Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Does Geographical Diversification in International Trade Reduce Business Cycle Volatility? ARIAN FARSHBAF 1, MAY 2014 Abstract: This paper studies empirically the effect for business cycle volatility in open economies of geographical diversification in international trade. Using a panel data of 133 countries over five sub-periods covering 1962-2006 and applying Instrumental Variables Generalized Method of Moments estimation to address potential simultaneity, it is shown that greater diversification of trade among trade partners, measured by the size of a Herfindahl index, is associated with significantly lower domestic output volatility, although the estimated impact for lower consumption volatility is not statistically significant. We also find that diversified trade insulates output fluctuations from various fiscal, financial, and international shocks. In addition, analyses of the newly constructed trade indicators introduced in this paper demonstrate that trading with economies that are larger, more developed, or with lower business cycle volatility reduces domestic output volatility. Finally, home business cycles being less synchronized with those of trading partners is associated with lower output volatility, once an external shock is controlled for. JEL Classifications: F15, F40, F44 Keywords: Business Cycle Volatility, International Trade, Geographical Diversification, Trade Partners, International Business Cycle Synchronization 1 This paper is based on a chapter of my Ph.D. dissertation on “Business Cycles Volatility and Global Trade” completed at the Department of Economics, University of Southern California. Contact: [email protected]. † I am grateful to all members of my Ph.D. Committee: Professors Caroline Betts, Robert Dekle, Cheng Hsiao, John Matsusaka and Vincenzo Quadrini for their guidance, support and valuable advice throughout. I would like to also thank professors Jeffrey Nugent and Guillaume Vandenbroucke for their helpful feedback and advice. All errors and mistakes are mine. 1 1 Introduction Economic globalization, as measured by significant increases in international flows of goods and capital relative to the size of national economies, has introduced both challenges and opportunities for individual nations by exposing them to an array of international intra-temporal shocks, while providing for mitigation of idiosyncratic shocks through inter-temporal trade. The thesis of this paper is that geographical diversification in trade ‒ trading with a larger number of countries with relatively equal trade shares ‒ can reduce the domestic output risk of exposure to specific intra-temporal shocks in trade partners, and reduces domestic consumption risk by providing a more diversified set of lending and borrowing opportunities. In short, this paper tests the hypothesis that greater geographical diversification reduces business cycle volatility, conditional on a country’s level of trade openness. Business cycle volatility2 is viewed as a major indicator of national economic performance, perhaps mostly due to robust links with other macroeconomic performance indicators. Barlevy (2004) and Mendoza (2000) show a significant welfare loss ‒ much higher than the “trivial” amount argued by Lucas (1987) ‒ as a result of higher macroeconomic volatility, while Ramey and Ramey (1995) among others show a strong negative correlation between business cycle volatility and long-run growth. On the other hand, lower business cycle volatility can be regarded as the result of successful stabilization policies pursued by policy makers. Empirical work on the subject of business cycle volatility (hereafter “BCV”) falls into three major groups. One strand studies BCV as a time series, with a focus on a describing the timeseries behavior for a specific country or group of countries. Blanchard and Simon (2002), is one of the most frequently cited works of this kind, where various properties of the growth rate of the U.S. output such as persistence and volatility are studied. They find that a decline in output volatility, rather than lack of relatively large shocks, is behind the extended U.S. economic expansion in the 1990s. Kose, Otrok and Whiteman (2003) decompose business cycles into components attributable to domestic and international factors through an unobservable common factor analysis. Usually time series studies are primarily descriptive and non-structural and no conditional relationships between business cycle volatility and other variables are examined. 2 Business cycle volatility is defined as the standard deviation of a macroeconomic aggregate’s growth rate or its cyclical volatility as will be explained in detail in Section 2. 2 A second line of literature tries to account for variations in volatility across nations and over time, by considering unconditional correlations with factors proposed as potential sources of macroeconomic fluctuations. Karras and Song (1996) is an example which finds volatilities in exchange rates and money supply, the degree of openness to trade, and smaller government size are positively correlated with higher business cycle volatility. The third line of research, which the current paper falls within, focuses on measuring the conditional correlation of business cycle volatility with a specific factor or related factors believed to be potentially important in influencing business cycle behavior. Ferreira da Silva (2002) for example examines econometrically the impact of financial development on BCV. Similarly, in our analysis, we do not try to account for as much variation in the volatility of macroeconomic fluctuations as possible ‒ although appropriate country specific characteristics and major factors that contribute to aggregate volatility are included as controls. The goal, rather, is to investigate the impact of an important aspect of international trade, namely geographical diversification, on the volatility of fluctuations. The relationship between trade openness as one of the proxies for economic globalization3 with business cycle volatility has been extensively investigated in the literature. Due to the ambiguous stance of theory regarding the relationship, it has been tackled as an empirical task with a close-to unanimous finding that trade openness is a destabilizing factor for business cycles even after controlling for many country characteristics. Di Giovanni and Levchenko (2009), Karras and Song (2003) and Easterly, Islam and Stiglitz (2001) are among many who reached this conclusion. By contrast, Cavallo (2008) reaches his main finding ‒ namely trade openness has a net stabilizing effect on macroeconomic fluctuations ‒ after separating the effect of terms of trade (TOT) volatility from the overall impact of openness. Some studies have produced caveats to the general conclusion that trade openness aggravates domestic output volatility. For example, Kose, Prasad and Terrones (2003) investigate a potential non-linearity for financial openness-output volatility relationship and find that the stabilizing effect of financial openness 4, the capital account equivalent of trade openness, is realized only after a certain threshold of 3 Farshbaf (2012b) challenges treating “de facto openness” alone as a proxy for economic globalization. Financial openness is defined as gross stocks of capital flows to GDP, based on data constructed by Lane and Milesi-Ferretti (2001). Gross capital flows is the sum of inflows and outflows in terms of foreign direct investment (FDI), portfolio investment and bank lending. 