Exchange rate policy and trade balance: a cointegration analysis of the Argentine experience since 1962

Abstract

Using multivariate cointegration tests for nonstationary data and vector error correction models, this article examines the determinants of trade balance (TB) for Argentina over the last forty to fifty years taking into account that the short-run impacts of currency depreciation on the TB behaviour may differ from the long-run effects. Our investigation confirms the existence of long-run relationships among TB, real exchange rate (RER) and foreign and domestic incomes for Argentina during different RER management policies. Based on the estimations, the Marshall-Lerner condition is checked and, by means of impulse response functions, we trace the effect of a one-time shock to the RER on the TB not finding support for a J-curve pattern in the short-run.

Notes

¹Despite the USA has been the major trading partner for Argentina in the analysed period (for instance, in 1980 USA represent 31% of Argentinian exports and imports, in 2000 this percentage was 30.5%, only Brazil, during the nineties has been more important in the Argentinian TB), we have extended the same analysis by considering as ‘foreign income’ the World GDP volume index from the data statistics provided by The World Bank. The long- and the short-run analysis yield estimations which are lining up with these ones reported in this article. These estimations can be obtained from the authors.

²The computed test statistics for the unit root null are greater than the MacKinnon (1996) critical values though at the 1% level the ADF test yields really close to the corresponding tabulated value. The ADF result coincides with the usual persistence of the variable and, therefore, the stationary consideration of the TB followed in the related literature. See Nadenichek (2006) and Kim (1996) not only for a clear and brief discussion but also for examples on the stationary treatment of the TB.

³Johansen and Juselius (1990) suggest that, in case that the statistics yield conflicting results, the maximum eigenvalue test may be better.

⁴In fact, we have tested different periods and ML condition begins to be satisfied after 1994.

⁵The stability of the estimated VAR and VEC has been checked in each of the sub-periods from the inverse kp roots of the characteristic AR polynomial, where k is the number of endogenous variables and p is the largest one. The stability condition is verified if all the roots have modulus less than one. In a VEC estimation with r cointegration relations this implies that kr roots should be equal to unity.

⁶Although, we note that the impulse response functions for the sub-periods 1962–2000 and 1978–2005 show a similar behaviour.