Rising BRIC and Tectonic Shifts in Sub-Saharan Africa s Integration into the Global Economy. Abstract

Transcription

1 Rising BRIC and Tectonic Shifts in Sub-Saharan Africa s Integration into the Global Economy Abstract This paper analyzes the dynamic of Sub-Saharan Africa (SSA) business cycle synchronization (BCS) with the rest of the world (RoW). A particular attention is paid to the BRIC (Brazil, Russia, India and China) which is one of the important forces that are shaping the global economy. There are indications that SSA s BCS and trade are concomitantly and gradually shifting from advanced economies to the BRIC. Empirical estimates confirm that greater trade integration of SSA countries mainly explained the shifting patterns in their business cycle synchronicity. Moreover, it is documented that trade with the BRIC has the strongest impact on BCS. Its effect stands three times much higher than that with the G7, which is the main trading bloc of the region, although in decline. Furthermore, for all groups of countries but the G7, trade triggers output correlation, independently of the trading partners position in the business cycle. Finally, trade with G7 and the BRIC started affecting cyclical output correlation only from 1990 and onwards reflecting the recent BRIC-driven commodity boom and growth spell it ignited in SSA. However, on a deception note, the verdict on intra-ssa trade is somehow different, with this type of trade acting as an accelerating factor of BCS from 1970 to 2000, and as a neutral factor, thereafter. Such a recent decoupling effect of trade does not bode well with ongoing efforts to boost intra- Africa trade as a means to deepen regional integration and achieve monetary zone. Keywords: Business Cycle Synchronicity, Trade, Sub-Saharan Africa, and BRIC. JEL Classification: E3, F1, F3.

2 2 Table of Contents I. Introduction... 3 II. Theoretical considerations... 4 III. Empirical Methodology and Data... 6 A. Methodology... 6 B. Data... 7 IV. Results... 8 A. Some Stylized Facts... 9 B. Baseline C. Asymmetric Impact D. Evolution over time E. Robustness checks V. Conclusion... 17

3 3 I. INTRODUCTION The growing economic prominence of Brazil, Russia, India and China, known as the BRIC, is one of the major forces shaping the global economy. This rise has manifested itself through stronger trade and financial ties with the rest of the world, including with Sub-Saharan Africa (SSA). As they gained strength economically, these countries have become not only key markets for SSA exports and imports but also financiers, investors and donors, therefore acting as important growth powerhouses for many SSA countries. Not surprisingly the impressive commodity-price-driven growth sustained during much of the last decade has been attributable in part to the growing economic dynamism of the BRIC. Such a development suggests that output correlation between SSA and BRIC might have firmed up recently (Samake and Yang 2011). Although the question of synchronization is both relevant and timely, especially given the prospects of further rise of BRIC s share in the global economy and the intensification of their trade and financial relations with SSA, no study has hitherto explored this. The few existing studies on the relations between SSA and some BRIC, in particular China, analyze constraints and policy challenges (Broadman 2007) or potential gains and losses associated with increasing trade flows between these two trading partners (Kennan and Stevens 2005; Jenkins and Edwards 2006). To the best of our knowledge, our paper is the first attempt to explicitly investigate the impact of growing trade and financial flows between SSA and the rest of the world, with a particular focus on the BRIC, on business cycle synchronization, henceforth referred to as BCS The purpose of this paper is therefore twofold. It sheds light on the possible synchronization of SSA s business cycles with those of various economic blocs and identifies the underlying factors. Main findings are as follows. First, looking at the data for the period , SSA is being increasingly integrated into the global economy. The average synchronicity between SSA countries and the RoW is positive and trending upwards over time. Closer synchronization also coincides with the growing trade between SSA and the global economy. Synchronicity and trade integration have varied markedly across blocs (G7, BRIC, non-g7 OECD, SSA and Remaining Countries) and time. Of all groups under consideration, the G7 countries are those which SSA countries are more synchronized with. However, ties with the BRIC appeared to have particularly gained steam. There are indications that SSA s BCS and trade are concomitantly and gradually shifting from advanced economies to the BRIC. This observation lends support to our arguments that with deeper trade with the BRIC, drivers of the SSA s business cycles have also changed. Nowhere are these changes more conspicuous than that between SSA and the BRIC. Building on these observations, the econometric analysis demonstrates that greater trade integration of SSA countries mainly explained the shifting patterns in their business cycle synchronicity. Our results show that closer cycle correlation has moved in tandem with growing trade ties; and this is true irrespective of the groups of countries which SSA trades with. Moreover, it is documented that trade with the BRIC has the strongest impact on BCS. Its effect stands three times much higher than that with the G7, which is the main trading bloc of the region, although in decline. Furthermore,

