The expansion in government and poor state of the economy got me thinking about how government growth is reflected in measured gross domestic product. So here is a wonky look at the treatment of government in the Bureau of Economic Analysis GDP data.

Data notes: By "government," I mean total federal, state, and local. For 2009, I'm using the average of second and third quarter data. All data from BEA Tables here.

GDP measures total production. In 2009, government production was 20.7 percent of U.S. GDP.  Government production is roughly the sum of government value-added (the stuff it produces itself) and government purchases. The first item, government value-added, was 12.4 percent of GDP and mainly consists of employee compensation. For example, the Pentagon produces output by adding together fighter pilots, which it hires, and fighter jets, which it buys.

A more commonly cited measure of government is total government spending. In 2009, that was 38 percent of GDP. The difference between this number (38 percent) and the production number (20.7 percent) is 17.3 percent, and represents the sum of government interest payments and transfer payments to individuals and businesses.

Figure 1 shows how the three measurements of government size have changed over time. Government production has remained fairly stable as a share of the economy, but total government spending has soared. The growing gap between these two lines mainly represents the massive growth in transfer (or subsidy) programs, such as Social Security.

How Does Government Growth Affect Measured GDP?

Consider how the recent rise in government spending might have affected measured GDP. First, let's look first at the production part of government spending. The important thing here is that we don't know how much government workers actually produce because their output is generally not sold on the market. As a consequence, the BEA measures their output as the sum of their compensation amounts. Also, we know the dollar value of the things the government buys, but we don't know how much those intermediate goods actually produce when in the hands of the government. So the government production portion of GDP seems kind of shaky, despite the superb efforts of the BEA to assemble all the data.

Anyway, let's say the government adds a new worker with pay of $100,000, the BEA measures GDP being boosted by $100,000. But it might be that the worker doesn't actually produce anything useful, and he adds zero to the economy's actual output.

If the government hires that worker away from the private sector, private GDP would go down by about $100,000. As a result, overall measured GDP would be unchanged. But that would be incorrect because the economy's actual output fell by $100,000.

So let's say the government spent $100 billion to hire a million new government workers. Let's say half of those workers produced as much value as their salaries, but the other half produced nothing of value. The result of this government expansion would be that the BEA would overestimate U.S. GDP by $50 billion. (I am assuming that the government's hiring doesn't change the unemployment rate. I'm also ignoring the distortionary effects of higher taxes).  

Now let's look at the transfer or subsidy portion of government, which equals 17.3 percent of GDP.

Let's say the government increases transfers by $100 billion, perhaps by increasing Social Security benefits, and funding it by higher taxes on wages.

If there are no behavioral responses among taxpayers and benefit recipients, measured GDP would be unchanged, which would be the correct answer.

But of course there would be behavioral responses. The higher taxes would induce people to work less and the higher Social Security benefits would induce people to save less and retire earlier. The results would be that output would fall, and that would be accurately reflected in measured GDP.

In sum, my purpose here was not to explore how a growing government affects the economy, which is a huge subject. Instead, it was to explore whether measured GDP accurately reflects changes in the size of government. The answer appears to be that the transfer part of government spending (17.3 percent of GDP) would be accurately reflected in a shrinking GDP, but that the production portion of government spending (20.7 percent of GDP) may not be. If workers produce less output when they work for government than when they work in the private economy, the latter portion of measured GDP will be overstated.