4 3 financial openness is attained, which potentially explains why emerging markets have not yet reaped benefits of financial liberalization. A similar analysis performed in Farshbaf (2012b) shows that trade openness and output volatility also share a non-linear relationship; trade openness increases output volatility up to a threshold level of openness, estimated at around trade/GDP ratio of 88 percent, after which level increases in openness reduce output volatility. In addition, it documents that countries which are more financially integrated are less susceptible to the increased output volatility associated with greater trade openness. In the current paper, the role of geographical diversification is studied as a potential counter to the adverse output volatility impact of various shocks for a given level of trade openness and country-specific characteristics. Despite empirical investigation of the trade openness-volatility relationship, the literature has been silent on the role of geographical diversification of trade. The only direct work we later became aware of is a mimeograph by Bacchetta et al (2007) that examines the effect of export diversification on output volatility. By contrast, there are numerous works studying the impact of sectoral concentration in trade for economic outcomes such as growth and fluctuations. For example, Di Giovanni and Levchenko (2009) find the specialization that emerges from opening up to trade increases the volatility of output growth. Our paper attempts to bridge the existing gap in the empirical literature with a comprehensive investigation of the impact of geographical diversification in international trade on business cycle volatility. An important contribution of this paper is the construction of four new international trade indicators that capture various economic characteristics of trade partners for each country in the dataset, using bilateral trade data. We examine how, on average, across nations and over time, trading with more developed nations, larger economies, countries with more volatile business cycles, and countries which share with the domestic economy higher degrees of correlation among business cycle indicators influences the volatility of business cycles at home. We regard this study as a portfolio view of international trade, where a country’s trade partners constitute the assets in the portfolio and the performance of the portfolio is assessed by the reduction in business cycle volatility, which in turn depends on the economic characteristics of the trade partners and the correlation of business cycles among them. 4 In addition, we account for the possibility that the level of geographical diversification is endogenously determined as a policy choice outcome, perhaps selected to be a function of current macroeconomic volatility. This naturally raises the issue of simultaneity and calls for an appropriate identification strategy. Using a panel of 133 countries 5 over 5 sub-periods covering 1962-2006 and following instrumental variable General Method of Moments technique which addresses potential simultaneity, the paper shows that for a given level of trade openness, more geographically diversified international trade in either imports or exports, captured by a lower value for a Herfindahl index (HI), reduces business cycle volatility. Further, for a given level of exposure to international trade and geographical diversification, trading with economies that are more developed, larger and with less volatile output fluctuations is also associated with lower output volatility at home. Our measure of international business cycle synchronization, which is a weighted6 multilateral correlation of a country's business cycles with those of its trading partners, is also strongly associated with lower volatilities of major domestic macroeconomic aggregates which is counter to the intuition of international business cycle theory. However, once we control for some external shocks, like terms of trade volatility, results are reversed in favor of the theory. In other words, countries are able to benefit from international risk sharing most ‒ in terms of lower aggregate volatility at home ‒ when this measure of international synchronization is lowest, on average and conditional on external shocks. Another important question addressed in our analysis is whether geographical diversification has any mitigating effect on shocks of various origins. The results indicate that at higher levels of geographical diversification in trade, the adverse effects on output volatility of fiscal, financial and international shocks are significantly reduced while there is no statistically significant impact for money supply shocks. One noteworthy observation from descriptive statistics section is that despite increased geographical diversification measured by HI over the past 5 decades in our sample, countries, on average, have maintained trade with relatively similar type of international partners in terms of 5 See Table A3 in the appendix for the full list of countries. Throughout this paper, “weights” refers to using the relative share of imports or exports of the international trading partners in the calculation of an indicator. 6 5 their level of development and size of the economy; namely the highly advanced economies with large shares in the trade volume of the rest of the world. Findings in this paper have three main implications. One is that geographical diversification of international flows of goods and services is an important measure of globalization, which has not only expanded with the rise in trade openness but has a significant impact on macroeconomic performance. Second, since the estimated impact of enhanced diversification by only 10 percentage points in the sample is a reduction in output volatility of about 30 percent, this implies that trade diversification ‒ suitably conditioned ‒ could be a potentially important policy goal to increase domestic welfare. Finally, for stabilizing purposes, economic characteristics of trade partners, like the size of their GDP and development level, over which trade is diversified are also important influencers and might be the key in understanding the mechanisms through which geographical diversification smoothes business cycle volatility. The remainder of this paper is organized as follows: Section 2 and 3 describe the data and variables, and present some stylized facts about evolution of the main variables of interest over time and across different country samples. Section 4 contains the empirical analyses that examine three different relations between volatility and diversification variables. Section 5 summarizes our findings followed by references and appendix. 2 Data and Variables The basis for our empirical analysis is a panel data set of annual observations for 133 countries over 1962-2006. The entire sample period is then divided into 5 sub-periods with a length of 9 years each. For the variables representing each period, either the beginning of period values, period averages or standard deviations of the log differences or de-trended series are used. For aggregates such as GDP per capita, used to characterize individual countries, the beginning of period value is used to represent a country's development status for example; for the Herfindahl index that captures geographical diversification, period averages are taken; and for volatility measures, we apply the standard deviation of the log differences, growth rates or de-trended variables for each sub-period. In this, we follow conventions in the existing empirical literature. 6 This paper relies heavily on bilateral international trade data for the construction of trade variables, including the Herfindahl index (HI) for geographical diversification, and other variables that capture various characteristics of international flows. The source for the bilateral trade data is Dyadic Trade Data which is based on the International Monetary Fund's Direction of Trade Statistics. The obtained raw trade flow data is organized into 58 annual matrices of bilateral trade for the period 1950-2007 where rows and columns represent the flow of trade from one country to another in current U.S. dollars. The bilateral trade matrices are used to find the relative shares of trade ‒ either imports or exports ‒ with each other country that are applied as weights for computing the HI and other trade indicators, in combination with national income data from Penn World Table, as described below. We also use as control variables some series from the World Bank's World Development Indicators database. For the identification process, instrumental variables are from Andrew Rose's compiled dataset that contains many geographical and cultural time invariant variables. A complete reference to data sources is included in the appendix. The panel data that is constructed consists of 133 countries and the main criteria in choosing them were that a) a country maintain its independence throughout the sample period of 19502007, and b) bilateral trade data be available for the most part of the period. For example, the states that became independent following the dissolution of former Soviet Union in 1991 are excluded from the panel. Further, as it is customary in almost all empirical business cycle volatility literature, with some degree of arbitrariness, the entire period is divided into 5 subperiods of 9 years each. While in most papers periods of 10 years' length are used, this would lead to some loss in informative data in more recent years or in earlier periods. Our organization of the data is such that the panel makes use of most of the available data and therefore led to the following 5 sub-periods: 1) 1962-70, 2) 1971-79, 3) 1980-88 4) 1989-97 and 5) 1998-2006. The construction of the main variables follows. 