An Empirical Analysis of the Economic Integration Between Mexico and the United States and Its Connection with Real Exchange Rate Fluctuations (1980–2000)

Abstract

This article provides a new perspective about the links between the Mexican and US economies by studying the behavior of some Macroeconomic variables during the 1980–2000 period. It uses time series techniques to show that the Mexican GDP, its components and even real money balances had a robust long-run relationship with the US economic activity and the bilateral real exchange. The tighter nexuses appear to have begun in the early eighties and not in the nineties, as is often thought. The relationships here found were not modified by the inception of NAFTA or other events.

ACKNOWLEDGMENT

Comments and suggestions of Rafael Gomez‐Tagle, Adalberto Gonzalez and seminar participants at the Bank of Mexico and the Bank of Canada are gratefully acknowledged. Rocio Elizondo was a very capable research assistant. The author is solely responsible of all errors and omissions in the paper, and his views are personal and should not be taken as representing those of Bank of Mexico.

Notes

¹Two papers that refer to this fact are Kamin and Rogers (2000) and Bergoeing et al. (2002).

²Actually, the aggregate that was used as an intermediate target by the Mexican central bank is the monetary base but as the deposits of commercial banks have been zero since the beginning of the nineties, both aggregates are now identical.

³We can consider the real exchange as an I(0) or I(1) variable and apply the Pesaran et al. (2001) test for the existence of a long-run relationship. With either assumption that test yields the result that each one of the long-run relationships we test below is legitimate at the 1% level of significance.

⁴For comparison, the corresponding correlations for Canada and the United States are very high.

⁵We say “the effect of real exchange rate on Mexican GDP” but it might well be the case that such line of causality does not exist and that both variables are simultaneously affected by some particular negative shocks. We continue with the hypothesis that real exchange rate depreciations are contractionary because we are not aware of a final answer to the problem of this correlation but we certainly are open to other possibilities. Whatever the reason behind the sign of that correlation, our statistical results are robust enough although they should not be taken as a prove of causality.

⁶The Johansen-Juselius (1992) approach starts with a VAR where the number of cointegration vectors and constraints in the long-run and short-run parameters are carried out through maximum likelihood tests. The single-equation error correction approach is an earlier method suggested by Banerjee et al. (1993) that we use to simplify the presentation of our results. For the latter method we start with an autorregressive distributed lags specification that is then simplified by using the general-to-specific approach to reach a final form. The results are so robust that the results with either method are basically the same. For the second method we use the critical values tabulated by Ericsson and MacKinnon (1999). As the series involved are nonstationary either method is a proper way to look for long-run relationships.

⁷This is a feature found in the PC-give program.

⁸As was mentioned before, cash is the aggregate that holds the closest and most stable relationship with prices and output in Mexico.

⁹See Garces-Diaz (2001).

¹⁰This a concept from spectral analysis that can be regarded as a measure of correlation between two variables at different frequencies. See Brockwell and Davies (1991) p. 436 for a definition.

¹¹See Albuquerque (2001).

¹²The equations show no signs of nonlinearity. The RESET statistics reported in Tables III, IV, and V can be used as simple tests for nonlinearity, as discussed in Granger and Teräsvirta (1993). The low value of those statistics indicates that a linear specification is appropriate in each case.

¹³This result is robust throughout different samples. We used the periods 1957–2000 and 1980–2000 with similar results. We utilized both the built-in function of S-plus to obtain the square coherency at all frequencies and the program written by Albuquerque (2001) that calculates the long-run correlation at frequency zero with a nonparametric method.

¹⁴It can be shown that the ratio of investment to output has an inverse relationship with the real exchange rate: it decreases (increases) with a depreciation (appreciation).

¹⁵Nonlinearities might be present somewhere else but definitively not in the relation between output and the real exchange rate.

Additional information

Notes on contributors

Daniel Garces-Diaz

Daniel Garces-Diaz is a senior economist at the Mexican central bank. He obtained his Ph.D from UC Berkeley and his research includes topics on international trade, economic integration, exchange determination and monetary economics. He can be contacted at Banco de México, Av. 5 de Mayo #18, 3er. Piso, Mexico D.F. 13500, Mexico. Tel.: [phone]; E-mail: [email protected]