The GDP Deflator and the Inflation Rate

There is confusion between the GDP deflator and other measures of prices such as the CPI and the PCE deflator. Here's one way to think about it that might help to clear things up.

The CPI (or the PCE) attempts to measure how the prices of a typical market basket of goods changes over time. The idea is to measure the impact of price changes on the consumption bundle of the average household.

However, that's not what the GDP deflator measures. Just as we can think of a typical bundle of goods that a household consumes, we can also think about the price of a unit of GDP. GDP is composed of four elements, consumption, investment, government spending, and net exports. For example, real GDP might be 600 C's, 200 I's, 150 G's and 50 NX's. If so, we can think of a unit of GDP as .6 units of the consumption good, C, .2 units of the investment good, .15 units of the investment good, I, and .05 units of net exports, NX.

Imagine, then, going to the store and purchasing a unit of GDP off the shelf. That package, the unit of GDP you are purchasing, would come in these proportions, and the GDP deflator is the price that you would have to pay for this unit of GDP. But there is no reason to think that this unit of GDP will be the same as a unit of the market basket consumed by a typical household. To start with, investment goods are not part of the household's consumption bundle. Secondly, GDP is what we produce domestically, within our borders. Some of what we produce is traded for goods produced outside our borders. The GDP deflator will not reflect this, but the CPI will since the imported goods would be in the market basket used to track prices over time.

The point is that the basket of goods tracked by the GDP deflator, which is a unit of GDP, is not the same as the typical basket of goods consumed by households (which is dominated by the C component of GDP). The GDP deflator has a very specific purpose, and it's name tells you exactly what that purpose is. It should be used to deflate nominal GDP to obtain real GDP. It is not a measure of household inflation, nor is it intended to be, and using to measure the rate of inflation rate faced by households is not appropriate.

Here's David Altig with another way of looking at this:

Does the GDP deflator lie?, macroblog: Though last week’s report on U.S. gross domestic product (GDP) growth in the second quarter is second-hand news by now, I’ve taken note that Barry Ritholtz’s views on the news has, in particular, continued to rumble through the blogosphere. Barry is not happy with the GDP deflator, and samples approvingly from a Barron’s article by Aaron Abelson:

“GDP, in common parlance, stands for stands for gross domestic product, or the aggregate value of all the goods and services produced on these blessed shores... These days, alas, those initials more typically signify “gross deceptive pap”...

“Comes now the so-called preliminary estimate that claims second-quarter GDP grew by a much more robust 3.3%.

“The key here is the GDP deflator, which purports to adjust GDP for the impact of inflation; it’s a curious calculation in that, contrary to its moniker, it seems designed to do the exact opposite of deflating GDP.

“Thus, according to this accommodating measure (accommodating, that is, if you’re determined to put a good face on a dreary report), inflation grew at an improbably restrained 1.33% in April-June. And maybe it did—but not in the good old U.S. of A. However, obviously more important than accuracy to those doing the calculating is this simple equation: The lower the deflator, the greater the growth of GDP…

“Of course, even by the government’s not entirely extravagant figuring, the consumer price index was up a hefty 8% in the latest quarter. Perhaps the computer that tallies the CPI doesn’t talk to the computer that measures the deflator.”

Strong words, but if you ask me, misguided. Barry actually makes the case against the case in this picture, about which he notes:

“It’s no coincidence that the current situation resembles past ones where oil prices had spiked. Since more than half of the U.S. Crude consumption is imported, the price and quantity go into all GDP calculations as a negative.”

Exactly. Let me provide an elaboration of the spot-on point made at The visible hand in economics blog. For the sake of argument assume that every drop of oil consumed in the United States is imported, and everything imported to the United States is oil. If we leave exports out of the picture for simplicity, we can think of U.S. consumption as consisting of GDP—everything produced in the United States—and imported oil.

Suppose, then, that the price of oil rises precipitously. If both incomes and oil consumption are relatively fixed in the short-run, what would we expect to happen? The answer is more expenditure on imported oil and less spending on everything else. As the demand for domestically produced goods and services falls, so would their prices. (Or more generally, they would rise at a slower than normal pace.) Since domestically produced goods and services by definition constitute GDP, GDP-deflator inflation will be low, while the consumer price index (which would include nonexported GDP plus imports) could well be quite high.