4 4 for all groups of countries but the G7, trade triggers output correlation, independently of the trading partners position in the business cycle. Finally, trade with G7 and the BRIC started affecting cyclical output correlation only from 1990 and onwards reflecting the recent BRICdriven commodity boom and growth spell it ignited in SSA. However, on a deception note, the verdict on intra-ssa trade is somehow different, with this type of trade acting as an accelerating factor of BCS from 1970 to 2000, and as a neutral factor, thereafter. Such a recent decoupling effect of trade does not bode well with ongoing efforts to boost intra- Africa trade as a means to deepen regional integration and achieve monetary zone. These empirical findings are robust to the usual tests conducted in the literature. What to do with these findings? They are important at least on two counts. First, a greater BCS implies that SSA countries are now more vulnerable to shocks from the BRIC, therefore pointing to the need for SSA s policy makers to pay greater attention to policy stances in BRIC. In particular, a coupling of SSA s business cycles to BRIC s would have implications for the current structure and the functioning of exchange rate arrangements or those in the making across SSA, most of which are using or expected to use the currencies of some advanced economies as their anchor currencies. Second, a stronger BCS also suggests that SSA s course towards strong and sustainable growth and its efforts to diversify away from primary commodities might be shaped by the pace and sustainability of growth in BRIC and their bilateral trade patterns. The remainder of this paper is structured as follows. The next section quickly reviews the theoretical foundations of the key determinants of BCS on the other hand. Section III discusses the empirical strategy and data used in the analysis. Section IV presents and comments empirical findings. Finally, Section V offers some concluding remarks. II. THEORETICAL CONSIDERATIONS This section presents the determinants of business cycles synchronization as they relate to SSA countries. The theoretical point of departure is the framework developed by Frankel and Rose (1997, 1998), in which trade constitutes the key driver of BCS. From a theoretical standpoint the impact of increased trade on BCS is ambiguous. On the one hand, stronger trade relations are likely to cause increased cycle correlation when intraindustry trade dominates such relations. This is likely to be case when global production and supply chains feature trade relations. Favorable output shocks, industry-specific in nature, that occur in a particular economy may boost the demand for foreign goods, therefore driving economic expansion in trading partners (Frankel and Rose 1998, Imbs 2004, Baxter and Kouparitsas 2005, Caldéron and others 2007 and Inklaar and others 2008). On the other hand, if inter-industry trade dominates, then greater trade integration implies increased specialization. In such circumstances, industry-specific shocks result in lower output correlation (Krugman 1993, Imbs 2004).

