Garett Jones  

Government Hiring: Raising GDP by Definition

Welfare as a Magnet for Immigr... Work Experience and Job Perfor...
GDP, Gross Domestic Product. The number gets a lot of attention, deservedly.  You'd be foolish to use it as your sole economic statistic but you'd be just as foolish to ignore it and go with your gut.  

Today I'd like to draw attention to one of the peculiarities of GDP.  For your consideration:

Scenario 1. Tomorrow, ExxonMobil spontaneously hires an unemployed petroleum engineer for $100K per year.  She spends a year looking for new oil, finds nothing.  

Scenario 2. Tomorrow, the federal government spontaneously hires an unemployed petroleum engineer for the same $100K.  She spends a year looking for new oil, finds nothing. 

So, how do these two alternative scenarios impact the official GDP figures? 

Scenario 1 has zero impact on GDP: No oil to sell=no extra consumer purchases=no extra GDP.  As the Bureau of Economic Analysis says, "Personal consumption goods and services purchased by persons..."

Scenario 2 raises GDP by $100K.  As BEA says, "Government consumption expenditures...consists of...compensation of employees..."

Hiring a worker who (through no fault of her own) accomplishes absolutely nothing raises GDP if the government does the hiring.  Hiring a worker who (through no fault of her own) accomplishes absolutely nothing does nothing to GDP if the private sector does the hiring.  

Why? Because GDP counts government salaries as "government expenditures" as soon as the government hires a person.  But the "consumption" and "investment" parts of GDP only count genuine purchases by the private sector (leaving the oddities of imputed spending for the coda below).  

So if a private sector product spends years in the incubator, burning through thousands of person-hours of work and millions of dollars of salary--but never sees the light of day--then the product never shows up in GDP.  But if the government had hired those same workers who worked just as long on a similarly fruitless project, their labor would give a big boost to GDP. 

Government hiring creates GDP by definition.  Private hiring only creates GDP if the worker actually creates a product.  

Coda: I'm not making an argument for changing the BEA's rules to include private sector R&D or even investment in organizational capital in GDP.  Whenever possible, let's keep GDP as a measure of genuine arms-length, market-test-passing purchases of goods and services.  GDP already includes a lot of imputed consumer spending, as Michael Mandel reminds us; we don't need to add imputed spending to the investment side as well. 

Comments and Sharing

COMMENTS (24 to date)
Matt writes:

Would you argue for excluding government expenditure on labor to resolve this discrepancy? We don't need to add, but why not subtract for consistency's sake?

I also wonder how much of the fiscal multiplier can be put up to this (recording the salaries of government employees in addition to whatever portion of those salaries they spend on final goods and services)...

Our compulsive need to reduce incredibly complex phenomenon down to one number enables no end of mischief. Just look at the CPI or unemployment numbers.

Daniel Kuehn writes:

I think I'm confused... the income approach to GDP seems like it would capture this. Are you meaning that there's a commensurate loss in profits in the private sector if they don't find oil?

Garett Jones writes:



Hiring a useless private sector worker has no net effect on GDI, though it has a gross effect. The higher useless wages cause lower profits. So when the CFO hires his useless nephew GDP is unchanged.

Hiring a useless government sector worker has a net and a gross effect on GDI. So when the Deputy Secretary hires his useless nephew GDP rises.

I really this this matters for purposes of measuring GDP volatility: G is more stable than C or I by construction, not just because of policy choices.

[Glossary for nonspecialists: GDI=Gross Domestic Income, equal to GDP via the circular flow.]

blink writes:

How about if the worker is productive? Suppose the worker finds either $100K or $200K in oil, which is sold. Under private employment, GDP increases by $100K or $200K respectively. What about under government employment? Is it always $100K (=worker compensation) or is it output + $100K?

Garett Jones writes:


Yes, that's just what I'm saying:

When the private sector makes a pointless hire, it comes out of profits. Wages rise, profits fall, No (net) effect on GDP.

When the government makes a pointless hire, it counts in GDP: Wages rise, but there are no profits to fall.

Another way to see this: The private sector contribution to GDP is measured as outputs, but the government contribution to GDP is measured as inputs.

Brent writes:

@ Daniel

Yes, I had the same question right away, and that's how I answered it. If firms pay the worker 100k in the worker is unproductive, then that is a cost without corresponding revenue... i.e., less profit.

