The procedure used for testing the validity of gold statistics is described in some detail so as to make easily possible a test for other countries and periods. At the outset it must be recognized that our conclusions are, to some extent, a matter of individual judgment, and our opinions may of course be rejected by others. It would be hard to choose between such conflicts in opinion because--as will be seen--statistical problems are involved for which there exists at present no statistical theory which one would wish to fall back upon. The entire case is of value, however, for all students of economics, for it shows in an important field what precautions must be taken before establishing theories or "Verifying" them.
The problem may best be broken down into four parts: (1) the formulation of the underlying hypothesis, (2) the description of the sample and the results, (3) the explanation of the failure of the statistics to confirm the hypothesis and a discussion of certain difficulties even if it had been confirmed, and (4) the consequences of (2) and (3).
The hypothesis is rather easily formulated. The observation of gold movements from country A to country B are made in A's export and in B's import statistics. The two should coincide, allowing for such factors as costs of transportation and time lags. In neighboring countries both virtually disappear; in countries more distant from each other, say Europe and Australia, a time lag may have to be considered. Costs of transportation, insurance, handling charges, and so on, which all enter significantly in the determination of gold points, play no appreciable role here. Gold is so valuable that these costs all but disappear from the statistics, as is evident from the proportion they make with the value of gold shipments.3 In view of the rounding off of numbers in foreign trade statistics, the factor of distance (costs of transportation)-- which plays a great role in determining the value of less valuable____________________