When the Wall Street Journal reported last month that the White House was mulling tweaks to American trade statistics, some economists laughed — others gasped. There are two key takeaways: 1) trade statistics are important, for lots of reasons, and should be protected from political interference; but 2) we also have to acknowledge that they are much less tidy than is often assumed.
Here’s the story. Allegedly, the White House was pushing U.S. government statistical offices to boost the U.S. trade deficit with its main trading partners — on paper. The idea was to tweak the trade stats by registering the value of imported goods destined for eventual re-export only as imports (when they enter the United States), but not counting these same goods as exports when they are moved on to a third country.
In principle, excluding re-exports from trade statistics altogether is not an unreasonable idea. In fact, the World Trade Organization (WTO) and the Organization for Economic Cooperation and Development (OECD) are working hard to compile data that counts only value added in each country to avoid an artificial inflation of trade statistics.
That approach is completely different, however, from fully counting goods when they enter the United States but ignoring those that are being re-exported elsewhere. That makes no sense.
U.S. trade statistics serve a political goal
In other words, President Trump apparently was ready to bend statistics deliberately to persuade Congress that free trade in North America was a disaster for the United States — a position that resurfaced last week when the president mulled ignoring WTO decisions that didn’t suit him.
[interstitial_link url=”https://www.washingtonpost.com/news/monkey-cage/wp/2017/02/15/what-is-nafta-and-what-would-happen-to-u-s-trade-without-it/?utm_term=.1ffa51ea7f1d”]What is NAFTA and what would happen to U.S. trade without it?[/interstitial_link]
Distorting figures to draw a custom-tailored picture of the economy contravenes everything that official statistics should stand for, especially their impartiality. In this sense, criticisms of the alleged White House suggestions are entirely justified.
But don’t assume the import/export data we have is accurate
Critics miss a less flattering truth about trade statistics, however: Import and export data are much more messy than their champions suggest and millions of users assume.
In principle, all trade flows are recorded twice. The exporter counts when goods leave the country and the importer counts the goods when they arrive at their destination. As mirror statistics, Mexico’s exports to the United States should not differ very much from U.S. imports from Mexico. Here’s the trouble — the gap in these figures is huge, and it has been growing, rather than shrinking.
Take merchandise trade with Mexico, roughly 90 percent of total trade across the United States’ southern border, and much easier to measure than trade in services. According to the U.S. Census Bureau, the 2015 U.S. trade deficit with Mexico was $63.4 billion. The Mexican government, in contrast, put the figure roughly twice as high, at $122.1 billion (the import figures are here, the export ones here).
That’s puzzling, to say the least. To get a clearer picture, we dug into the authoritative OECD STAN Bilateral Trade in Goods Database. From there, we traced the discrepancy in U.S. and Mexican trade figures over time, and compared that with similar kinds of data mismatches between other country pairs.
As Figure 1 illustrates, gaps between Mexican (blue bars) and U.S. estimates (red bars) of the bilateral merchandise trade are not unique to 2015. They go back at least two decades but diverge more sharply in recent years. Since 2013, the difference between the two estimates has become so large that the discrepancy corresponds to more than 10 percent of the total value of all trade between the two countries.
Figure 1: Mexican trade estimates (blue bars) and U.S. estimates (red bars) have often diverged, but the gap has widened in recent years.
Such discrepancies are not limited to the U.S. trade relationship with Mexico. For instance, the U.S. trade deficit with China in 2015 amounted to $388 billion according to U.S. statistics, but only $260 billion according to Chinese data. And the U.S. deficit with Germany, estimated at $78 billion by U.S. agencies, is closer to $58 billion according to German sources. The U.S. government tabulates the U.S. trade deficit with Canada as $21 billion, but Canadian data reports the deficit to be almost four times as large — $90 billion.
[interstitial_link url=”https://www.washingtonpost.com/news/monkey-cage/wp/2017/03/07/will-trumps-unilateral-trade-approach-work-history-says-no/”]Will Trump’s unilateral trade approach work? History says no.[/interstitial_link]
Why so many data mismatches? It’s not a question of faulty U.S. data per se — we find similar discrepancies in other bilateral trade reporting as well. Looking at discrepancies between country-pairs in absolute numbers (rather than the surplus/deficit of any specific country) for five large economies reveals that the differences are significant in almost all relationships (see Figure2).
Figure 2: Discrepancies in OECD STAN bilateral merchandise trade mirror statistics as a percentage of totals reported by importing economy, average values 1990-2011. The graph indicates, for example, that the difference between reported annual merchandise exports from the U.S. to Germany and the imports from the U.S. reported by Germany on average amounted to 25.25 percent of the total value of German imports from the U.S. during 1990-2011. But it’s not just a U.S. data issue alone — the other four countries listed here reported similar discrepancies in trade data with each other. DATA SOURCE: OECD STAN bilateral trade database rev. 3.
So where do these discrepancies come from, and what do they mean for users of trade data?
The most obvious explanation would be that the valuation of imported goods includes customs, insurance and freight charges, while exports don’t. But this accounts for a small part of the story — in fact, contrary to this logic, the discrepancies are in many cases driven by higher reported values of exports compared with corresponding imports.
We also don’t see any clear signs of deliberate tampering with the aggregate data. Rather, the problem is the immense (and rarely acknowledged) complexity of data collection behind these figures — one that grows worse as production itself becomes more and more globalized.
Trade flow estimates build on billions of custom forms (many collected manually), complemented by uncountable household and enterprise surveys.
Information can get lost, such as when goods cross borders undeclared. Companies may be tempted to lie on the forms to circumvent capital controls or shift profits around. (Smart transfer pricing allows multinationals to adjust the price tags when they move goods across borders for further processing inside the company and thereby declare more income in low-tax jurisdictions). Currency conversion can draw on different exchange rate data. And despite all the harmonization efforts from international organizations, national agencies remain free to collect and aggregate data as they prefer.
What is the U.S. merchandise trade deficit with Mexico? Chances are it’s somewhere between the figures both countries report, but the honest answer would be that we don’t really know. Offering a point estimate — a single hard number — feigns a degree of accuracy that does not match real-world uncertainties. Given these structural problems and huge gaps in reported data, it is misleading to suggest that official trade statistics capture an “accurate” picture of imports and exports.
Of course, we need economic statistics — and need to collect them as accurately as possible. And we should defend economic statistics against overt attempts to twist them for political purposes. But any mature discussion about economic policy needs to acknowledge the huge margin of error in the numbers we have. Anyone stylizing present-day statistics as “the economic truth” can only lose the argument in the end.
Lukas Linsi is a postdoctoral research fellow at the University of Amsterdam.
Daniel Mügge is professor of political arithmetic at the University of Amsterdam and directs the FickleFormulas project.