Armington Elasticities for Belgium [ Working Paper 12-24 - 03/10/2024]

Information
Classification

This paper estimates the Armington elasticity for the Belgian economy for different levels of (dis)aggregation. At a highly disaggregated level covering about 4,500 HS products, the estimated value of the elasticity ranges from 1.12 to 70.69. Across CPA product groups, the value of the elasticity ranges from 2.40 to 16.63. The corresponding aggregate elasticity ranges from 1.98 to 2.48.

PDF & Download

Publication (en - 2.79 Mb)

Authors

Beni Kouevi-Gath (A)

A : Author, C : Contributor

Publication type

Working Papers

The Working Paper presents a study or analysis conducted by the Federal Planning Bureau on its own initiative.

Executive summary

The Armington elasticity, or the response of import flows to a shock to international prices plays a crucial role in open macroeconomic and international trade models, especially for a small open economy such as Belgium. In fact, the predictions of those models depend heavily on the value of the parameter. And yet, the value of the Armington elasticity is still highly debated and empirical studies offer a range of values that is too large to help settling the issue. Moreover, the values taken by this parameter to replicate business cycles features in macroeconomic models tend to be too low compared to those from micro-econometric studies. As a result, it is challenging to calibrate macroeconomic models by relying on the existing literature.

The goal of this study is to estimate Armington elasticities that could be used to calibrate a one-sector as well as a multi-sector open macroeconomic models of the Belgian economy. To this end, it first estimates the elasticity across the highly disaggregated 6-digit products of the harmonized system (HS). It then aggregates the data for classification of products by activity (CPA) product groups and performs estimations at that level as well. The resulting estimates can be used to calibrate a multi-sector model of the Belgian economy such as DynEMItE[1]. For a one sector-model such as the Belgian version of QUEST III R&D[2], the elasticities estimated across CPA product groups are aggregated to obtain a unique aggregate elasticity.

The estimation of the Armington elasticity is confronted with at least two endogeneity issues. The first is the simultaneity bias which comes from the simultaneous determination of both prices and quantities. The second is due to measurement errors as prices are not observable in trade datasets and are proxied with unit values. To tackle these issues, we use the framework developed by Feenstra (1994). The latter deals with endogeneity by estimating a simultaneous system of equations containing both the supply and the demand of imports. In addition, it makes use of the panel structure of trade datasets to construct internal instruments.

The methodology is applied to data from the BACI dataset to estimate Armington elasticities for different levels of (dis)aggregation. Results point to a strong heterogeneity in the estimates, not only across HS 6-digit products, but also across CPA product groups. For the former, the estimates range from 1.12 to 70.69. For the latter, the estimated elasticities take values between 2.40 to 16.63. The results suggest that the calibration of macroeconomic models should account for this heterogeneity across product groups. They also suggest that in order to limit the effect of heterogeneity bias in a one-sector model, it is desirable to perform estimates with disaggregated data before aggregating them properly. Doing so implies an aggregate Armington elasticity that lies between 1.98 and 2.48 for a one-sector model.

[1] See Verwerft (2022)

[2] See Roeger, Varga, and in 't Veld (2008)