How to lie with NGDP statistics?

I'm not very good at this. But I think we need to ask the question. Others could answer it better than me.

Scott Sumner tells me that Greece is considering issuing NGDP-linked bonds, where the amount the government pays to bondholders is proportional to Nominal Gross Domestic Product. Like Scott, I rather like the idea of NGDP-linked bonds. But they do have one potential downside compared to regular bonds, and we need to think about it.

If a government issues NGDP-linked bonds, it has an incentive to lie about NGDP statistics. It has an incentive to say that NGDP is lower than it really is, or really could be. Because the lower is NGDP, the less money the government has to pay to those who hold NGDP-linked bonds.

It's a bit like (though not exactly like) the incentive an individual has to lie about his income, and say that his income is lower than it really is, so he pays less income tax.

[Update: maybe "manipulate" is a better word than "lie".]

I don't know how big a problem this would be for Greece, or any country. Maybe there are ways around it too, like external accountants.

But I'm going to try to think of cases where the exact definition or measurement of NGDP is a bit fuzzy, where a determined government and compliant national statistics office could fudge the numbers:

1. Government expenditure vs transfer payments. Government expenditures on newly-produced goods and services are included in NGDP, but transfer payments (where the government is just handing out cash, and not buying anything) are not included. Hmmm. Suppose the government cut pay to the police by $100, but increased welfare payments for their kids by $100. NGDP falls by $100. (Somebody should to be able to cook up a better example, but you get the idea.)

2. Investment vs intermediate goods. Suppose firm A buys (say) software for $40 from firm B, and uses it to produce final goods worth $100. If you treat that software as an investment expenditure (because it's a long-lasting capital good) then NGDP is $40 + $100 = $140. But if you treat that software as expenditure on an intermediate good (because it gets used up in producing the final good) then the value added of firm A is $100 - $40 = $60, and the value added of firm B is $40, giving NGDP as $60 + $40 = $100 (the value of final goods and services produced). (If we were calculating Nominal Net Domestic Product we would subtract depreciation on the software from Nominal Gross Domestic Product, but depreciation is notoriously difficult to define and measure, which is why we usually talk about NGDP rather than NNDP.)

3. There's the whole underground economy thing, where the people producing the goods and services don't report them to the government (because they want to avoid taxes), and so the government doesn't report it in NGDP. The government still has an incentive to go after the underground economy to get extra tax revenue, but it has less of an incentive if it has NGDP-linked bonds, because some of that extra tax revenue will go to the bondholders.

4. What else?

How big a problem is this? Can it be fixed?

(I don't think this is as big a problem for NGDP targeting, because I don't see as big an incentive for the government to lie.)

On the upside, maybe one of the most boring and arcane parts of teaching macro is going to get exciting and relevant!

Comments

[Bob Smith]

To what extent is an organization like Eurostat dependent on national ststitics offices? I would have thought very, but I wonder if they would have capacity to do that sort of work themselves? You'd still have measurement issues, but at least could be sure that you're measuring the same thing.

I think your point is that even with an independent government agency, the government could still indirectly cook the books just by gaming how NGDP is measured. Its like tax planning in reverse, find out what is defined as being income and arrange your affairs to recharacterize to avoid having amounts that fall within that definition.

[Nick Rowe]

Bob: "To what extent is an organization like Eurostat dependent on national ststitics offices?"

Good question. I don't have a clue.

" Its like tax planning in reverse, find out what is defined as being income and arrange your affairs to recharacterize to avoid having amounts that fall within that definition."

Yes. Like tax avoidance, rather than outright tax evasion. It wouldn't need to outright lie if it substituted transfer payments for wages.

And if someone has an ongoing incentive to figure out ways of doing something, you can bet they will eventually figure out a lot more devious ways of doing it than we ever could.

[Donald A. Coffin]

For the US, and other countries that do something similar...imputed values. In the US, newly-constructed housing counts as part of gross investment. But buried in personal consumption expenditures is the consumption value of owner-occupied housing. It's imputed as "owner-equivalent-rent." [In 2013, this was about $1.3 billion (nominal).] This is a fairly large item. In olden days (i.e., when I was an undergraduate), some farm consumption--value of food produced on the farm for consumption--was also imputed (this is so small now as to be all but irrelevant).

[Donald A. Coffin]

That should be $1.3 TRILLION, not billion.

[Frank Restly]

Nick,

By issuing, does Scott mean Greece is selling NGDP futures for Euros or bestowing NGDP futures without receipt of payment?

If government is selling them, then Ponzi finance becomes a possibility - government makes payment on existing NGDP linked bonds by selling more NGDP linked bonds. Why would a government need to lie in this case? Governments normally are able to roll over the principle of their debt but generally make interest / ROI payments from tax revenue. That need not be the case with gullible enough investors.

If a government is bestowing them then are they perpetual or are they fixed term? If fixed term, where does the Greek government obtain funds to retire them as they mature?

But let me try to answer your question:

"If a government issues NGDP-linked bonds, it has an incentive to lie about NGDP statistics."