5 5 The above argument is also applied to countries production bases. The literature suggests that the degree of economic specialization affects BCS. Krugman (1993) proposes that economic specialization drives industry-specific supply shocks and increases the magnitude of asymmetric shocks. If that is the case, trade is likely to increase business cycle decoupling. On the contrary, two economies that feature similar production structures are likely to be subject to analogous shocks, therefore displaying similar business cycle patterns (Imbs 2004). In sum, the more countries production bases differ, the more their business cycles move away from one another. Other factors such as financial integration and macroeconomic policies have also been found to have a significant predictive content over BCS. The literature has also purported that financial integration drives BCS. Another factor that affects cycle correlation is financial integration. The theoretical and empirical literature suggests that the direction of such causation could go either way. One the one hand, models of financial contagion show that a negative shock in a banking sector may result in the reduction of credit supply, therefore adversely affecting domestic real sector. Under deeper financial integration, this shock is transmitted internationally as banks reduce their overseas lending in order to repair their balance sheets or to be able to continue lending in countries affected by the shock (Allen and Gale 2000, Morgan and others 2004, Perri and Quadrini 2010, Enders and others 2010). On the other hand, standard international business cycle models point to a positive correlation between financial integration and BCS. Countries witnessing favorable productivity shocks and financially connected to the rest of world attract capital flows from countries that are not impacted by shocks. Conversely, when confronted with negative shocks, countries experience capital outflows in favor of other countries that are not affected by these disturbances. As a result, business cycles diverge (Backus and others 1992, 1994). The business cycle literature also posits that pursing similar macroeconomic policies increases the likelihood of business cycle correlation (Böwer and Guillemineau 2006). The other way round might also hold true. Similar business cycles induce analogous policy reaction and macroeconomic policy synchronization. Inklaar and others (2008), for instance, reported that, in OECD countries, the impact of macroeconomic convergence-particularly fiscal and monetary on output correlation is as strong as that of TI. Also, factors such as those that capture the degree of exposure to global shocks say oil dependency or the level of indebtedness, determine whether business cycles converge or diverge. Among these factors, a well-documented and consistent finding from this literature suggest that trade is the most important engine of BCS, especially in developing countries such as SSA countries (Baxter and Kouparitsas 2005, Caldéron and others 2007, Inklaar and others

6 6 2008, and Tapsoba 2009). III. EMPIRICAL METHODOLOGY AND DATA A. Methodology We assess the impact of trade integration on BCS by estimating coefficient, α 1, in the following equation: S i, j,t α0 α1ti, j,t ε1,i.j,t (1) Si, j,t represents the correlation between the output of country i and that of partner j and in time t; T stands for trade intensity, thereafter termed TI, between country i and that of i, j, t country j and in time t; and ε represents the vector of error terms. 1,i, j, t The sign of the coefficient, 1, captures the direction in which TI influences cycle correlation between SSA countries and the rest of the world. This coefficient is positive when the converging effect of trade dominates, and negative when its diverging impact prevails. The value of the coefficient measures the magnitude of the impact of trade on BCS; hence capture the degree of SSA s output co-movement with various groups of countries. The instrumental variable (IV) technique is used to estimate equation (1) 1. Following Frankel and Rose (1997, 1998), basic variables of gravity model are considered instruments for TI. These variables comprise the logarithm of the product of outputs; the logarithm of the distance between the main cities of the countries within the pair; a dummy variable for common border, with this variable equal to 1 if countries within the pair share a common border, and 0, if not; another dummy variable for common language, with this variable being set at 1 if countries within the pair have a common language spoken by at least 9 percent of population, and 0, if not. We also address potential heteroscedasticity by means of pair-clustering method. This method assumes that observations for a pair of countries are not independent over decades. Put simply, observations of the first decade may affect those of the second and the third 1 A simple way to estimate equation (1) is to use Ordinary Least Squares (OLS). However, OLS may generate biased coefficients because of the endogeneity of TI. To be sure, countries displaying BCS are likely to trade more (or less) during common expansions (or common recessions). Moreover, monetary zone arrangement boosts both trade and macroeconomic policy. As a result, a positive effect of TI on synchronicity may be due to such arrangement. This is known to be the simultaneity bias.