2.1 Business Cycle Volatility There are two main approaches to the construction of BCV measures in the literature. In some analyses, the volatility variable is calculated as the standard deviation of the cyclical component of the time series or of its growth rate over past quarters or years. Therefore, the constructed 7 variable is a continuous function, like a moving average, as in Blanchard & Simon (2002). Elsewhere, which is the case for most cross-sectional and panel data analyses like ours, the whole period is divided into sub-periods and the volatility is computed as the standard deviation of the growth rate or of the de-trended time series in each period, hence allowing the characterization of variations across periods. For example, the period covering 1971-79 in our panel data has the most pronounced output volatility among the full sample period, likely due to the contemporaneous oil supply shocks. The latter approach has the advantage that the length of each period can be altered to check the robustness of the results subject to the choice of sub-sample, or it can also be reduced to a cross-sectional analysis by taking each sub-period as one and examine the evolution of the relationship over time as is done in some papers. In most cases that we analyze, our estimated relations are robust to such changes. The de-trending method for the underlying macroeconomic data series, and choosing the corresponding parameters, and whether to use per capita or total values are also among the choices that must be made in constructing BCV variables. Dominant in the empirical literature is use of the standard deviation of the growth rate of per capita output or its components, over each sub-period. Alternatively, the standard deviation of the filtered time series, usually using Hodrick-Prescott (HP) filter with a weighting parameter value of 100 recommended for annual data, is used. In this paper, BCV is computed based on all four methodologies. Although the regression results are sometimes sensitive to this choice, the overall pattern remains the same. This paper follows the dominant approach of using the standard deviation of the growth rate of per capita aggregate in the remainder of this paper. Table A1 in the appendix displays the correlations among various BCV measures. Formally, for any country i at any period t, BCV of GDP per capita is defined as: 𝐵𝐶𝑉𝑖,𝑡 ≜ 𝑆. 𝐷. {∆ 𝑙𝑜𝑔(GDP𝑃𝐶𝑖,𝑠 )}, 𝑠 ∈ {𝑡 − 8, 𝑡}, 𝑡 ∈ {1970, 1979, 1988, 1997, 2006} (1) 2.2 Herfindahl Index for Geographical Diversification in Trade The Herfindahl Index for geographical diversification is defined as the sum of the squares of imports (exports) of each trade partner as a share of total imports (exports) of the country in 8 question. Obviously this and other trade based indices, all have an import and an export counterpart. By definition, the Herfindahl index for imports and exports ‒ henceforth HIM and HIX ‒ can vary between zero and unity with a larger number indicating less diversification. In the following formula (Mj/M) stands for the relative share of imports from country j to country i for which, geographical diversity in imports measure is being constructed: 𝑀j 2 𝐻𝐼𝑀𝑖 = ∑ ( ) , 𝑀 𝑗 = 𝑖𝑛𝑑𝑒𝑥(𝑎𝑙𝑙 𝑡𝑟𝑎𝑑𝑒 𝑝𝑎𝑟𝑡𝑛𝑒𝑟𝑠) (2) 𝑗 Here 𝑀(= ∑𝑗 𝑀𝑗 ) is total imports of country i from all trading partners. The Herfindahl index for exports (HIX) is computed analogously. It should be noted that these measures of trade diversification convey no information about the characteristics of trade partners. Two countries might have same HI but that does not tell us anything about the type of countries they are trading with, whether trade partners are relatively large or developed economies for example, although these features might also be important influences on volatility. A country might gain more in lower volatility by having commercial relations with a large and highly developed economy like the U.S. rather than with a smaller and less developed country with the same trade share. In terms of the expected impact of these measures of diversification on output volatility, higher diversification means a more evenly distributed trade among a given number of international partners, which we hypothesize would reduce the aggregate risk of specific international shocks and permit more channels of borrowing and lending in the wake of a shock of any specific origin. 2.3 Size of International Markets We use the weighted average GDP of the economies that trade is conducted with as a proxy for the size of international markets that a country has access to, with the weights being the shares in total exports or imports. TP𝐺𝐷𝑃𝑖 = ∑(𝑆𝑗 . 𝐺𝐷𝑃𝑗 ) , 𝑗 = 𝑖𝑛𝑑𝑒𝑥(𝑎𝑙𝑙 𝑡𝑟𝑎𝑑𝑒 𝑝𝑎𝑟𝑡𝑛𝑒𝑟𝑠) (3) 𝑗 9 Here Sj is the import or export share of partner j in the total imports or exports of country i. In theory, the higher the value of TPGDP, it is more likely that the domestic country is able to import excess domestic demand from abroad or export excess supply to international markets through international capital flows. Therefore, ceteris paribus, we expect this variable to have a stabilizing effect on domestic aggregate fluctuations. 2.4 Development Level of Trading Partners Trading with more developed countries can potentially be beneficial in various ways and is indicative of certain administrative capabilities of a country; establishing and conducting commercial relations with more advanced economies requires maintaining higher standards in various stages of doing business. In the long-run, such trades can influence own growth through technological transfers. Therefore, we expect this variable might proxy for a wide range of institutional and administrative capacities associated with lower levels of output volatility at home country. TP𝐺𝐷𝑃𝑝𝑐𝑖 = ∑(𝑆𝑗 . 𝐺𝐷𝑃𝑝𝑐𝑗 ) , 𝑗 = 𝑖𝑛𝑑𝑒𝑥(𝑎𝑙𝑙 𝑡𝑟𝑎𝑑𝑒 𝑝𝑎𝑟𝑡𝑛𝑒𝑟𝑠) (4) 𝑗 In the above formulation, TPGDPPC is the weighted average GDP per capita of trade partners where, as before, the weights Sj's are their import or export shares in the totals of the country for which the indicator is being constructed for. It should be noted that in the computation of both TPGDPPC and TPGDP, international partners' output and output per capita from the year 2000 are used in order to control for their growth and focus on the type or size of the economies that trade is being conducted with. 2.5 Business Cycle Volatility of Trading Partners This variable can be thought of as a measure of the magnitude of exposure to international shocks that a country is facing as a result of trade, computed as the weighted average of trading partners' business cycle volatilities. The latter variable is calculated by taking the standard deviation of the growth rate of GDP per capita over the past 9 years, which is consistent with the measure of BCV that we are using for the country itself in empirical analyses. 10 TP𝐵𝐶𝑉𝑖 = ∑(𝑆𝑗 . 𝐵𝐶𝑉𝑗 ) , 𝑗 = 𝑖𝑛𝑑𝑒𝑥(𝑎𝑙𝑙 𝑡𝑟𝑎𝑑𝑒 𝑝𝑎𝑟𝑡𝑛𝑒𝑟𝑠) (5) 𝑗 A higher TPBCV for home country implies that it is trading with, on average, relatively more economically volatile countries, potentially subjecting demands for its exports or supply of its imports to higher fluctuations, which in turn may be reflected in higher GDP volatility at home. A similar variable appears in Bacchetta et al (2007). 2.6 IBC Synchronization The construction of this new variable is an attempt to measure the level of business cycle synchronization of a country with respect to its international trading partners. International business cycle synchronization or IBCSync, is basically a multilateral correlation variable that describes how a country's business cycles co-move with those of its trading partners. Formally, the construction of this variable involves first, computing the cyclical component of the real output time series for each country using HP filter and then calculating the correlations between the cyclical components of the country in question with those of all its trading partners. Next, these bilateral correlations are averaged with weights according to their shares (Sj) in the total volume of trade (sum of exports and imports or either of them) of country i. 𝐼𝐵𝐶𝑆𝑦𝑛𝑐𝑖 = ∑(𝑆𝑗 . 