Depressions in the Colombian economic growth during the twentieth century: a Markov switching regime model

Abstract

In this article, we modelled the Colombian long run economic growth (1925–2003) using a two-regime first order Markov switching model. We found evidence of nonlinearity in the annual rate of economic growth. The results show that changes between regimes are sudden and sporadic. The Colombian economy remains in the sustainable growth regime most of the time. The turning points from the Markov switching model capture very well the behaviour of real output through time. In fact, they identify the four main depressions of the century.

Acknowledgements

We are grateful to Juan Nicolas Hernández for excellent research assistance. We thank Luis Eduardo Arango, Munir Jalil, Jesús Otero and Carlos Esteban Posada for their comments. The views expressed are those of the authors and not of the Central Bank of Colombia or of its Board of Directors.

Notes

¹ See Hamilton (1994) pp. 677–701

² One advantage of this model, as Moolman (2004) underlines, is that no previous information concerning the dates when the economy was in each regime or the size of these regimes are required. Also, the probability of being in a specific regime is inferred from the data.

³ See Filardo and Gordon (1998).

⁴ An alternative approach based on the decomposition of a univariate and multivariate nonlinear process into their permanent and temporary components is used by Clarida and Taylor (2003), who applied the univariate case to study nonlinearities in the US business cycle. Threshold models and smooth transition autoregressive models are also widely used to study nonlinearities in macroeconomic variables; see e.g. Arango and Melo (2006), Andreano and Savio (2002) and Stanca (1999).

⁵ See Posada (1999), Fernández and Gonzáles (2000), Misas and Posada (2000) and Urrutia and Fernández (2003).

⁶ During 1951 to 1974 the SD of the annual rate of growth was 1.5% and during 1984 to 1997 it was 1.2% been the lowest of the period under analysis.

⁷ During 1925 to 1950 the SD of the annual rate of growth was 3.25%, the largest of the period under analysis.

⁸ See The Boards of Directors’ Report to the Congress of Colombia, Banco de República, March 2000.

⁹ Following Cruz (2005) and Layton and Smith (2000), we also considered an alternative specification that allows for three different regimes (i.e. depression, sustainable growth and booms). This specification was not supported by the data though.

¹⁰ However, when the variances are estimated individually, the variance of the depression regime is numerically higher than the variance of the sustainable growth regime.

¹¹ A similar model is presented in Hamilton (1990).

¹² Similar magnitudes are found in Hamilton (1989) for the US's GDP rate of growth.

¹³ For practical reasons, Equation 8 is truncated in i = 100 000.

¹⁴ Previous studies have found that the average length of the Colombian business cycle is approximately 8 years; see for example, Posada (1999) and Fernández and Gonzáles (2000).

¹⁵ Similar results were found by Arango and Melo (2006) for the Colombian industrial production index, their proxy for economic activity.

¹⁶ In 1949, the economy grew 8.7%, in 1948, 4% and in 1946, 9.6%. The economic slowdown of 1950 was the consequence of tight economic policies (Ocampo,1987).