7 7 decade. Another problem we may be confronted with is one of addressing concern about spatial inter-dependence. An observation of the pair of countries A and B may depend on that of the pair of countries B and C. Yet we could not account for such spatial inter-dependence, hence adjust covariance matrices, as observations are missing. B. Data This section describes how key variables are computed and sources from which primary data used for this computation are drawn. The dependent variable is the degree of BCS between of country i and partner j at a given time t. Business cycle is measured by the de-trended component of the logarithm of real GDP (Gross Domestic Product) growth. 2 GDP data in constant US dollars are pulled from the IMF World Economic Outlook databank. Consistent with widely accepted approaches in the literature on BCS, the cyclical component of output is computed using the Hodrick and Prescott s filter, hereafter, the HP filter (Frankel and Rose 1998, Rose and Engel 2002, and Darvas and others 2005). The use of the HP filter is, however, subject to two criticisms. First, the HP filter is interpreted as a high-pass filter that removes longer fluctuations with a frequency of more and includes those fluctuations in the trend. Given this feature, this filter might not always be relevant from a policy perspective, particularly in SSA countries where institutions and policy responses are weaker than elsewhere. Turning the light on short and medium-term fluctuations is of key relevance for these countries and the HP filter may not provide such a focus. Second, there has been some debate about the value of the smoothing parameter, lambda, for annual data. Hodrick and Prescott (1997) initially set the value at 100. Rigorous evidence from Ravn and Uhlig (2002) subsequently puts the value of the above parameter at We follow the latter in the paper. Alternative filters to the one developed by Hodrick and Prescott (1997), in particular the Baxter and King s (1999) band-pass linear filter, henceforth referred to as BK, also receives greater attention (Frankel and Rose (1997, 1998), Rose and Engel (2002), and Darvas and others 2005). These filters address the shortcomings of the HP-based filter. 3 Despite its drawbacks, the HP filter generates, however, results that are qualitatively similar to those of alternative filtering methods (Frankel and Rose 1997 and 1998, Caldéron and others 2007). We evaluate the sensitivity of our baseline findings to alternative filtering methods or in other words to alternative measures of cyclicality in the section on section on robustness. It is worth recalling that BCS is measured through a 10-year moving average of coefficients of correlation of the 2 Some authors, like Frankel and Rose (1997, 1998), Darvas and others (2005), and Inklaar and others (2008) use alternative proxies of economic activity such as industrial production or employment rate. Such data are not available for SSA countries. 3 The BK filter addresses the shortcomings associated with the HP filter by combining a high-pass filter and a low-pass filter and by setting the length of cycle according to the assumptions of the authors. We assume that the length of the business cycle is between 2 and 8 years, which virtually matches the time span that was originally recommended by Baxter and King (1999) 6 to 32 quarters (i.e. 1.5 to 8 year)

8 8 business cycle of country i and that of partner j. The second variable of interest is TI. The degree of TI between countries i and j is equal to the value of bilateral trade divided by the sum of the total trade or the sum of the output of countries i and j. Measures of TI used in this article are those used by Frankel and Rose (1997, 1998) and Baxter and Kouparitsas (2005). Trade intensities (TI1 and TI2) are Mijt Xijt Mijt Xijt computed as follows: TI1ijt and TI2ijt. (X M ) (X M ) Y Y it it jt We also test, in the section on robustness the sensitively of our findings to alternative measures of TI indicators, particularly with respect to the way normalization is done. For such small economies as those from SSA, normalizing bilateral trade by total trade or output of the pair is likely to generate very marginal numbers. Because of this, we favor expressing bilateral trade relative to SSA countries trade and output along the following line: Mijt Xijt Mijt Xijt TI3ijt and TI4ijt. X represents the bilateral exports, Free On ijt (Xnt Mnt) Yit Board (FOB), of country i to country j and M is the bilateral imports, Cost-Insuranceijt Freight (CIF), of country i from country j. X it ( M it ) stands for total exports FOB (total imports CIF) of country i. Y it represents the nominal GDP of country i. Bilateral trade, total exports, total imports and nominal GDP data are derived from the IMF s Direction of Trade. Since numbers on TI for SSA countries might still be small, even when using the above measures, we opted for expressing TI in logarithmic form, thus enabling some degree of variability and ensuring that we obtain a meaningful and readable coefficient on TI. 4 Our dataset covers 44 SSA countries and their trading partners, and the period under consideration spans from 1970 to Data on BCS, TI, contains at most [...] pairs of countries. In total, the dataset includes at most 116,833 observations. As data are missing for some countries and for some years, we end up with an unbalanced panel. jt it jt IV. RESULTS This section presents the empirical assessment of the impact of TI on business cycle correlation between SSA countries and their trading partners. This analysis will be prefaced by the review of BCS and TI trends, the purpose of which is to identify potential regularities across on the association between these two variables across groups of countries and across time. 4 The estimates are still robust even without applying logarithmic form to TI variables.

