I’m working on a post about the Autor et al paper from earlier this year regarding the impact of trade with China on US manufacturing, and in doing so I’ve been playing around with some trade data from the US Census Bureau. Here I’d just like to put up some of what I’ve found.
First, just the basics:
The country specific data that I’m using only go back to 1987, unfortunately. Looking at trade overall we see that while the US has consistently run a deficit, it began to balloon around 1997 until 2006 or so. After the global recession (which really had a big impact on trade!) the deficit has remained stable at a level slightly less than the previous peak.
Looking at the evolution of our trade partners over time, here’s what happened to our top five trade partners plus China from 1987 until today.
Here we can see the remarkable rise of China in context, and it really is remarkable. Note the acceleration after China joined the WTO at the end of 2001. It’s hard to be sure yet, but it also looks like the curve reached another inflection point around 2006 or 07, and since then Chinese imports have been increasing at a decreasing rate. Also apparent is the steady rise of trade with Mexico, which also looks to have taken off after NAFTA in 1994. Taiwan’s imports have basically remained steady the past 30 years, as have Japan’s.
Looking in terms of percent of total imports, the patterns of Chinese and Japanese trade are almost exact inverses of one another. Just before the beginning of Japan’s first lost decade, they provided nearly the same share of US imports as China does today. I was just a toddler during the 80s, but from what I understand there was a real fear that Japan was going to overtake our economy, much as people fear China’s rise today. I tend not to worry too much about China overtaking us economically for a number of reasons, and seeing Japan stall out and stagnate for essentially 20 years is one of them. On the other hand, China does have an order of magnitude greater population, and they’ve got quite a ways to go before getting to our level of per capita GDP. I’m sure Mike has a word or two about that.
Finally, I wanted to see if the nature of trade has changed over time. In industrial organization, often something called the Herfindahl Index, or HHI, is used in measuring the level of concentration in an industry. It ranges from 0 to 1: values closer to 0 indicates a large number of small firms, and values closer to 1 indicate a small number of large firms. So higher levels suggest a decrease in competition. I constructed an HHI type index for all of the countries that trade with the US using their ‘market share’ of total US imports and exports for each year in the dataset.
Again, higher values indicate higher concentration: more trade with fewer countries. Now, overall the range of values is pretty low and pretty similar across imports and exports. As a final quick check, I wanted to see if these variations were simply due to the value of the dollar relative to other currencies. When the dollar is strong, our exports are less competitive and so the trade HHI should be higher in those times due to reduced trade, and vice versa for imports.
As quick regression I ran
hhi = year + twdi + nafta + wto
where hhi is the trade HHI *100^2 (basically to make things easier to see), twdi is a Trade Weighted Dollar Index, and nafta and wto are dummy variables for the years in which NAFTA and China’s membership in the WTO are in effect.
lm(formula = im ~ Year + twdi + nafta + wto, data = twdi) Residuals: Min 1Q Median 3Q Max -61.324 -22.092 -9.477 22.190 98.867 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -7356.091 5114.772 -1.438 0.16329 Year 4.212 2.550 1.652 0.11161 twdi -1.473 1.159 -1.271 0.21579 nafta -89.175 30.399 -2.933 0.00726 ** wto -70.701 32.236 -2.193 0.03821 * --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 40.82 on 24 degrees of freedom Multiple R-squared: 0.5255, Adjusted R-squared: 0.4464 F-statistic: 6.644 on 4 and 24 DF, p-value: 0.0009525
lm(formula = ex ~ Year + twdi + nafta + wto, data = twdi)Residuals: Min 1Q Median 3Q Max -77.265 -20.362 -3.355 15.429 53.024Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 10419.4969 4071.2267 2.559 0.0172 * Year -5.1497 2.0296 -2.537 0.0181 * twdi 6.7052 0.9224 7.269 1.65e-07 *** nafta 63.1336 24.1972 2.609 0.0154 * wto 21.9839 25.6586 0.857 0.4000 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1Residual standard error: 32.49 on 24 degrees of freedom Multiple R-squared: 0.8132, Adjusted R-squared: 0.7821 F-statistic: 26.13 on 4 and 24 DF, p-value: 1.936e-08
As expected, exports are highly correlated to the TWDI, and NAFTA seems to significant to both imports and exports, but since this is time series data for a relatively short period of time, I’d take that with a grain of salt. It’s interesting that the TWDI doesn’t seem to affect the import index; I’ll have to think of why that would be.
That’s all for now, though I could spend days poring over this. What I really wish we had was more historical trade data, as pretty much the entire time period covered by this Census Bureau data is from the era of liberalized trade. To even go back another couple decades would be really interesting.