Garett Jones writes:


Let me guess, based on what I know about how government calculates state university's contribution to GDP:

Private sales of government-made gasoline would just show up in C. As long as government gas sales are larger than the cost of the gasoline operation (likely?) that's the end of the story. If not, then the voluntary gas revenue shows up in C, while the uncovered costs (costs of operations - total sales) show up in G.

eric falkenstein writes:

I remember Hyman Minsky always telling me that Investment and Government Spending were the same because both were in a sense exogenous. Further, Robert Eisner used to always argue government spending was great because the productivity spillovers were under-reported by a factor of 2 or something. Ad to that that we are always below full employment in real time (so, multipliers are always >>1), and there's never a reason not to spend more government money at any time.

I think the key to countering this argument is to look at what they actually spend money on, such as shoring up state pensions, or transfer payments.

Let's Be Free writes:

Great post. It’s often forgotten that the GDP equation is really nothing more than a static macroeconomic accounting identity. It sheds no light on the microeconomic focused value creation processes that support GDP growth (like value added production and exchange) or impede growth (like deadweight losses caused by moving resources away from productive activity or interventions that impede productivity). This is where Krugman misses the boat more often than not. Math is not economics and economics is not math. GDP=C+I+G+(X-M) doesn’t tell you whether an economy is doing well or not anymore than Assets=Liabilities+Equity tells you whether or not a business is being well managed. It’s not the score that counts so much as how the score changes over time. Arbitrary and unproductive G is a loss to the collective weal. And rather obviously, though there are counter examples for sure, an employee hired in the private sector has a much greater chance of being productive than one who is hired by the government.

Richard writes:
Scenario 1. Tomorrow, ExxonMobil spontaneously hires an unemployed petroleum engineer for $100K per year. She spends a year looking for new oil, finds nothing.

It's sort of distracting when you take a job that is stereotypically male and make the hypothetical person in your example a "she." I was reading a book on crime once and it said something like "Now imagine a criminal with a long history of violence. She..." I just stopped reading. I had the picture of a tattooed male thug in my mind and did not like the jolt I got from being reminded how much radical feminism has triumphed over the last few decades. I don't like this kind of stealth social engineering through language. Please reconsider.

Ben Hughes writes:

Does this imply that GDP is "overestimated" (that's the wrong word, but can't find the right one...) for governments that exhibit a high degree of spending as a proportion of GDP (net transfer payments)?

Richard writes:

To the substance of the issue, could this and other statistical manipulation partly explain why Krugman and others argue that government spending in order to stimulate the economy "works"? Since it looks like they can always point to a higher GDP by definition.

Jon writes:

We'd be better omitting G from GDP. Sure that not enitirely fairly. There is some G in there, but we can be cognizant of that in our heads and avoid nonsense such as the multiplier of digging holes is at least one. Seems like a win.

Bill Woolsey writes:

This argument confuses GDP with consumption.

Now, if spending by oil firms on employees looking for new oil doesn't count as investment, but rather is counted as a cost of producing existing oil, then it has been mismeasured.

The cost of the search should be counted against the value of the new oil found. If none is found, then there will be future losses.

By the way, all of the value of the new oil would (and should) be counted as future GDP. The past labor expense of finding it should be depreciation to get net national income.

In my view, the key issue is the need for current resources. Government workers and oil searching workers both require resources today.

A shift of workers from cooking food in restaurants (providing current consumption services) to either searching for new oil in a private firm or else pushing paper in a government office, involves the use of current resources and involves an opportuntity cost of the foregone restaurant meals.

Whether the government services are valuable or oil is actually found is not the point. Treating the shift as a worsening of the technology for oil production today is a conceptual mistake. Treating the expansion in the use of labor by government the same as the restaurant blowing up is also a mistake.

By the way, newly produced capital goods are not counted according to the discounted value of the future returns they actually generate. Historical GDP is not revised according to the realized returns from capital goods.

GDP is supposed to be measuring current output. I think it is a mistake to critique it as if it was supposed to be measuring welfare.

It is appropriate to critique it if it misses current output. And labor used as an investment in future oil search (or any research and development) should be counted as current output.

I am not sure about all the expenses of oil search, but....