The question isn't so much about the incentive to lie (both the buyer and seller of any security or good have that incentive). It is a question of a government (as legal authority) maintaining it's neutrality when two parties accuse each other of lying and it is one of the accused parties.

[Nick Rowe]

Donald: Ah yes! The stock of housing is treated differently from the stock of other consumer durables. NGDP includes housing services, but does not include washing machine or car services, if you rent them from yourself. Presumably, if you converted a laundromat into a customer-owned coop, NGDP would go down. There must be some ways to reduce NGDP along these lines.

Frank: No.

[Prakash]

Hi Nick,

I had commented on both the MR and money illusion threads with this potential solution. Europe could potentially agree to the NGDP linked bonds if Greece were to agree to the following.

Cut the pay of everyone in the central government and central gov pensions by 25 to 35 % and to replace it, give them all the same deal that is currently being offered to the creditors. The same calculation that replaces an EUR 100 loan replaces an EUR 100 salary. This removes the obvious one way fudge as it would hurt their salaries personally. If perquisites are a large part of someone's package, cut and replace a larger percentage of the salary.

I think this solution would keep everyone honest and precisely why greece may not agree to the same.

[JKN]

Nick,
There is really no such thing as NGDP -- it's an abstraction. NGDP calculation could become as complex as the tax code, given the right (NGDP bond) incentives. I love NGDP bonds, but there are as many truck-sized loopholes in NGDP calculations as US federal tax calculations, because NGDP is entirely conceptual.

It's not the best exercise as a result. You could probably run a small-scale prediction market (not price-setting, clearly) a la Sumner, but you could never do big money and NGDP bonds.

[acarraro]

The Italian public pension system is indexed on Italian NGDP already (even if you cannot trade it, the incentive exists already). I don't see much difference between NGDP and CPI calculations in terms of reliability and complexity. They are both extremely difficult calculations based on assumptions and empirical data... Inflation bonds are pretty liquid and widely traded. Argentina has cheated on both measures. If anything the risk is lower since low inflation is generally seen as good news, while low growth isn't (so the government would pay a bigger PR cost in cheating)...

Overall I think the risk is low to non-existent...

[Nick Rowe]

Prakash: Aha! That's like paying managers in stocks/shares.

But that is very similar to acarraro's "The Italian public pension system is indexed on Italian NGDP already"

[bill woolsey]

The rule should be that if here is a change in calculating of GDP, the old definition applies to bonds issued under the old definition.

This doesn't help when the the government changes what it does to impact measured government output. For must of us, nothing we do can impact measured GDP significantly. I have ngdp debts so I marry my housemaid to lower NGDP and cut my debt payments. Not enough impact to be worthwhile.

The government, on the other hand, might be able to do this.

Now, if the government can subvert the office that measures NGDP, then they don't have to do anything. Just tell the officials to report that nominal GDP in Greece is zero (well, 10% lower?)

[Nick Rowe]

bill: yep. The government might subsidise marriage to reduce NGDP. Or provide a subsidy for kids to live with their parents, or parents to live with their kids and grandkids.

[anon]

There is an interesting interaction with NGDP targeting here. NGDPT is often advocated for large currency areas, where subregions may have an incentive to under- or overstate local NGDP statistics in order to influence policy in their preferred direction. For instance, in the Eurozone case, Germany might be better off under a tighter policy, whereas Spain would want easier money. How do you keep local statistics offices from misreporting NGDP data in order to do this? Is there a case for switching to a monetary policy target that roughly tracks NGDP, especially in the short- to medium-run, but is more robust and less vulnerable to manipulation? If so, what might such a target look like?

[Frank Restly]

Nick,

Presumably payments on NGDP bonds are made from tax revenue? Presumably Greece tries to maintain a constant ratio of tax revenue to NGDP? And so do they really help themselves by lying about NGDP or are they keeping two sets of books - one for tax collections and one for NGDP contract payments?

If a government provides a subsidy for kids to live with their parents won't this reduce both NGDP and tax revenue?

[baconbacon]

Nick-

What about backloading government contracts for past the expiration date of bonds? Say your NGDP bonds expire in 5 years, your defense department contracts to buy 10 tanks- 9 at a cut rate to be deliverd before the 5 year period is up and number 10 to be delivered at 5 years and 1 day which will cost 5X as much as the first 9.

[Nick Rowe]

anon: Hmm, that's a new twist.

Frank: yes, but their incentive to increase tax revenue is less than it otherwise would be.

bacon: Time-shifting NGDP! Devious!

[Mark]

Countries already have inflation linked debt. And are generally trusted to not cheat the statistics. Even if they are suspected of cooking the numbers it's just another element of risk to be priced

[Peter Fulcher]

Argentina in its 2005 restructure issued some 'GDP warrants'. Subsequently the Argentina government regularly overstated GDP, thus paying more on the warrants than necessary. Matt Levine (Bloomberg) recounts the story here http://www.bloombergview.com/articles/2014-03-28/argentina-saves-some-money-by-embracing-gdp-reality
PF