This study reconstitutes the DPRK trade statistics of 1990-2008 on the basis of the mirror data reported by the country's trading partners. We collect all available mirror data for the DPRK trade since 1990, which amounts to about 300 thousands in HS 6 digit, examining and assessing the possibility of errors and fallacies in each data and then rearranging them according to their degree of reliability. We present those examined and rearranged data as "the reconstructed DPRK trade statistics".
For this purpose this study has carried out several works from identifying the basic problems of the mirror data for the DPRK trade to establishing an appropriate methodology to assess the reliability of each mirror data. And the conclusions of the works can be summarised as follows.
First, the mirror data reported by DPRK's trading partners have suffered a variety of problems, which could mislead the country's real trade. They might omit the real trade as well as report the wrong trade the country have not actually done. Also they have technical problems, including statistical break before and after the late 1990s and the different reporting scheme for the North-South Korean trade. All these problems raise concerns about the reliability of exiting mirror data Second, to resolve these reliability concerns, we construct "the reference commodity set" composed of those commodities that we can be sure the DPRK have actually traded, as well as "the reference trade volume of each commodity" in the reference commodity set. For the reference commodity set are used the commodities reported in the North-South Korean trade, the DPRK-China trade and the DPRK-Japan trade, all of which could be hardly mis-reported. And the reference trade volume could is determined as the maximum trade volume of each commodity among the North-South Korean trade, the DPRK-China trade and the DPRK-Japan trade. Third, using the reference commodity set and trade volume, we classify all the existing mirror data for the DPRK trade by four categories according to their reliability. 1) Class 1: trade of commodities included in the reference commodity set, and its trade volume does not exceed the reference trade volume, 2) Class 2: trade of commodities included in the reference commodity set, but its trade volume exceeds the reference trade volume, 3) Class 3: trade of commodities that are not included in the reference commodity set, 4) Class 4: trade of commodities classified as HS 999999 etc. Fourth, a variety econometric tests show that those classification from class 1 to class 4 are reasonable and thus it would be desirable to use mainly the class 1 data, which are relatively most reliable, when we examine the DPRK trade in the real world.