarios closer to zero.
Using the GDP ratio is a perfectly valid way to test that although it's not perfect. It adjusts for population and inflation but not in a perfect way. All other things equal you would expect that tax collections should mirror GDP growth, at least over the very short term (one or two years). Over longer periods of time there may be changes in the GDP composition that would change that but it is highly unlikely that an almost 1.5% of GDP drop in two years would happen by itself.
In any case since after the tax cut even nominal tax collections fell you probably don't need to even use the GDP ratio.
What exactly is proved or indicated by that graph?It's difficult to prove things with a graph. I guess that's my point in a nutshell. It argues against the points about growth made in the linked post (one more time...). And as for the scale, last time I checked $5,000 on 30,000 was twice the percentage difference as on $60,000. So if one looks at the slope of the line (we are talking about growth, right?) a constant slope is a decreasing growth rate. That's a hard visual adjustment.
The fact that over time nominal tax collections rose under Reagan does not in any way prove that they were as high as they would have been without the tax cuts [..and ensuing paragraph]
What's with the strawman? I spend an entire post saying you can't prove either the supply-side or anti-supply side case by spinning national accounts data and you give me two gratuitous paragraphs as if I did? Is it just that you want to say this somewhere? Anyway, for the spectators, data is here if you want it.
I still think you're reading a lot into "paying for themselves" but I can't affirm or contradict anyone's generalised mind-reading of supply-siders.
Then I misunderstood your point about tax cuts. Sorry. I thought you were saying that tax cuts had paid for themselves as I defined it above.
As for GDP ratios although they can't 'prove' they offer very, very strong evidence. Especially if you couple it with the drop in nominal collections. The combination of a drop in total income tax collections the year after the tax cut was approved plus the fact that income taxes dropped 1.5 points of GDP in two years, when normally you would have expected them to be stable, is telling you that tax cuts did not pay for themselves, that, in fact, they reduced tax collection from what they would have been.
I don't read much into supply siders. There are all kinds of people making all kinds of claims. I tend to think that the term itself is meaningless, something that only exists in press reports not in real economic thinking.
I was just pointing out that the idea that tax cuts paid for themselves (which I now understand you don't claim) is false and that you can point that out, however imperfectly, by using GDP ratios (which I think you do claim and therefore we disagree on that).
One last point. The link you provided has aggregate tax data. If you are going to discuss the impact of a cut on income tax rates it's better to look at income tax collections. Available here.
Saturday night at 11:00 and we're doing this basically real time. DO we have lives?
As for GDP ratios although they can't 'prove' they offer very, very strong evidence. Especially if you couple it with the drop in nominal collections. The combination of a drop in total income tax collections the year after the tax cut was approved plus the fact that income taxes dropped 1.5 points of GDP in two years, when normally you would have expected them to be stable, is telling you that tax cuts did not pay for themselves, that, in fact, they reduced tax collection from what they would have been.
OK, I can't leave this alone. No supply-sider would suggest that government should maintain a constant or growing share of a growing national product. It's not consistent with the religion. 'Pay for themselves' means produce enough dollars to continue the programs as is - nothing to do with share of GDP, and every supply-sider would devoutly wish taxes/government as a percent of GDP to shrink.
Furthermore, you can't possibly expect anyone to accept your version of "what they would have been". You have keys to the parallel universe where taxes weren't lowered and other policies weren't put in place? Please.
Finally, I suppose the one year dip in receipts is all important and the fact that over the ensuing years revenues grew substantially means nothing? You think marginal rate incentives occur immediately (even when you don't use the new rates on your return for a year)? Interesting, even your hero Krugman argues (constantly) that income tax cuts take time to produce whatever stimulus they will.
Your characterization of pro-tax cutting arguments, like your convenient definition of "pay for themselves" is a strawman that holds no particular meaning.
I fail to see how changing the accounting side - fooling around with taxing current income vs. future income (income taxes vs. deficit borrowing) - should have any effect on government *spending* as a percentage of GDP. Taxes as a percentage of GDP, sure, but la-de-da; I don't think "we'll pay for all this shit later" is an expressed creed of the movement. Maybe subconsciously revealed, but not expressed.
The marginal preference for total government spending should track *income*, to first order, not tax rates.
Comments are Closed.