Voila! A simple Econ-101 explanation, with nary an insult hurled at the good folks from the Bureau of Economic Analysis.

That said, there are plenty of reasons to be cautious in interpreting last week’s report. Mark Thoma has a fine roundup of many fine points by many fine bloggers. To that list I’d add comments by Spencer at Angry Bear, William Polley, Lim at The Skeptical Speculator, Ben Leeson at Working Thoughts, Zubin Jelveh and Felix Salmon (both at Portfolio.com), to name a few. But I would delete the suspicion that low GDP-deflator-based inflation suggests shenanigans are afoot.

Comments

[Edd]

I just want to make a question. If it´s a mistake to measure inflation with the GDP deflator, would it be considered a mistake to deflate GDP with the CPI in order to obtain real GDP?

[JKH]

Thank you, thank you, and thank you – so much disinformation on this topic out there right now.

A key point you make, I think:

“It is not a measure of household inflation, NOR IS IT INTENDED TO BE, and using it to measure the rate of inflation rate faced by households is not appropriate.”

(My emphasis – some think it should and must be a measure of household inflation. They then reject its usefulness and validity when it’s not.)

One suggestion – the following has the potential I think to be a touch misleading or confusing:

“Imagine, then, going to the store and purchasing a unit of GDP off the shelf.”

In order to simulate a truer physical microcosm, the store needs to be a global distributor, and the consumer needs to be a global traveller of somewhat split personality in order to purchase the export content of a GDP unit.

Along with the exclusion of import inflation, the fact that the deflator includes export inflation also tends to be overlooked, thereby complicating a comparison with CPI even further.

Finally, some who would agree with your explanation unfortunately go to the extreme point of denying that the GDP deflator is a measure of inflation. This is false. It is a measure of inflation. It just isn’t a measure of domestic household inflation, as you point out above. Its inflation components are what they are, including inflation in respect of the export content of the GDP unit you describe.

[Edd]

I just want to make a question. If it´s a mistake to measure inflation with the GDP deflator, would it be considered a mistake to deflate GDP with the CPI in order to obtain real GDP?

[JKH]

Ed,

a) It’s not a mistake per se to measure inflation with the GDP deflator. It’s only a mistake to use the deflator to measure inflation on some set of goods and services other than what is produced in GDP.

E.g., it would be a mistake to attempt to measure household inflation using the GDP deflator, as noted by Mark Thoma above.

But it’s not a mistake to use the GDP deflator to measure inflation for GDP.

b) Yes, it would be a mistake to deflate GDP with the CPI. As noted above, the set of goods and services covered by each of these is quite different, so the result would make no sense. CPI is a measure of household inflation. CPI excludes export inflation and the GDP deflator includes it. And CPI includes import inflation reflected in domestically purchased goods and services, whereas the deflator excludes it.

[spencer]

It is a fact of life that nothing is perfect. Every once in a while because of the way we calculate GDP the deflator is distorted.

Live with it, it is not a big deal.

[anon]

Barry Ritholtz and a few others thought it was a huge deal. This helps make it clearer.

[dWj]

I wonder if one potential source of confusion is that the "basket" of which the price is being derived seems, in the usual breakdown, to include negative weights for imports. If you subdivide consumption into consumption of domestically produced products and consumption of imports, the latter should be excluded when deflating the GDP because the latter is excluded from GDP; when we say "C+I+G+NX", we're counting it in the C and backing it out in the NX.

In any case, the GDP deflator is calculated as a derived construct; it's not as though BEA adds up nominal GDP, calculates the deflator, and then divides one by the other, as I think a lot of people are imagining. It calculates nominal GDP and real GDP, adding up nominal units of peanut butter, manhole covers, etc., for the former and real units of peanut butter, manhole covers, etc. for the latter; the deflator is the ratio that can be calculated at the end.

[Brian MacLean]

It doesn't affect your main argument but in the following sentence didn't you mean .6 units of C, .2 of I, .15 of G, and .05 of NX?
"If so, we can think of a unit of GDP as 6 units of the consumption good, C, 2 units of the investment good, 1.5 units of the investment good, I, and one half unit of net exports, NX."

