A two-step approach to examine the dynamics of market convergence

Abstract

We present an improved approach to examine convergence of markets such as those for equity, bonds or commodities. The approach is motivated by Monte Carlo simulations and consists of two steps. First, we test for regime-shifts in the cointegration paths and cointegration with structural breaks. If equilibrium errors are stationary, we then obtain the degree of convergence by rolling speeds of adjustment in a vector error correction model. Our approach is illustrated by an application on stock market convergence.

Keywords:

• market convergence
• rolling cointegration
• structural breaks
• speed of adjustment

JEL Classification:

• C22
• C51
• G15

Notes

¹ Calculations were performed using gretl. Program code is available from the authors upon request.

² An alternative approach is to estimate only for regimes before and after the break. In this case, the mean estimates are equal to the true parameter

³ All calculations were also repeated using stock indices in home currency. In this case, the Kejriwal and Perron (2010) tests yield an additional significant break in the model GER–UK at 1992/08 which coincides with the withdrawal of the United Kingdom from the European Exchange Rate Mechanism (ERM) in September 1992. The identification of this break in contrast to the model in US dollar provides evidence that currency revaluations influence the portfolio behaviour of international investors. The Kejriwal (2008) test applied to this two-break case does not reject cointegration.

⁴ When exchanging the roles of countries in Equation 2, the Kejriwal and Perron (2010) tests again yield one significant break in each model with a break fraction of about 0.4. Cointegration cannot be rejected in this case either.