𝐶𝑜𝑟𝑟(𝐵𝐶𝑖 , 𝐵𝐶𝑗 ) ) , 𝑗 = 𝑖𝑛𝑑𝑒𝑥(𝑎𝑙𝑙 𝑡𝑟𝑎𝑑𝑒 𝑝𝑎𝑟𝑡𝑛𝑒𝑟𝑠) (6) 𝑗 A high value for IBCSync implies that the country's cyclical component of GDP is closely following those of its partners or the country is, on average, trading with economies whose cyclical behavior is similar to its own. The idea for constructing this variable was that risk sharing and hence smoothing of business cycles are more likely, the lower the correlation of a country's business cycles with those of its trade partners. This is possible, at least in theory, by borrowing from booming trading partners when the country itself is experiencing recession and on the other hand the domestic country's boom phase impact for output volatility can be curbed by lower demand for its exports from the rest of the world if they are in recession. Hence from this perspective, being less internationally synchronized should help smooth business cycles at home. As will be shown in section 4.5 the 11 preliminary empirical results are consistent with this hypothesis, only after controlling for one source of volatility in net exports such as term of trade. 3 Descriptive Statistics This section provides a descriptive summary of the main variables of interest based on sample averages for trade diversification using our newly constructed trade variables. As discussed, the panel data consists of 133 countries over the time span of 1962-2006 divided into five subperiods of 9 years: 1962-70, 1971-79, 1980-88, 1989-97 and 1998-2006. We also look at different groups of countries separately. Tables 1-4 report sample averages for the full sample, and also for the OECD, Latin America, and for “other countries” which excludes the latter two groups from the full sample. BCVY and BCVC stand for the business cycle volatility measures of output and consumption, respectively. Consumption volatility is studied to measure the degree to which trade diversification improves consumption smoothing through inter-temporal trade opportunities. Most trade variables have import and export versions and this is indicated by a suffix “.m” or “.x” at the end of each variable. The behavior of BCV statistics over alternative sub-periods may reflect the macroeconomic experience of the world economy since the Second World War. For almost all four country groupings (Tables 1-4), including the full sample, the cross-country average BCVY peaks during the 1971-79 period, the era of oil supply shocks. This period also is characterized by a greater measure of business cycle synchronization for countries in each grouping. For example, the weighted average business cycle correlations across trading partners jumped from about 0.04 in 1960s to about 0.13 in 1970s for the entire sample and continued rising thereafter. This rise in international synchronization during 1970s is visible across all four country groupings and even within the highly inharmonious Latin America group, the only group with an average negative IBC stance in the 1960s (a negative IBCSync variable means that the country's business cycle is moving on the opposite phase of its trading partners, on average). For Latin America (Table 3), the indicator rises to levels of synchronization during 1970s that are only seen towards the beginning of 21st century in the rest of the world. Overall, these are all signs of a dominant global shock. 12 As Tables 1-4 suggest, the magnitude of the BCV of GDPPC declines after the 1970s subperiod, and in 3 cases it settles at levels lower than those that prevailed during the initial period of 1960s. This decline relative to the initial period is nearly 9% for the entire sample (Table 1), and about 30% for the OECD group (Table 2), while the Latin America countries (Table 3) settle at almost the same levels of output volatility that existed in 1960s, but it still is 25% lower than those in 1970s. This general moderation, registered by our relatively large panel dataset, is more pronounced by the OECD group. Cross-grouping comparison reveals that, not surprisingly, the OECD group has the lowest aggregate volatility during all sub-periods, with volatility levels about half those of the full sample. The following observations are worth making regarding consumption volatility: First, contrary to the notion of consumption smoothing, for the entire sample the volatility of consumption component is, on average, higher than that of output. The relative volatility of consumption to that of output (BCVC/BCVY), a measure of consumption smoothing, is below the critical value of unity only for the most advanced (OECD) countries as shown in Table 2, on average. The most anomalous case belongs to the Latin America group (Table 3) during 19891997 sub-period, which follows the 1980s debt crises and the 1994 speculative crisis that hit these economies. Interestingly, these crises seem to have the highest impact on consumption rather than output volatility. To summarize, consumption smoothing is simply not occurring according to the predictions of macroeconomic theory, for most of the countries in the sample. In addition across Tables 1-4, except for the OECD group, the volatility of consumption reaches its high-point during the period 1989-1997. The international supply shocks of 1970s seem to be a potential reason for increased output volatility while the financial crises of 1980s through 90s might account for the large volatility of consumption. This observation suggests that in trying to explain the consumption smoothing anomaly one should look at financial factors that affect households’ consumption-saving behavior. Finally, for all four country groupings, the relative volatility measure (BCVC/BCVY) has remained more or less at the same levels over time in the past 5 decades, with relatively little intra-group variations. In other words, despite all innovations and developments in the financial sector, domestic or international, and integration of international commercial and financial markets that globalization has brought about, we do not see any obvious improvement for 13 consumption smoothing. We leave further analysis of this feature of the data for future work, and focus on output volatility as a measure of BCV in assessing the role of trade diversification. Trade openness or the total volume of trade (X+M) to GDP has increased more than 40% over the entire sample on average since 1960s as can be seen in Table 1. We also see steady declines in Herfindahl indexes for imports and exports, indicating a general movement towards higher geographical diversification for both import and export exchange. Yet, as Table 1 suggests, exports have systematically enjoyed higher levels of diversification while we see frequent instances of temporary increased concentration for exports in the averages over time. The individual time series for both indexes can actually be quite volatile for specific countries in some cases, as Figure series A1 in the appendix for a select group of countries illustrates. They correspond to the annual, non-averaged HI time series for the entire period of 1960-2007. It should be noted that portrayal of this variable on the same scale even for such advanced countries like the United States (USA) and the Great Britain (GBR) would not be insightful as, for example, GBR's variations in HIX are within the lowest band of the diagram for USA. Next we study the variables that capture trade partner characteristics. Reported TPGDP is in 6 10 format for simplicity, and both TPGDPPC and TPGDP values here are based on output per capita and output from the year 2000 for all periods. This is done to focus only on the type of economies that each country is trading with; GDP per capita being a proxy for development level and GDP being a measure of the size of the trading partner and proxy for the size of international markets. Admittedly, this approach has the disadvantage of ignoring the big changes that some countries have gone through over the last 5 decades. Nonetheless, our empirical results are robust to the choice of year ‒ either current or year 2000 ‒ for the construction of these indicators. Latin America (Table 3) seems to enjoy commercial relations with richer and larger countries, on average, than OECD and the “other countries” group do. This is at least in part due to being located in the same continent as the largest economy in the world (the United States) which has a relatively high share of trade with these countries. We have constructed the relative shares of the U.S. and also G-7 countries7 in the total trade of each country. As Table A2 in the appendix 7 G-7 countries include: Canada, France, Germany, Italy, Japan, United Kingdom, and United States. 