"However, there are certain cases for which the NIPA treatment differs from that
used in business-tax-accounting. For example, the exploration and drilling costs
associated with unsuccessful drilling activities (“dry” holes) are treated as expenses by
the petroleum industry but as investment in the NIPAs."

There is also has been a move to shift all research and development to investment rather than wrongly treating it as an expense of current production.

Bill Woolsey writes:

"As part of the upcoming 2013 comprehensive revision, BEA is planning to treat research and
development spending and the production of long-lived artwork as capital investment in its core accounts
(see Jennifer Lee and Andrew G. Schmidt, “Research and Development Satellite Account Update,” Survey
of Current Business 90 (December 2010): 16–55, and see Rachel H. Soloveichik, “Research Spotlight:
Artistic Originals as Capital Assets,” Survey 91 (June 2011): 43–51). In addition, BEA is exploring the
feasibility of creating satellite accounts that would report investment in a variety of other intangible assets,
such as individuals’ investments in human capital (see Ana Aizcorbe, Carol E. Moylan, and Carol A.
Robbins, “BEA Briefing: Toward Better Measurement of Innovation and Intangibles,” Survey 89 (January
2009): 10–23)."

Steve Roth writes:

By exactly the same logic:

If government hires me and I *do* find oil, GDP is unchanged. My work is valued at cost.

Govt services aren't sold in the market, so it's hard to assign a value to them. Not a big revelation.

Steve Roth writes:

Same broken argument that TC used in TGS:

A. The only way to know the value of a good is to see its price -- what someone pays for it.

B. Government goods are not sold, so there is no price. So we don't know their value, only their cost.

C. The value of government goods is lower than their cost.

But if you want to assert that A and B are true, you can't claim C (without contradicting yourself).

A and B (if adopted) prove that there's no way to know C except via faith.

andy writes:

The same with corruption - corruption makes the money being miscounted as G instead of T(ransfer)...the more corruption, the higher GDP.

MingoV writes:

This doesn't matter to the GDP example, but it matters in other ways. A petroleum engineer who finds no oil did not accomplish nothing. Negative findings are important. She provided information that drilling in certain regions would waste money and time. That information has significant value despite the fact that it adds nothing to the company's yearly revenues (or the GDP).

abhay writes:

In System of National Accounts, oil exploration expenses are capitalised. So the impact of GDP will be similar in both scenarios.

L writes:

The value of the oil that ExxonMobil produces, in terms of its contribution to GDP, is equal to what people ultimately pay for that oil (through their purchases of final products for which oil is an intermediate input of production). When Exxon sells oil, they set the price in order to, among other things, recoup the costs of performing R&D. Although the engineer's salary may not be explicitly added to GDP in scenario #1, it is certainly included in GDP through its impact on personal consumption expenditures (PCE).
(Or, if you want to think of the engineer’s salary as capital investment, you could say that Exxon raises prices to compensate for the negative return on that investment).

For scenario #2, I think it helps to NOT think of the "G" in GDP=C+I+G+(X-M) as simply a measure of what the government spends on consumption and investment... instead, I think of it as an estimate of the value of the goods and services that the government produces. It is impossible to measure the value on government-produced goods and services in the same way that we measure the value of the oil that Exxon produces -- because the government does not SELL the things that it produces! If it did, there would be no need for the “G” in the GDP equation and the two scenarios would be identical (the engineer’s salary would contribute to GDP through increases in PCE). Since we CAN’T do that, the best we can do is to set the value of government-produced goods and services equal to the costs of their production, which would rightly include the engineer’s salary.

(PS - I don’t think it matters, but I don’t think the costs of oil exploration would be capitalized in the National Accounts if they were performed by the government).

PrometheeFeu writes:

@Garett Jones:

I don't think the discrepancy is actually there Garett. If the government hires that engineer, they have to do one of three things:

1) Reduce spending elsewhere by $100,000. No GDP effect.

2) Raise taxes by $100,000 which means income falls by $100,000. No GDP effect.

3) Borrow money. No GDP effect today, lower GDP when the debt is paid.

So really, what you are saying only applies if the government borrows money. But if the private company borrows money, the same thing happens, doesn't it?

PrometheeFeu writes:


How is it not social engineering to keep using the male pronoun thereby reinforcing the idea that only males can be oil & gas engineers despite the fact that males are not the only ones who are oil & gas engineers?

Comments for this entry have been closed
Return to top