Every quarter the government releases its estimate for the gross domestic product of the United States. At the same time, it also releases an estimate for the gross domestic income of the United States. Confusingly, if you crack open an economics textbook it will tell you that GDP and GDI are the same thing — identical by definition, as if you were to release a person's mass and a person's weight-at-sea-level-on-planet-Earth simultaneously.
And yet Wednesday the government announced that it is going to start tracking a third measurement of the United States of America's economic activity — gross domestic output, which is defined as the average of GDP and GDI.
Yes, the average of two identical quantities.
And no, it's not a joke. There is good reason to believe that this will improve the accuracy of our understanding of the state of the economy, and that in the future we may find ourselves using GDO as our standard metric of economic activity. GDP itself is relatively new to the game, and used to be overshadowed by gross national product, so there's no reason to think it should reign supreme forever. But more than a tweak to government accounting, the switch is a window into the gap between the pristine clarity of economic theory and the murky swamps of economic reality.
So what the heck are these numbers?
The canonical equation that defines GDP goes like this:
GDP = Government Purchases + Consumer Purchases + Business Investment + Exports - Imports
And the canonical equation that defines GDI goes like this:
GDI = Employee compensation + Profits + Taxes - Subsidies
The key trick, however, is that these quantities are equal.
- GDP measures the total value of everything that is sold. It does this by adding everything the government buys to everything businesses buy to everything consumers buy, and then adjusting for things bought by or from foreigners.
- GDI measures the total earnings accrued by selling things. It does this by taking business profits, plus money businesses pay to their employees, plus taxes businesses pay to the government, less subsidies businesses receive.
The total value of everything sold is the same thing as the total earnings amassed by selling things. It just is.
But it isn't:
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/3921834/Screen%20Shot%202015-07-30%20at%202.51.03%20PM.png)
The good news is these two bars do add up to being pretty similar-looking. The bad news is that $300 billion is actually quite a lot of money. The statistical discrepancy is about equal to the entire economic output of Israel or Denmark. It's similar in size to the combined economic output of Alaska, Maine, South Dakota, Wyoming, Rhode Island, Montana, and Vermont.
Those are, admittedly, small states. But if the government just forgot to count them in a survey, people would think it was a big problem.
Improving measurement with GDO
The problem is that while GDP and GDI are, in theory, identical in practice they are constructed by adding up data sets that, like all human endeavors, are measured by human beings. We can't measure household consumption perfectly and we can't measure workers' compensation perfectly either. In life, nothing is perfect.
But just as you can improve your understanding of the state of public opinion by averaging several polls to reduce sampling error, by averaging together GDP and GDI you should be able to reduce the extent to which measurement error biases your estimate.
Taking two different stabs at measuring the same thing and then averaging them together, in other words, is more likely to get you close to the true quantity than taking either measurement approach alone.
Economics in the real world is hard
At a conference, an economist once observed to me that one problem with fancy economists with jobs at prestigious schools is that they get to spend all their time working with theoretical quantities. Less-fancy economists with less-fancy jobs need to spend more time working with the murky and often annoying process of actually assembling economic data.
That can leave them more attuned to the practical policy problems that arise due to measurement error and ambiguity.
For example, in 2009 the Obama administration and their friends in Congress outlined an economic stimulus package that they thought would be roughly appropriate in size given the scale of the recession. What they did not know what that subsequent revisions to the data would reveal that the economy collapsed much more rapidly in the winter of 2008-'09 than the initial estimates had shown. Had they been more mindful of the error-prone nature of initial data, they might have designed their legislative proposal somewhat differently.
Another example is that the tax code provides for investment income to be taxed at a lower rate than labor income.
This is supposed to encourage well-to-do people to save and invest in a way that encourages long-term economic growth rather than squandering their earnings on yachts and fancy cars. But it also creates a large incentive for people to simply find ways to classify their income as investment income. Thus the general partners in hedge, venture capital, and private equity funds managed to identify a legal loophole through which much of their management fees can be classified for tax purposes as investment gains.
This costs the Treasury money, of course, but it also creates misleading official statistics — labor income vanishes and reappears as investment income even though nothing in the real world has changed.