9 9 A. Some Stylized Facts We first attempt to draw some stylized facts based on key variables of interest, in particular measures of BCS and indicators of TI described in the section on data (Figure 1). We first identify potential synchronization trends. For the period , average synchronicity between SSA countries and the RoW is positive and trending upwards over time. This trend differs across partner blocs. Of all groups under consideration, the G7 countries are those which SSA countries are more synchronized with. The period of interest is split into three sub-periods: , and This breakdown uncovers a closer synchronization of outputs between SSA countries and other groups of countries, starting from 1980 and onwards. More importantly, average output correlation across all country pairs not only strengthened but also seemed to be converging after The comovement of output between SSA and the BRIC appeared to have particularly gained steam. Closer synchronization also coincides with the growing trade between SSA and the global economy (Figure 1). The direction and volumes of Sub-Saharan Africa s bilateral flows have changed. Nowhere are these changes more conspicuous than that between SSA and the BRIC. Bilateral trade flows between SSA and the BRIC expanded between 2001 and 2010 as opposed increases for trade within the region and with G7, respectively. Also, rising trade flows with the BRIC seemed to have gained momentum after 2005, thus coinciding with the commodity price boom of the second half of 2000s. In line with the literature, there are indications that SSA s BCS and trade are concomitantly and gradually shifting from advanced economies to the BRIC. This observation lends support to our arguments that with deeper trade with the BRIC, drivers of the SSA s business cycles have also changed.

10 10 Figure 1. SSA and the Global Economy SSA output synchronicity y ( SSA output synchronicity ) SSA-G7 SSA-BRIC SSA-Non BRIC TI 1 TI SSA-BRIC SSA-OECD SSA-Other SSA-SSA TI 3 TI Source: Authors' calculations

11 11 B. Baseline To confirm potential association between greater synchronization and growing trade links, we perform a bivariate analysis between these two. We therefore regress our measures of business cycle movement on those of TI measures. An intercept and year dummies are included in all regressions. The purpose of the year dummies is to account for potential common shocks to SSA and their trading partners. Because of potential dual causality between BCS and TI, Ordinary Least Squares (OLS) estimator may not yield unbiased coefficients. We control for endogeneity by reverting to the instrumental variable (IV) estimation strategy. This is the method that is consistently used across the analysis, unless indicated otherwise. The results are reported in Table 1. They validate our contention that greater trade integration of SSA countries causes increased coupling of their business cycle with that of the RoW. We then consider the trade relations between SSA and various groups of countries, namely the G7, BRIC, non-g7 OECD, SSA and Remaining Countries. Our results show that closer cycle correlation has moved in tandem with growing trade ties; and this is true irrespective of the groups of countries which SSA trades with. To compare the relative magnitude of the effects of SSA s trade with these various groups, we compute standardized coefficients on TI (Table 2). One result stands out: trade with the BRIC has the strongest impact on BCS. Its effect stands three times much higher than that with the G7, which is the main trading bloc of the region, although in decline. Table 1. Trade Intensity and BCS. SSA and ROW (IV estimates) Comovement HP Overall G7 BRIC Non-G7 OECD SSA Remaining countries TI *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) TI *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) N 116, ,350 11,703 11,668 5,244 5,233 28,619 28,549 27,873 27,823 43,394 43,077 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included. Table 2. Trade Intensity and BCS. SSA and ROW (Standardized IV estimates) Comovement HP Overall G7 BRIC Non-G7 OECD SSA Remaining countries TI1 6*** *** 0.221*** 0.179*** *** *** ( ) (0.0125) (0.0260) ( ) (0.0111) (0.0133) TI2 4*** *** 0.213*** 0.183*** 8*** *** ( ) (0.0131) (0.0249) ( ) (0.0112) (0.0136) N 116, ,350 11,668 11,668 5,233 5,233 28,549 28,549 27,823 27,823 43,077 43,077 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included.

