Trade Integration and Business Cycle Synchronization among East African Community Countries

ABSTRACT

This study uses a panel of 31 African countries for a period of 17 years to assess whether trade integration among these countries has led to the synchronization of their business cycle. This is an important issue to consider for entry into a monetary union. Unlike most previous studies, a time-varying index is applied to measure business cycle synchronization. The heterogeneous panel estimators reveal a positive relationship between the two variables among these African countries. This empirical finding should be considered by the East African Community (EAC) authorities for further debate on the readiness of the EAC Monetary Union.

KEYWORDS:

• Business cycle synchronization
• trade integration
• EAC

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/08853908.2022.2128945

Notes

¹ An OCA is defined as a geographic area that could benefit from using a common currency.

² To our knowledge, only Tapsoba (2010) discusses the same issue among 53 African countries. However, he uses correlation coefficients to measure the business cycle synchronization. Nevertheless, it is worth noting that several studies have assessed the monetary unification in Africa only using the OCA criterion of business cycle synchronization; however, to save space, this literature is not discussed. See Asongu, Nwachukwu, and Tchamyou (2017) for ample literature on existing and proposed African monetary unions, and Samba and Mbassi (2020) for CFA zones.

³ The sample and study period choices depend on the availability of data, which must be available for the variables considered in the current study. The list of countries is provided in Table A5 in the online Appendix.

⁴ They use different macroeconomic variables including real GDP, index of industry production, total employment, and unemployment rate.

⁵ As a robustness check, the author includes OECD countries in the sample to compare developed to developing countries.

⁶ The correlation coefficients between two countries are often estimated over a given time spanning between 5 to 10 years. They are then used as the dependent variable in the regression model, and all other regressors must be averaged over the same time span as well. This is suitable for cross-sectional analysis but not for panel data estimations, where a time-varying measure is strongly required (see Gianelle, Montinari, and Salotti (2017)).

⁷ See the summary statistics reported in Table A1 in the online Appendix.

⁸ See also Gianelle, Montinari, and Salotti (2017) and Kalemli-Ozcan, Papaioannou, and Peydró (2013).

⁹ The main reason is to capture as much trade as possible among these countries, which import more than export between themselves.

¹⁰ As stressed in Baldwin and Taglioni (2006), these dummies absorb the time dummies, hence eliminating the omitted variables bias and any other bias from incorrect computations of the included regressors.

¹¹ In an IV estimation, Frankel and Rose (1998, 1020–1023) used distance, common border, common language, and additional dummies of RTAs as valid instrumental variables. In addition, the gravity approach corrects some of the limitations of using the bilateral trade indicators used in the current study. This is because it provides a way of estimating the potential levels of trade based on a specific series of other determinants of bilateral trade (Bouët, Cosnard, and Laborde 2017).

¹² The year 2000 marks the beginning of the EAC between the three original founders of the EAC. Thus, one could assume that there could be trade effects with spill-over effects to these neighboring countries (Rwanda and Burundi, who joined 7 years later), which should not be ignored.

¹³ Based on the bilateral characteristic of the variables of interest, and to avoid the endogeneity with the dependent variable, both indices are calculated in a similar way.

¹⁴ First-stage regression results in Table A2 of the online Appendix show that both variables do not have any explanatory power on bilateral trade. Hence, they are taken as exogenous variables.

¹⁵ Following DeHoyos and Sarafidis (2006), the Pesaran (2015) CD test, which is suited for both balanced and unbalanced panels, is applied under the null hypothesis of cross-sectional independence/weak cross-sectional dependence; large values indicate that there is strong correlation between panel units (Pesaran 2021). For most of the estimated models, the null hypothesis is rejected, indicating that that there is cross-sectional dependence in the residuals.

¹⁶ As explained in Pesaran (2006), the intuition behind this estimation procedure is to filter the individual-specific regressors using cross-sectional aggregates, such that as the number of units tends to infinity, the differential effects of unobserved common factors are eliminated. For brevity, the estimated coefficients of the cross-sectional averages of observed and unobserved variables are not reported.

¹⁷ Both tests are performed using Stata version 16.0, where a Stata command multipurt for Maddala and Wu’s (1999) panel unit test (MW assumes cross-section independence) and the Pesaran (2007) CIPS panel unit root test (CIPS assumes cross-section dependence) is used. The null hypothesis for both tests is non-stationarity across all panels, i.e., the series panel contains a unit root.

¹⁸ To save space, the MW results are not reported but are available upon request for interested readers. From this test’s results, the null hypothesis of no stationarity is rejected in all panels and confirms that only a fraction of the panel series is stationary.

¹⁹ Specifically, the test of weak instruments follows the developments and discussions in Stock, Wright, and Yogo (2002). As further explained by these authors, the proposed test performs well when diagnosing instrument relevance in a model with one endogenous variable, which is the case in the current article.

²⁰ This is in line with the findings of Böwer and Guillemineau (2006), showing that financial integration (in terms of FDI flows) may appear not robust in terms of business cycle synchronization. Caporale and Girardi (2016) also found a limited role of capital flows in business cycle comovement.

²¹ The reader should also refer to Table A2 in the online Appendix showing the positive estimated coefficient of the “CFA members” variable, indicating that countries using CFA trade more compared to countries that do not use CFA.

²² Stata version 15.1 was used to generate these figures.

²³ There are too many pairings (each dot refers to a different country pair) for countries in CFA zones compared to EAC countries. Thus, for clarity, only some are shown.