14 shows, the average share of the U.S. in the total trade of Latin America countries is about 36% as compared to 15% for the full sample. A notable observation emerges after reviewing the last three variables. In general, TPGDP and TPGDPPC do not seem to follow a clear trend when we consider both export and import versions while a downward trend for HIX and especially HIM is more prominent, on average. In addition, quantitatively, the variation coefficients for TPGDP and TPGDPPC are between 3-5%, while the same normalized measures of variation for HIM and HIX are between 11-18%, respectively, for the entire sample and over 5 sub-periods in Table 1. In other words, despite measurable diversification increases by countries reflected in their Herfindahl indexes, average characteristics of trading partners show higher stability and no clear trend. These observations point to the possibility that international commercial relations have been maintained with countries of relatively similar economic size and development level throughout the globe and over 5 decades, and these trade partners are the most advanced and largest economies in the world, on average. The TPBCV variable averages suggest that the mean exposure to business cycle volatility from trading partners does not show much variation over time and across different groups (Tables 1-4), with the exception of the period 1971-79 in which it reaches the maximum. Average output volatility of trade partners declined since the 1970s and has settled at levels that are quite comparable to our sample's initial period in 1960s. Perhaps more importantly, this variable is lower in value than the BCV level of a typical country during any period. More specifically, while the average output volatility (BCV) of countries in the full sample (Table 1) is about 5.4%, the TPBCV is only around 2.3%, indicating a diversification across economies that are at least half as volatile as the country itself, on average. 15 Tables 1-2: Average Period Statistics Full Sample of 133 Countries 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average BCVY 0.047 0.063 0.056 0.059 0.043 0.054 BCVC 0.058 0.077 0.075 0.093 0.063 0.074 BCVC / BCVY 1.30 1.22 1.34 1.53 1.48 1.38 (X+M)/GDP .61 .65 .66 .71 .85 .70 HIX 0.220 0.183 0.173 0.173 0.181 0.185 HIM 0.195 0.153 0.145 0.140 0.126 0.151 TPGDPPC.x 22278 22707 22076 23868 23389 22894 TPGDPPC.m 22139 21845 21766 23449 22017 22257 TPGDP.x 2658 2767 2760 2801 3098 2824 TPGDP.m 2792 2524 2511 2651 2590 2610 TPBCV.x 0.027 0.043 0.032 0.027 0.025 0.031 TPBCV.m 0.026 0.044 0.033 0.027 0.027 0.031 IBCSync.x 0.042 0.135 0.138 0.142 0.207 0.138 IBCSync.m 0.044 0.120 0.141 0.156 0.211 0.140 Sample of 28 OECD Countries 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average BCVY 0.028 0.035 0.029 0.029 0.020 0.028 BCVC 0.028 0.028 0.027 0.024 0.018 0.025 BCVC / BCVY 0.95 0.84 0.95 0.83 0.91 0.89 (X+M)/GDP .32 .40 .46 .58 .80 .51 HIX 0.150 0.145 0.133 0.133 0.131 0.138 HIM 0.152 0.136 0.134 0.130 0.118 0.134 TPGDPPC.x 22539 21812 21629 25565 25932 23528 TPGDPPC.m 22108 21517 21569 25827 25379 23314 TPGDP.x 2471 2326 2364 2554 2758 2497 TPGDP.m 2516 2212 2274 2619 2554 2437 TPBCV.x 0.024 0.041 0.030 0.025 0.020 0.028 TPBCV.m 0.024 0.046 0.030 0.024 0.022 0.029 IBCSync.x 0.191 0.300 0.345 0.324 0.380 0.310 IBCSync.m 0.197 0.287 0.350 0.338 0.361 0.309 16 Tables 3-4: Average Period Statistics (continued) Sample of 33 Latin America Countries 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average BCVY 0.037 0.054 0.057 0.047 0.039 0.047 BCVC 0.053 0.063 0.076 0.104 0.061 0.072 BCVC / BCVY 1.51 1.29 1.32 1.99 1.35 1.49 (X+M)/GDP .62 .66 .65 .75 .83 .70 HIX 0.241 0.235 0.247 0.240 0.255 0.244 HIM 0.218 0.175 0.191 0.206 0.179 0.194 TPGDPPC.x 26823 25964 25913 26746 25926 26260 TPGDPPC.m 26423 24272 24189 25936 23773 24879 TPGDP.x 4636 4526 4708 4661 4780 4664 TPGDP.m 4751 4001 4314 4781 4422 4449 TPBCV.x 0.025 0.034 0.033 0.025 0.023 0.028 TPBCV.m 0.025 0.042 0.036 0.026 0.027 0.031 IBCSync.x -0.057 0.232 0.102 0.027 0.363 0.141 IBCSync.m -0.055 0.174 0.128 0.047 0.382 0.144 Sample of 78 “Other Countries”8 8 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average BCVY 0.061 0.076 0.065 0.073 0.051 0.066 BCVC 0.073 0.101 0.092 0.114 0.080 0.093 BCVC / BCVY 1.37 1.34 1.50 1.61 1.73 1.52 (X+M)/GDP .71 .73 .74 .74 .87 .76 HIX 0.245 0.182 0.162 0.168 0.178 0.185 HIM 0.209 0.155 0.136 0.125 0.114 0.145 TPGDPPC.x 20532 21956 20925 22375 21731 21549 TPGDPPC.m 20643 21187 21062 21849 20316 21025 TPGDP.x 2039 2342 2229 2301 2696 2336 TPGDP.m 2223 2147 1980 1972 2013 2059 TPBCV.x 0.029 0.047 0.033 0.028 0.028 0.033 TPBCV.m 0.027 0.044 0.033 0.029 0.029 0.032 IBCSync.x 0.019 0.036 0.069 0.111 0.097 0.071 IBCSync.m 0.018 0.039 0.063 0.124 0.101 0.073 “Other countries” sample consists of the full sample countries excluding those in OECD or Latin America groups. 17 Finally, our measures of international business cycle synchronization (IBCSync) reported in the last two rows of Tables 1-4 show that the weighted average correlations of business cycles among trading partners have risen to levels many folds those that prevailed in the 1960s. Interestingly, the OECD countries' synchronization in the 1960s (Table 2, column 1), was at the levels that the whole sample, on average, has reached only today (Table 1, column 5) and not surprisingly they have been the most internationally synchronized group throughout all periods, by this measure. Latin America on the other hand, is the only group that experienced a negative IBCSync value in the 1960s, and ended up with levels comparable to those of OECD group. To summarize, generally speaking, despite expansion of geographical diversifications in the broadest sense as measured by HI, countries seem to be trading with relatively similar economies, over time, as reflected in the average GDP, GDP per capita and volatility measures of their trading partners. Trade partners are more developed, larger in size and more stable than the country itself, on average. This observation is consistent with the fact that the highly industrialized nations have retained rather large shares in the trade volume of other countries, revealing the important role that these advanced nations consequently play in the domestic economy of other nations. We finally examine the unconditional relationship between our measure of geographical diversification (HI) against business cycle volatility of output and also against GDP per capita in Figures 1-2. For the sake of brevity, only diversification in exports in the last sub-period (19982006) is portrayed. Figure 1 depicts the scatter plot of the HIX-BCV relationship which shows a positive unconditional correlation as confirmed by the simple regression line; countries with less geographical diversification (or higher HIX) tend to have higher macroeconomic fluctuations. In this diagram, OECD countries are denoted by black hollow squares and are clearly centered towards the origin of the diagram, indicating that they enjoy relatively lower output volatility and higher diversification in trade, on average. On the other hand dark circle and gray diamond markers belong to Latin America and the remaining countries in the entire sample of 133 countries, respectively. Finally, as Figure 2 shows, HIX-lnGDPPC exhibits a negative relationship for the last sub-period, as one might have expected, namely that more developed countries tend to have higher geographical diversification. 18 Figures 1-2: Scatter Diagram of Herfindahl Index for Geographical Diversification in Exports against BCV and GDPPC, Period 1998-2006. .8 .6 HIX .4 .2 0 0 .05 .1 .15 BCV of GDPpc .2 .25 7 8 9 log of GDPpc 10 11 .8 .6 HIX .4 .2 0 6 Notes: Black hollow square, dark circle and gray diamond markers denote OECD, Latin America and the remaining countries in the full sample of 133 economies, respectively. 19 4 Empirical Analyses The purpose of the analyses in the next sections is to answer the following questions: Does geographical diversification in international trade reduce business cycle volatility of macroeconomic aggregates? Diversifying trade with respect to what type of country characteristics most promotes stabilization of business cycle fluctuations? Does geographical diversification mitigate the adverse impact of external and internal idiosyncratic shocks, and if so what type of shocks? 