[Mark Thoma]

Yes, that would have been better (and I am going to change it).

I think - from teaching - I reflexively avoid decimals and fraction sin examples and didn't catch the inconsistency (I even worried about 1.5)

Thanks.

[cm]

Another thing to be considered is that due to "hedonic adjustment", the "value" of some GDP components is not actually based on market prices times sales volume, but on imputations. For example, upwards adjustment of computer sales/investments by claiming that computers with faster processors, larger disks, larger displays, or newer software versions are supposedly leading to more productivity in business. Likewise with other goods to which "quality improvements" apply.

If you have such make-believe numbers in your calculation, it means GDP is not an objective "measure", but a subjective "model".

Other major economies are not applying such adjustments, and US GDP is constantly exhorted as being superior.

[a]

".6 units of C, .2 of I, .15 of G, and .05 of NX."

Is it possible to reason:

Deflator = 0.6 * Infl_C + 0.2 * Infl_I + 0.15 * Infl_G + 0.05 * Infl_NX,

where Infl_C *is* roughly the CPI. If so, what are the values (roughly) of the other inflations (just to get a sense of the magnitudes)? Is the problem that NX should be broken down into two components (Exports and Imports) with vastly different inflations?

[reason]

Is this just another way of saying relative price movements make interpreting real measures of income and consumption problematic? (One of my pet themes anyway).

[JKH]

a,

C inflation should correlate reasonably well with CPI, although the bases and measurement types are still different - "real time" GDP component versus basket.

I agree that exports and imports should be broken out. Net exports is a good measure for showing the intersection of GDP with the foreign sector. But it is a bad one for showing the composition of GDP across all sectors; i.e. imports are distributed as sector specific content across each of C, I, G, and E.

And I agree that sector specific inflations are very instructive. Here is a very useful comment from the Macroblog post on this point:

“Is it not simple quantitative analysis? The import deflator of -4.56% overwhelmed the PCE, Investment, Export and Government deflators of 2.93%, 0.11%, 1.37% and 0.98%, respectively.

BEA Table 1.1.8

Posted by: marmico | September 03, 2008 at 08:40 AM”

[a]

JKH - thanks.

[dblwyo]

Mark - thanks. Quite helpful. This really ripped thru the blogosphere and the discussions quite heated; unfortunately the conspiracy theorists likely triumphed in the end. Menzie China also took a pretty good pass, coupled with a very nice discussion of other factors in his "Why It Feels Like a Recession" post.
When you tunnel down the various components to looking at oil import prices they results were a little perverse. However there are two very straightforward fixes here. First look at Gross Domestic Purchases, essentially GDP net of trade. Second, to answer the question of economic trend, direction and turning points look at YoY changes. I believe you folks who actually understand this stuff look at log differences for the same result ?
Anyway real GDP growth this way 2.5% and 2.2% - so much for the surge meme :) Net of trade though, i.e. what's the domestic economy doing, it was 1.1% and 0.4% !!!
Don't know if you allow outside URL reference but the charts poking at all of this are up on my site.

[me]

Great job Mark. Where was this when dblwyo and I were getting killed over at Barry's Big Picture?

[anon]

I didn't think you got killed.

The Big Delusion is still at it today. It would help for starters to know the components of GDP.

[Barry Ritholtz]

My main complaint with the GDP deflator is that it creates these wildly anomalous outputs. That is a very significant flaw in the way our economic output is calculated.

A variety of data points suggest that the economy did not expand at an annualized 3.3% rate in the 2nd Quarter.

I do not think it is too much to ask of the model which produces GDP data to consistently reflect the actual economic output of the country -- even if some of the inputs are volatile.

[VG]

Okay, let's go back to basics.

Oil is the fundamental commodity of industrial civilization. It is used in every single good and service consumed in the United States. If the price of imported oil doubles, then, fundamentally, that must either: A. increase inflation; or B. hurt the general economy. And very likely it will do both.

But this GDP deflator doesn't say that. On the contrary, it says that the huge increase in the price of imported oil has somehow resulted in the lowest inflation in years AND has also increased the size of the general economy by a high rate.

This is impossible. Therefore it isn't true.

Is the GDP number supposed to tell us something about the Economy? Or is it just a bunch of numbers mashed together by an arcane formula?