12 12 C. Asymmetric impact The next question of interest is to find out if the uncovered positive association between TI and output correlation is influenced by position in the cycle. In this regard, we consider two scenarios: the first focusing on times during which the trading partner s business cycle is positive, and the second turning the light period when the trading partner s business cycle is negative. Results are summarized in Table 3. It turns out that, for all groups of countries but the G7, trade triggers output correlation, independently of the trading partners position in the business cycle. Good or bad fortunes in the BRIC, non-g7 OECD, within SSA and other remaining countries do not alter the positive impact of trade on business cycle correlation. The trade relations with G7 countries stand as the only exception. The positive association between TI and cyclical output correlation holds only when the economic tide is high in G7 countries. Trade with G7 has no impact on BCS in times when output is below long-term potential in the latter. On the one hand, these findings echo the view that the output in SSA is increasingly decoupling from traditional partners cycle, in particular G7 s, and anchoring on emerging countries, including the BRIC s. On the other hand, they lend credence to the conjecture that the recent economic revival in SSA owes much to growing demand in emerging countries, especially the BRIC, for the region s primary commodities. That raises concern about this development perpetuating and even deepening the economic dependence of SSA on primary commodities or, in other words, on the vagarities of international commodity markets. Partner's positive cycle Partner's negative cycle Table 3. Trade Intensity and BCS. Asymmetric impact (IV estimates) Comovement HP ALL G7 BRIC Non-G7 OECD SSA Remaining countries TI *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) TI *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) N 58,535 58,317 5,817 5,804 2,578 2,574 13,844 13,827 14,143 14,123 22,153 21,989 TI *** * *** *** *** ** ( ) ( ) ( ) ( ) ( ) ( ) TI *** *** *** *** ** ( ) ( ) ( ) ( ) ( ) ( ) N 57,479 57,230 5,754 5,732 2,618 2,611 14,509 14,472 13,536 13,506 21,062 20,909 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included. D. Evolution over time We further explore whether the magnitude of the impact of trade co-movement of outputs is uniformly distributed across time. As the findings relate to the whole period under consideration, we investigate how the relationship between TI and BCS evolves over time and across groups of trading partners. Results are presented in Table 4 and Figures 2a and 2b. Trade with G7 and the BRIC started affecting cyclical output correlation only from 1990 and onwards. This mirrors very much the recent BRIC-driven commodity boom and growth spell it ignited in SSA. The verdict on intra-ssa trade is somehow different, with this type of trade acting as an accelerating factor of BCS from 1970 to 2000, and as a neutral factor, thereafter. Such a recent decoupling effect of trade does not bode well with ongoing efforts to boost

