Authors: Prodromos Prodromidis
Title: The tool employed by the EU to estimate the fair import prices of the UK, under scrutiny
Abstract
In recent cases the Court of Justice of the European Union (CJEU) allowed the use of statistical values and databases in order to combat the undervaluation of goods imported into the European Union, and seemingly gives the European Anti-Fraud Office (OLAF) and EU customs authorities an unlimited license to reject the transaction values. The development marks a watershed in global trade and customs from a legal viewpoint. (E.g., Schippers and de Wit, 2023).
The economic implications for consumers and businesses across the EU are considerable. However, the statistical tool proposed by the EU authorities to identify undervalued goods and estimate custom duty losses, and approved by the CJEU (2022) in a case brought up by the EU against the United Kingdom (Case C-213/19 Commission v United Kingdom), is intriguing. The paper looks into the tools mechanics -its assumptions and the way it treats statistical values- and poses a good number of questions.
The tool was developed by the European Commission’s Joint Research Centre (JRC) (Arsenis et all, 2015). Using data taken from COMEXT (a reference database for detailed statistics on international trade, which is managed by Eurostat), OLAF and JRC run univariate linear econometric regressions explaining aggregate monthly import values of individual goods (e.g., eight-digit code textile or footwear items) made in particular country (e.g., China) it terms their respective quantities, for each and every EU member state, over a period of forty-eight months. In the process, OLAF and JRC analysts filter out (exclude) observations associated with unusually high or low values, and re-run the regressions to estimate the so-called cleaned average price per kg (CAP) of each eight-digit code good from made in a particular country imported in each and every EU member-state. Then, using only the CAPs which are associated with high or modest R2s, i.e., using the CAPs of the member states that feature a high or modest model fitness, the analysts calculate an arithmetic average for the entire EU, that is to say, a non-weighted average, of the said CAPs. This is taken to be the EU-wide estimated fair price (FP) of each eight-digit good imported from a particular country.
Next, a value corresponding to 50% of the FP is calculated. It constitutes the lowest acceptable price (LAP). Imports with a value lower than the LAP are viewed as presenting a significant risk of undervaluation, and, so, ought be subject to customs controls before clearance. With a CJEU’s decision (Commission vs UK, C-213/19), this risk-profiling LAP becomes an undervaluation threshold on the basis of which the FP is applied on the imported item and customs duty losses are calculated.
In our view the tool operates with the assumption that the prices of imports do not vary significantly over time. It is an assumption not supported by the facts. Consequently, the tool ignores long-term price trends, seasonal or cyclical patterns, and other temporary price variations over time. In addition, the tool relies on econometric regressions that often employ few or very few observations (e.g., 4-36 observations), against the tool-developer’s own suggestion that only estimates using large number of observations are to be trusted. It also treats time-series data as cross-sectional (it is a paradox); assumes that the relationship between values and volumes is linear, when, in fact, there is good evidence that (a) the relationship is quadratic, and (b) lower priced monthly observations probably ought to be included in the regressions. Instead, the tool removes such observations as if they all result from errors; and removes country observations (full sets of observations) if they seem unfit in the (rather naïve) univariate regression setting it employs. Additional factors suggested by economic theory and common sense, could have been considered in a multivariate regression setting. Their absence along with other shortcomings almost certainly yields flawed CAPs and, hence, flawed FPs and LAPs. Last but not least, the said prices are not correctly tested for precision and reliability.
In view of the above and in our view, each of the above pitfalls or drawbacks on its own is sufficient to disqualify the tool as a means to accurately estimate something so sensitive and crucial as customs duty losses. Yet, it is the tool that was submitted to the CJEU, and based on its assumptions and estimates the withdrawing from the EU, United Kingdom was fined and paid to the EU Commission close to 2.7 billion euro.
Keywords: import undervaluation, transaction value, statistical value, fair price estimation, customs duty losses
JEL classification codes: C20, F10, H87
References
Arsenis S., Perrotta D., Torti F. (2015). The estimation of fair prices of traded goods from outlier-free trade data. European Commission, Joint Research Centre Technical Report # 100018. Ispra.
JJEU (2022). Judgment in Case C-213/19 Commission v United Kingdom. Press release # 42/22 of March 8, 2022. Luxembourg: Court of Justice of the European Union.
Schippers M., de Wit W. (2023). The Use of Statistical Values to Combat Undervaluation in the European Union. Journal of World Trade, 57.2: 253–276.
