Recently the San Francisco Fed released a new study entitled “The U.S. Content of “Made in China””. The study argues that “[g]oods and services from China accounted for only 2.7% of U.S. personal consumption expenditures in 2010.” This figure was cited approvingly by a large number of publications and bloggers, including the LA Times, the WSJ, Fortune, The Street.com, and Matt Yglesias, who writes:
When Americans go buy stuff, they’re overwhelmingly buying things that are made in America:
Sorry, Matt. I’m going to explain why the SF Fed study shouldn’t be taken seriously. In fact, the study has two main flaws:
- The authors did not distinguish between dollar shares and quantity shares of imports. When imported goods are much cheaper than domestic goods, then the quantity share can be much larger than the dollar share.
- The input-output tables used by the authors contain no actual information about how much of Chinese imports are going to personal consumption. In fact, all imports are divided among sectors by a simple rule known as the “proportionality assumption.” So in reality, Chinese imports could constitute much more of PCE–or much less–than the SF Fed economists calculate.
Dollar Shares versus Quantity Shares
So first let me explain the difference between dollar shares and quantity shares. I’ve just finished the second edition revision of my intro economics text, Economics:The Basics. When explaining the basic concepts of supply and demand to students, it’s always important to clearly differentiate between quantity (as measured in physical units) and cost (as measured in dollars). The same cost can correspond to higher or lower quantities, depending on the price.
The same distinction applies to Chinese imports. It is clearly true that Chinese imports are priced lower per unit than the domestic-made products that they replace. Similarly, China-made imports are much cheaper than the Japan-made or European-made imports that they replace–that’s why U.S. retailers changed their sourcing over the past ten years.
As a result, if we measure the share of Chinese-made products in PCE, our answer is going to be much different if we calculate the dollar share, versus calculating the quantity share. An example will make this clear. Suppose that a U.S. shirt factory sells 100 shirts at $50 a piece, for a total cost of $5000. Now suppose a Chinese manufacturer comes into the market and offers to sell an identical shirt for $5 a piece. In the first year, the Chinese manufacturer sells 50 shirts and the American manufacturer sells 60 shirts. What share of the market do the Chinese shirts have?
Measured in dollars, the Chinese have 7.7% of the market ($250/($250+$3000)).
Measured in quantity of shirts, the Chinese have 45% of the market (50/110)
Which share is right? For sizing the impact of imports on U.S. jobs and manufacturing, the quantity share is much more relevant than the dollar share.
In fact, it’s very easy to construct examples where the dollar share of imports goes down, but the quantity share goes up. If China offered its imports to the U.S. for a near-zero price, then China’s dollar share of the U.S. would be close to zero (assuming that there was some U.S. manufacturing left) but the quantity share would be close to 100%.
The authors of the SF study are calculating the dollar share, not the quantity share. That’s why their number seems so low.
In order to calculate the quantity share, we need to know the relative price of Chinese imports compared to equivalent U.S. products. It would also be useful to be able to compare the price of Chinese imports with imports from other countries. (See a recent article in the Journal of Economic Perspectives, Offshoring Bias in U.S. Manufacturing). It makes an enormous difference whether Chinese made imports are 5% cheaper than the equivalent U.S. products, or 50% cheaper.
However, the Bureau of Labor Statistics does not collect such relative price data across countries. At no point does the BLS measure the difference in price between a shirt made in China versus one made in Italy or the U.S. In fact, when the sourcing of a particular good changes from one country to another, the import price index often treats it as a new product, even if it is functionally identical. (Take a look at the BLS explanation of its import price methodology here).
The second problem with the study is that the government statisticians have no information–repeat no information–about whether an imported goods or service is going to consumption, to capital spending, or being used as an intermediate input. How could they? That question is never asked on any economic survey form.
Here’s the description of the problem from the official BEA ‘bible’, Concepts and Methods of the U.S. Input-Output Accounts
Unfortunately, data on the use of imports by industries and final uses are not available from our statistical data sources. Thus, to develop an import matrix, we make the assumption that imports are used in the same proportion across all industries and final uses.
This is what’s known as the “proportionality” assumption. The proportionality assumption is *not* harmless, especially when it comes to calculating the contribution of a single country, such as China, to PCE (see for example the paper here). It might very well be that Chinese imports are much more concentrated in PCE than the proportionality assumption suggests, especially in areas such as computers where the Chinese are more likely to have a low-end product (Best Buy does not sell U.S.-made supercomputers, do they?) Or Chinese imports could go much more into investment goods than anyone realizes. The point is that there is no information to make a judgement.
So the calculations of the SF Fed economists are fundamentally based on a huge assumption which may or may not be true. At a minimum, they should have offered up their calculation as a range.
BTW, if the SF Fed economists still are prepared to defend their calculations, I’m happy to debate them in any forum.