Other

European and US capital markets: Which econometric approach is the best fit?

Domagoj Ćorić orcid.org/0000-0002-9230-2211 ; Dr. Franjo Tuđman Defense and Security University, Department of Statistics *
Matej Kožnjak ; Faculty of science – Department of Mathematics, University of Zagreb
Dražen Smiljanić orcid.org/0000-0002-5759-7683 ; Dr. Franjo Tuđman Defense and Security University

* Corresponding author.

Abstract

This paper examines the long-run cointegration between the German DAX and the US S&P 500 from January 2021 to December 2025, using daily closing prices expressed as natural logarithms. The central argument is that methodological choices prevalent in the existing literature systematically fail to detect genuine long-run equilibrium relationships due to the econometric costs of over-differencing and firststep OLS bias amplification. Drawing on the theoretical contributions of Granger and Newbold (1974), Granger (1981), Granger and Joyeux (1980), Engle and Granger (1987), and Phillips (1988), the paper reconstructs the conditions under which differencing destroys low-frequency spectral dynamics and renders standard cointegration tests unreliable. With this conclusion in mind, the paper tests two models that encompass long-term cointegration: ARDL and ECM models. The empirical analysis confirms that the ARDL model outperforms its competitor, making it the best-fit methodology for modelling cross-Atlantic equity market integration and carrying direct implications for portfolio diversification, financial stability monitoring, and applied econometric practice.

Keywords

cointegration, ARDL, error correction model, DAX, S&P 500, spurious regression, fractional integration, capital market integration

Hrčak ID:

346308

URI

https://hrcak.srce.hr/346308

Publication date:

14.4.2026.

Visits: 159 *