13 13 intra-africa trade as a means to deepen regional integration and achieve monetary zone. By contrast, non-g7 OECD is the only group for which the relationship between TI and output correlation does not qualitatively changed over time. In between all the above lies the findings on the remaining countries (RC) group. The direction and the statistical significance of the coefficient on TI between this group and SSA vary over time. It is positive from 1970 to 1979, negative from 1980 to 1989, then insignificant from 1990 to 2000, only to turn positive from 2000 to This instability partly indicates the extent to which the group is diverse, with very limited discernible common patterns. Table 4. Trade Intensity and BCS over decades (IV estimates) Comovement HP Decades TI1 N TI2 N Overall *** ( ) 14, *** ( ) 14, *** ( ) 21, *** ( ) 20, *** ( ) 34, *** ( ) 34, ( ) 44, * ( ) 44,433 G * ( ) 2, * ( ) 2, ( ) 2, ( ) 2, *** ( ) 3, *** ( ) 3, *** ( ) 3, *** ( ) 3,208 BRIC (0.0127) (0.0119) (0.0101) 1, ( ) 1, *** ( ) 1, *** ( ) 1, *** ( ) 1, *** ( ) 1,763 Non-G7 OECD *** ( ) 4, *** ( ) 4, *** ( ) 5, *** ( ) 5, *** ( ) 8, *** ( ) 8, *** ( ) 9, *** ( ) 9,855 SSA *** ( ) 3, *** ( ) 3, *** ( ) 4, *** ( ) 4, *** ( ) 8, *** ( ) 8, ( ) 10, ( ) 10,613 Remaining countries ** ( ) 4, * ( ) 4, *** ( ) 6, ** ( ) 6, ( ) 12, ( ) 12, *** ( ) 19, *** ( ) 18,994 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included.

14 14 Figure 2a. Impact of Trade on BCS over Time ( ) SSA and All sample TI 1 TI SSA and BRIC - TI TI 2 SSA and G7 - TI TI 2

15 15 Figure 2b. Impact of Trade on BCS over Time ( ) SSA and non-g7 OECD - TI TI 2 Intra-SSA - TI TI 2 SSA and Remaining countries - TI TI 2 E. Robustness checks

16 16 We conduct some robustness checks to ensure that our results do not depend neither on the identification of business cycle or trade intensity nor on estimation technique. First, we evaluate the sensitivity of our initial findings to an alternative measurement of cyclicality, the Baxter-King (BK) filter (Table 5). Results are roughly similar, albeit showing a rather modest synchronization. The above exercise is repeated using another measure of TI, with again similar findings (Table 6). The findings are qualitatively unchanged under the alternative measure of cyclicality, the BK filter. Table 5. Trade Intensity and BCS. SSA and ROW (IV estimates) Comovement BK Overall G7 BRIC Non-G7 OECD SSA Remaining countries TI *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) TI *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) N 116, ,350 11,703 11,668 5,244 5,233 28,619 28,549 27,873 27,823 43,394 43,077 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included. Table 5. Trade Intensity and BCS. SSA and ROW (Standardized IV estimates) Table 6. Trade Intensity and BCS. SSA and ROW (IV estimates) Comovement BK Overall G7 BRIC Non-G7 OECD SSA Remaining countries TI *** ** *** *** *** ( ) ( ) (0.0109) ( ) ( ) ( ) TI *** *** *** *** ( ) ( ) (0.0104) ( ) ( ) ( ) N 115, ,833 11,608 11,703 5,203 5,244 28,303 28,619 27,795 27,873 43,020 43,394 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included. Second, having found that the coefficients on TI are statistically significant and positive, we now check the robustness of our findings to alternative estimation procedures. We choose to perform Quantile and Tobit estimation strategies. The high degree of heterogeneity of our sample and significant within-group disparities, hence the existence of many potential outliers, militates in favor of Quintile regressions, while potential truncation, due to limited trade data reporting, makes the case for trying a Tobit model estimation. Results are summarized in Tables 7 and 8. TI measures enter all the regressions significantly, with the exception of those between SSA and the remaining countries. Overall, these findings show that trade intensification has resulted in SSA s output cycle synchronizing with those of their trading partners.

