The author's first job in economic research was the nightmarish task (so explained my boss) of finding a way of measuring corruption on behalf of a country client in the Middle East.
It was explained that I should start by researching something simple - like exports of German cars to the country under investigation, as listed in German export documents. And compare the results with imports of German cars as documented by the Middle Eastern country.
The results were spectacular. Over 10 times more cars were exported from Germany according to German data, than were supposedly imported according to Middle Eastern data, over the same period.
Something wrong somewhere! The project was hurriedly shelved.
This exercise took place over 45 years ago, and such data might be expected to have changed for the better. But recent evidence suggests that raw trade data may be even more misleading today.
A paper by two academics at the University of Amsterdam (“Globalisation and the growing defects of international economic statistics
", by Lucas Linsi and Daniel K. Mugge, published in the Review of International Political Economy, in January 2019) suggests that American authorities claim that US imports from France exceed US exports to France by some $14bn, while French sources indicate the difference to be less than $4bn. OK, not a factor of 10 as in the ancient example above, but a massive discrepancy, and one between the data produced by two developed countries that spend large amounts on producing economic data, not data differences between one developed country and one with poor quality national statistics.
Linsi and Mugge cite other examples. They show that major differences of this sort are not cherry picked outliers. They are the rule rather than the