4.1 Empirical Approach In examining the effect of geographical diversification on macroeconomic volatility, the most important source of endogeneity, besides the measurement error, is likely to be simultaneity. Specifically, it is possible that if a country experienced economic instability, policy makers might embark on efforts to expand international trade diversification, through measures such as engaging in bilateral, multilateral or regional trade agreements. Thus, the explanatory variable HI should be instrumented by appropriate exogenous variables. Fortunately there are several instrumental variable (IV) candidates for international trade variables that are frequently used in the literature. In practice, we estimate a two dimensional simultaneous equations model (SEM) of the following form: { 𝐵𝐶𝑉𝑖,𝑡 = 𝛼1 + 𝛽1 . 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠1 + 𝛿1 . 𝐻𝐼𝑖,𝑡 + 𝛿2 . 𝑇𝑃𝑘 + 𝑢𝑖,𝑡 𝐻𝐼𝑖,𝑡 = 𝛼2 + 𝛽2 . 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠2 + 𝜆 . 𝐵𝐶𝑉𝑖,𝑡−1 + 𝑣𝑖,𝑡 (7) where: 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠1 = {𝑙𝑛𝐺𝐷𝑃𝑃𝐶, , 𝑃𝑜𝑙𝑖𝑡𝑦, 𝑇𝑟𝑎𝑑𝑒 𝑂𝑝𝑒𝑛𝑛𝑒𝑠𝑠, 𝜎𝐺 , 𝜎𝑀 , 𝜎𝑖𝑛𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙 , … } 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠2 = {𝐺𝑒𝑜𝑖 } 𝑇𝑃𝐾 = {𝑇𝑃𝐺𝐷𝑃, 𝑇𝑃𝐺𝐷𝑃𝑃𝐶 , 𝑇𝑃𝐵𝐶𝑉, 𝐼𝐵𝐶𝑆𝑦𝑛𝑐} In the above equations, after considering a large set of potential control variables, some of them are included in the BCV equation to capture the general economic and political characteristics of a country that might be common influencers for both the dependent and independent variables. LnGDPPC is the log of GDP per capita, and Polity is an index ranging between -10 and 10, indicating most autocratic to most democratic political regimes. Trade openness is the ratio of total volume of trade to gross domestic output and is included so that the 20 effect of diversification can be analyzed for a given level of exposure to trade. On the other hand, σ's capture fiscal, monetary or external shocks, respectively, which are conventionally viewed as major sources of business cycle volatility. The HI equation in (7) is estimated by Geo, a set of cultural and geographical variables such as language dummies that specify whether the official language of a country is Spanish or French, by the average air distance of the capital city of the country from certain cities around the world (NYC, Tokyo and Rotterdam), binaries for being a landlocked or island country, elevation and total area of the country. The predetermined value of consumption volatility, 𝐵𝐶𝑉(𝑐𝑜𝑛𝑠)𝑖,𝑡−1 , is also included as we need at least one time-variant instrumental variable for the panel data estimation and is used instead of output volatility due to the suspicion of autocorrelation. A rich set of potential determinants or covariates of geographical diversification in international trade, their quantitative impact, and underlying mechanisms are discussed in Farshbaf (2012a). Finally TP variables capture various characteristics of trading partners and together with above variables enable us to examine various specifications to answer the research questions that were posed in the beginning of this section. In most of the following sections we apply Instrumental Variable General Method of Moments technique using the panel data described in section 2, and although not reported, period dummies are included throughout all specifications. 4.2 Business Cycle Volatility and Geographical Diversification Our conjecture is that for a given level of exposure to international trade as measured by total volume of trade to gross output, higher diversification among trade partners reduces domestic output volatility due to at least two reasons: First, enhanced geographical diversification means that shares of trade are relatively equitably distributed among many trade partners, so that the impact of disturbances from any individual commercial partner is reduced. In other words, aggregate international risk is decreased. Obviously, this assumes that the shocks to trade partners are in part idiosyncratic and not too highly correlated. Second, any given shock that hits the economy can be mitigated through inter-temporal trade with a rather large network of trade partners. 21 Table 5 summarizes the results, which allows us to address the following question: For given general political and economic characteristics in a country and level of openness to trade, does diversifying in international trade reduce business cycle volatility? The first specification (Table 5, column 1) is our benchmark model, and columns (2) and (3) include the Herfindahl index of geographical diversification for exports and imports respectively. Last two columns examine possible stabilizing effect of diversification on consumption volatility. The estimation method is IV GMM as explained in section 4.1. Although the implementation of IV methodology is out of our concern for simultaneity based on economic reasoning, we have also conducted the Hausman test for endogeneity of HI, which confirms our theoretical concern. As the last two rows in Table 5 illustrate, a higher value for either of Herfindahl indexes, which is equivalent to less geographical diversification across other countries, shows positive and statistically significant association with business cycle volatility in output. A comparison between the results in columns (2) and (3) indicates that the marginal effects of import and export counterparts of Herfindahl index are close in magnitude. Our estimates are also economically meaningful and might have important policy implications. For example if a country raises its export diversification by 10 percentage points (∆HIX = -0.10), which is what some countries in our sample accomplished during a decade or so, the implied reduction in output volatility (using sample mean BCV = 0.045) solely as a result of such efforts is about (0.015/0.045) ≈ 33%. Our measure of economic development level, lnGDPPC does not enter most of the specifications in Table 5 significantly, due to its high correlation with other variables. In Table 5 we also see that greater openness to trade, (X+M)/GDP, is associated with higher levels of output and consumption volatility, confirming the frequently documented findings in the literature that this variable is destabilizing. In addition, fiscal and monetary policy shocks (σG and σM), respectively measured by the standard deviation of growth rate of government expenditures and by the standard deviation of the growth rate in the M2 measure of money over each sub-period in the sample, are consistently and significantly destabilizing for output and consumption growth. 22 Table 5: Dependent variable: Business Cycle Volatility of GDP per capita Consumption per capita (1) (2) (3) (4) (5) lnGDPpc -0.0026 -0.0020 -0.0022 0.0037 0.0029 (0.0020) (0.0027) (0.0022) (0.0042) (0.0042) Polity -0.0014 -0.0012 -0.0014 -0.00283 -0.0028 (0.0003)*** (0.0004)*** (0.0003)*** (0.0007)*** (0.0007)*** (X+M)/GDP σ(G) σ(M) HIX HIM 0.00041 0.00036 0.00048 0.0009 0.0009 (0.00015)*** (0.00020)* (0.00016)*** (0.0003)*** (0.0003)*** 0.071 0.069 0.072 0.167 0.167 (0.013)*** (0.014)*** (0.014)*** (0.028)*** (0.028)*** 0.014 0.012 0.017 0.0221 0.0224 (0.006)** (0.006)* (0.006)*** (0.0123)* (0.0126)* 0.122 0.086 (0.057)** (0.0854) 0.149 0.077 (0.058)** (0.07736) Notes: The numbers reported below each coefficient in the parentheses are robust standard errors with *, ** and *** next to them indicating statistical significance at 10, 5 and 1 percent respectively. Estimation method: IV GMM regression including period dummies. IV's used for trade variables are: dummies for Spanish, French, elevation, total land area, landlocked dummy, distance from certain cities in the world and lag of consumption BCV. Obs. No. in each: 387. Finally, the statistically insignificant coefficient estimates for HIX and HIM in specifications (4) and (5) in Table 5 illustrate that geographical diversification in trade does not smooth consumption, which as we know from section 3, is more volatile than output throughout most of our sample, on average. However, in results not presented here, both HIM and HIX are highly significant stabilizers for domestic absorption ‒ a measure of total consumption by households, government and businesses. Moreover, as discussed in Farshbaf (2012b), once the interaction of trade openness and diversification is introduced in the regressions, the combined impact of this measure of global trade is capable of smoothing all major macroeconomic aggregates, including consumption and its relative volatility to that of output. To