17 17 Table 7. Trade Intensity and BCS. SSA and ROW (Quantile regression) Comovement HP Overall G7 BRIC Non-G7 OECD SSA Remaining countries TI *** *** *** *** *** ** ( ) ( ) ( ) ( ) ( ) ( ) TI *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) N 124, ,110 12,431 11,795 5,498 5,247 30,102 28,691 29,423 27,889 46,634 43,488 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included. Table 8. Trade Intensity and BCS. SSA and ROW (Tobit regression) Comovement HP Overall G7 BRIC Non-G7 OECD SSA Remaining countries TI *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) TI *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) N 124, ,110 12,431 11,795 5,498 5,247 30,102 28,691 29,423 27,889 46,634 43,488 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Intercept and Year dummies included. V. CONCLUSION The purpose of this paper is therefore twofold. It sheds light on the possible synchronization of SSA s business cycles with those of various economic blocs and identifies the underlying factors. Main findings are as follows. There are indications that SSA s BCS and trade are concomitantly and gradually shifting from advanced economies to the BRIC. Empirical estimates confirm that greater trade integration of SSA countries mainly explained the shifting patterns in their business cycle synchronicity. Moreover, it is documented that trade with the BRIC has the strongest impact on BCS. Its effect stands three times much higher than that with the G7, which is the main trading bloc of the region, although in decline. Furthermore, for all groups of countries but the G7, trade triggers output correlation, independently of the trading partners position in the business cycle. Finally, trade with G7 and the BRIC started affecting cyclical output correlation only from 1990 and onwards reflecting the recent BRIC-driven commodity boom and growth spell it ignited in SSA. However, on a deception note, the verdict on intra-ssa trade is somehow different, with this type of trade acting as an accelerating factor of BCS from 1970 to 2000, and as a neutral factor, thereafter. Such a recent decoupling effect of trade does not bode well with ongoing efforts to boost intra-africa trade as a means to deepen regional integration and achieve monetary zone.

18 18 References Allen and Gale 2000 Backus and others Baxter, M. and M.A. Kouparitsas (2005) Determinants of Business Cycle Co-movement: a Robust Analysis, Journal of Monetary Economics, 52 (1): Baxter, M. and R.G. King (1999) Measuring Business Cycles: Approximate Band-Pass Filters For Economic Time Series, The Review of Economics and Statistics, 81 (4): Böwer and Guillemineau 2006 Broadman 2007 Calderon, C., A. Chong, and E. Stein, 2007, Trade Intensity and Business Cycle Synchronization: Are Developing Countries any Different? Journal of International Economics, Vol. 71, pp Darvas, Z., A.K. Rose and G. Szapary (2005) Fiscal Divergence and Business Cycle Synchronization: Irresponsibility Is Idiosyncratic, NBER Working Papers 11580, National Bureau of Economic Research. Enders and others 2010 Frankel, J. and A. Rose, 1997, Is EMU More Justifiable ex post than ex ante? European Economic Review, Vol. 41, pp Frankel, J. and A. Rose, 1998, The Endogeneity of the Optimum Currency Area Criteria, Economic Journal, Vol. 108, pp Hodrick, R.J. and E.C. Prescott (1997) Postwar U.S. Business Cycles: An Empirical Investigation, Journal of Money, Credit and Banking, 29 (1): Imbs, J., 2004, Trade, Finance, Specialization and Synchronization, Review of Economics and Statistics, Vo. 86(3), Inklaar, R., R. Jong-a-Pin, and J. de Haan, 2008, Trade and Business Cycle Synchronization in OECD Countries - a Re-examination, European Economic Review, Vol. 52(4), Jenkins and Edwards 2006 Kennan and Stevens 2005

19 19 Krugman, P. (1993) Lesson of Massachusetts for EMU, in F. Giavazzi and F. Torres (eds), The Transition to Economic and Monetary Union Europe, New York: Cambridge University Press, pp Morgan and others 2004 Perri and Quadrini 2010 Ravn, M., and H. Uhlig, 2002, On Adjusting the Hodrick Prescott Filter for the Frequency of Observations, Review of Economics and Statistics, Vol. 84, Rose, A. and Engel, C., 2002, Currency Unions and International Integration, Journal of Money, Credit, and Banking, Vol. 34, pp Samake, Issouf and Yang Yongzheng, 2011, Low-Income Countries BRIC Linkage: Are There Growth Spillovers? IMF Working Paper 11/267 (Washington: International Monetary Fund). Tapsoba, S., 2009, Trade Intensity and Business Cycle Synchronicity in Africa, Journal of African Economies, Vol. 18(2),