summarize this sub-section, our results confirm the empirically well-established adverse effects of trade openness on BCV, as well as that of volatilities in domestic fiscal and monetary policies. Most importantly our conjecture regarding the impact of higher trade diversification for output volatility – but not for consumption volatility – is confirmed. At a given level of trade 23 openness, more geographically diversified trade is associated with lower domestic output volatility which, is likely to be indicative of a causal relationship from HI to BCV. 4.3 Mitigation of Shocks through Geographical Diversification In this section we study the possible mitigation effects of trade diversification on shocks of various origins. In other words we want to see whether higher geographical diversification allows the countries to smooth the impact of specific shocks they are subject to, shocks which might be rooted in domestic policies or be internationally originated. The results are shown in Table 6, which as in Table 5 presents the results of IV GMM estimation methodology with the same set of IV's described for Table 5. Throughout all specifications (1) - (5) some control variables are included. Our empirical strategy is to introduce one type of shock at a time and examine its interaction with the diversification variable. As it will be illustrated below, the estimated coefficients on the shock variable and the interaction term along with the possible range of diversification variable (HI) will capture the mitigation effect. We start by replicating the benchmark model from previous section in column (1). The possible mitigation of geographical diversification for the volatility induced by domestic fiscal policy shocks is examined in specification (2) of Table 6. In the benchmark model (1), the coefficient estimate of σ(G) is positive, implying that higher volatility in fiscal policy increases business cycle volatility. However as in specification (2), once we add the interaction term, the coefficient estimate on σ(G) becomes negative while its interaction with HIX enters the regression with the opposite sign, with the latter being statistically significant even at 1% level. The overall marginal effect of fiscal shocks using the point estimates from specification (2) can be written as: 𝜕 𝐵𝐶𝑉⁄ 𝜕 𝜎(𝐺) ≅ −0.1628 + (1.738) ∙ 𝐻𝐼𝑋 , 𝐻𝐼𝑋 ∈ [0,1] (8) Quantitatively, for a country operating at the least level of geographical diversification or HIX=1, the estimated marginal effect of fiscal shocks on BCV is 1.575. This destabilizing effect falls to a magnitude of 0.073 for the median value of HIX = 0.136, which interestingly is quite 24 comparable to the estimated coefficient of σ(G) throughout other regression settings, where it enters alone without any interaction. In this regard, our estimates seem to be quantitatively and qualitatively consistent. In other words, at high levels of HI, that is lower geographical diversification, fiscal policy volatility is associated with higher macroeconomic volatility, but this impact is mitigated as diversification rises. Table 6: Diversification and Mitigation of Shocks. Dependent variable: BCV of GDP per capita lnGDPPC Polity σ (G) TO HIX (1) (2) (3) (4) (5) -0.0009 0.0009 0.0006 -0.0015 -0.0005 (0.0026) (0.0022) (0.0023) (0.0041) (0.0027) -0.0014 -0.0017 -0.0013 -0.0011 -0.0018 (0.0003)*** (0.0003)*** (0.0004)*** (0.0006)*** (0.0004)*** 0.0756 -0.1628 0.0547 0.0612 0.1187 (0.0185)*** (0.1012)* (0.0207)*** (0.0216)*** (0.0395)*** 0.0005 0.0001 0.00009 0.0001 0.0002 (0.00015)*** (0.00006)** (0.00006) (0.00006)** (0.00005)*** 0.1720 -0.0150 -0.0643 -0.0464 -0.0137 (0.0693)** (0.0567) (0.0711) (0.1137) (0.0474) σ(G) x HIX 1.738 (0.535)*** σ(TOT) -0.240 (0.178) σ(TOT) x HIX 1.997 (1.053)** σ(RER) -0.040 (0.127) σ(RER) x HIX 0.623 (1.044) σ(r) -0.0013 (0.0007)** σ(r) x HIX 0.0104 (0.0054)* obs.# 442 442 283 214 313 Notes: The numbers reported below each coefficient in the parentheses are robust standard errors with *, ** and *** next to them indicating statistical significance at 10, 5 and 1 percent respectively. Estimation method: IV GMM regression including period dummies. IV's used for trade variables are: dummies for Spanish, French, elevation, total land area, landlocked dummy, distance from certain cities in the world and lag of consumption BCV. 25 The volatility of the terms of trade is the most frequently used proxy for international shocks in the literature and in specification (3) we observe that its adverse effect on BCV can also be mitigated through higher geographical diversification. It is worth mentioning that there is a decline in the significance of trade openness, once we control for TOT volatility, which is in line with Rodrik (1998) and Cavallo (2008)'s findings that the major reason behind international trade's destabilizing effect is the terms of trade channel. Another international shock is volatility in the real effective exchange rate, which weights bilateral real exchange rates by the corresponding country shares in trade as constructed by IMF for its IFS database. As reported in column (4) of Table 6, the results follow the same pattern as the above cases. In other words, the destabilizing effect of volatility in real effective exchange rate tends to be mitigated at higher levels of geographical diversification, except that the estimated coefficients are not statistically significant at the conventional levels. Finally, I use variations in the real interest rate to capture the uncertainty or volatility in the financial markets or monetary policies. The last column in Table 6 confirms that, diversity in international trade partners can lessen the adverse effects of this type of volatility on the macroeconomic fluctuations. Again, at low levels of international diversification (HI close to 1) the estimated marginal effect of σ(r) on BCV is a positive number but this destabilizing effect is reduced as diversification expands, or in terms of our variables as HIX falls. Although not reported, measures of nominal monetary policy shocks such as the standard deviation of the growth rate of different measures of money supply do not exhibit statistically significant interaction effects. It should be noted that as the last row of Table 6 shows, the number of observations in the panel for specifications (3) and beyond are much less than the ones we had for specifications (1) - (2) which, might be one reason we get higher standard errors and hence lose some statistical significance for the coefficient estimates. The reason for this decline in the number of observations is the unavailability of data in some periods for variables such as the terms of trade, real effective exchange rate and real interest rates. A similar exercise is conducted in Farshbaf (2012b) in which the interaction of geographical diversification with trade openness is introduced to capture the overall impact of global trade on business cycle volatility. Those results indicate that although the stand-alone impact of trade 26 openness is destabilizing, the overall effect of the diversified or global trade, on major macroeconomic aggregates including consumption, can actually become stabilizing for high levels of geographical diversification. To summarize, the overall results conform with our expectation that higher levels of trade diversification have a mitigating effect on the output volatility induced by various types of domestic and international shocks. Although a structural model is beyond the scope of this paper, the results suggest that higher diversification means there are a wider array of international opportunities with which to manage risk. 4.4 Economic Characteristics of International Trading Partners Next, we include newly constructed variables described in section 2 that capture characteristics of international trading partners with which a country diversifies its trade. In particular, we would like to learn how, for a given level of trade openness and geographical diversification, trading with more developed countries as proxied by GDPPC, with larger economies as captured by GDP and with more stable countries in terms of their BCV, all on a weighted average base, affects macroeconomic fluctuations at home. We follow the same IV GMM estimation strategy due to the potential simultaneity between BCV and trade variables. Estimation results are reported in Table 7. Columns (1) and (3) in Table 7 are our benchmark models in which, the controls along with openness to trade and Herfindahl measure of geographical diversification for exports and imports (HIX and HIM) are respectively included for the purpose of comparison with the specifications that add economic characteristics of trading partners, one at a time. To summarize the findings in Table 7, we observe that trading with more developed countries or larger economies is associated with smoother business cycles at home as shown by the negative and statistically significant coefficients for TPGDPPC and TPGDP, while trading with more volatile economies, on average, has the opposite effect but is only marginally statistically significant. It should be recalled that TPGDPPC and TPGDP variables are constructed using GDP and GDPPC values from the year 2000 to focus on the type of the country with which, trading is 27 being conducted rather than allowing the fact that some countries have grown rich, affect our results. Nonetheless, the conclusions do not change once we allow real growth in output and output per capita of the trading partners; although the results are not reported they are available upon request. Table 7: BCV and Characteristics of Trading Partners. Dependent variable: BCV of GDP per capita lnGDPpc Polity σ (G) σ (M) TO HIX (1) (2) (3) (4) (5) (6) (7) (8) -0.0020 0.0007 -0.0022 0.0017 0.0008 -0.0001 -0.0008 -0.0020 (0.0027) (0.0023) (0.0022) (0.0023) (0.0023) (0.0021) (0.0023) (0.0022) -0.0012 -0.0011 -0.0014 -0.0014 -0.0011 -0.0013 -0.0014 -0.0014 (0.0004)*** (0.0004)*** (0.0003)*** (0.0003)*** (0.0003)*** (0.0003)*** (0.0004)*** (0.0004)*** 0.069 0.068 0.072 0.072 0.075 0.074 0.071 0.070 (0.014)*** (0.014)*** (0.014)*** (0.013)*** (0.013)*** (0.013)*** (0.015)*** (0.015)*** 0.012 0.016 0.017 0.014 0.015 0.015 0.016 0.017 (0.006)* (0.006)** (0.006)*** (0.006)** (0.006)** (0.006)** (0.007)** (0.007)*** 0.0004 0.0004 0.0005 0.0004 0.0002 0.0003 0.0004 0.0005 (0.0002)* (0.0002)** (0.0002)*** (0.0001)*** (0.0002) (0.0001)** (0.0002)** (0.0002)*** 0.122 0.151 0.141 0.139 (0.057)** (0.048)*** (0.053)*** (0.047)*** HIM TPGDPPC.x 0.149 0.123 0.094 0.177 (0.058)** (0.053)** (0.059)* (0.070)** -2.158 (0.696)*** TPGDPPC.m -2.126 (0.724)*** TPGDP.x -0.007 (0.003)*** TPGDP.m -0.005 (0.002)* TPBCV.x 0.292 (0.212) TPBCV.m 0.397 (0.339) Notes: The numbers reported below each coefficient in the parentheses are robust standard errors with *, ** and *** next to them indicating statistical significance at 10, 5 and 1 percent respectively. Estimation method: IV GMM regression including period dummies. IV's used for trade variables are: dummies for Spanish, French, elevation, total land area, landlocked dummy, distance from certain cities in the world and lag of consumption BCV. Obs. No. in each: 387. 28 One surprising result in Table 7 is that, being exposed to more volatile countries on average (TPBCV), is only slightly destabilizing in terms of domestic output, as judged by the statistical significance. This is to a large extent explained by the fact that TPBCV does not vary much over time or across countries, as discussed in section 3. Another interesting result is that the estimated coefficients of HIM lose some degree of magnitude and statistical power when TPGDPPC and TPGDP variables are included in the regressions, nominating access to more developed and larger international markets as stabilizing mechanisms for import diversification. This loss in the explanatory power of the HIM coefficients is most prominent when the TPGDP variable is included, as shown in Table 7 under specification (6), indicating that the stabilizing mechanism of import diversification at least partially works through access to larger markets. This is consistent with one of the main conjectures posed in this paper that international trade diversification across countries provides a wide array of international channels for domestic absorption to smooth domestic output, and this result suggests that the larger these international markets are, the higher are the chances of reducing BCV. One compelling explanation for these results is that trading with a larger number of advanced countries is stabilizing because these countries are the most economically stable countries. To test this hypothesis, regression specifications in the previous table (Table 7) are augmented by controlling for TPBCV or average output volatility of trading partners. Table 8 shows the regression results where only the main variables of interest are reported. The estimated impact of trading with more advanced and larger economies is not reduced in either magnitude or statistical significance, after controlling for average output volatility of trading partners. In other words, trading with economically larger and more advanced nations' strong association with higher stability at home is due to reasons other than their being highly stable. 29 Table 8: BCV and Other Characteristics of Geographical Diversification. Dep.variable: BCV of GDP per capita TO HIX (1) (2) (3) (4) (5) (6) (7) (8) 0.0004 0.0004 0.0004 0.0004 0.0002 0.0002 0.0003 0.0003 (0.0002)** (0.0002)** (0.0001)*** (0.0001)*** (0.0002) (0.0002) (0.0001)** (0.0001)** 0.151 0.145 0.141 0.142 (0.048)*** (0.047)*** (0.053)*** (0.054)*** HIM TPGDPPC.x -2.158 -2.279 (0.696)*** (0.739)*** TPGDPPC.m 0.123 0.126 0.094 0.092 (0.053)** (0.059)** (0.059)* (0.066) -2.126 -2.159 (0.724)*** (0.728)*** TPGDP.x -0.007 -0.008 (0.003)*** (0.003)*** TPGDP.m TPBCV.x -0.193 0.012 (0.208) (0.177) TPBCV.m -0.005 -0.005 (0.002)* (0.003)* -0.010 -0.009 (0.274) (0.277) Notes: The numbers reported below each coefficient in the parentheses are robust standard errors with *, ** and *** next to them indicating statistical significance at 10, 5 and 1 percent respectively. Estimation method: IV GMM regression including period dummies. IV's used for trade variables are: dummies for Spanish, French, elevation, total land area, landlocked dummy, distance from certain cities in the world and lag of consumption BCV. Obs. no. in each: 387. 4.5 International Business Cycle Synchronization In this section we touch on the issue of business cycle smoothing through the introduction of a new index constructed to measure multilateral correlations of home BCs with those of international trading partners or IBCSync. The idea for the construction of this variable is that the chance of smoothing business cycles would be higher, if the country in question and its commercial partners were at different phases of business cycles, so that correlations between home and foreign business cycles are lower. Table 9 summarizes our econometric results using only random effects estimation, as most of the variations in the data seem to be coming from cross rather than within countries. The IBCSync variable used here is based on total trade shares as weights, rather than import or export shares alone. 30 Table 9: International Business Cycle Synchronization & BCV of Various Macroeconomic Aggregates D e p e n d e n t Y lnGDPPC Polity σ (G) TO HI IBCSync v a r i a b l e : B C V o f DA C/Y† Y C I 0.0007 0.0094 -0.0110 0.0100 -0.0293 -0.0020 (0.002) (0.004)** (0.011) (0.003)*** (0.042) (0.003) -0.0013 -0.0028 -0.0016 -0.0016 -0.0173 -0.0012 (0.0003)*** (0.0005)*** (0.0016) (0.0004)*** (0.0062)*** (0.0004)*** 0.12 0.23 0.59 0.18 0.48 0.08 (0.01)*** (0.02)*** (0.07)*** (0.02)*** (0.26)* (0.01)*** 0.0002 0.00023 .00032 .00024 0.00105 .00013 (.00005)*** (.00008)*** (.00024) (.00006)*** (.0009) (.00059)** 0.033 0.057 0.140 0.041 0.014 0.025 (0.016)** (0.027)** (0.077)* (0.022)* (0.303) (0.018) -0.0124 -0.0223 -0.0741 -0.0172 -0.2429 0.0071 (0.007)* (0.012)** (0.033)** (0.009)** (0.131)* (0.007) σ (TOT) 0.081 (0.026)*** R-sq within 0.14 0.08 0.13 0.14 0.01 0.11 R-sq between 0.47 0.61 0.35 0.51 0.26 0.43 obs. # 570 570 570 570 570 291 Notes: The numbers reported below each coefficient in the parentheses are robust standard errors with *, ** and *** next to them indicating statistical significance at 10, 5 and 1 percent respectively. Estimation method is Random Effect. † By C/Y notation we mean relative BCV of consumption to that of output. We consider business cycle volatility measures of different aggregates, where all are on real per capita bases and business cycle volatility is calculated as the standard deviation of the growth rate of each aggregate over the sub-period of 9 years length. In particular, the dependent variables are the business cycle volatilities of output, consumption, investment, domestic absorption (C+I+G), the relative volatility of consumption to that of output and finally, output again in a different specification, respectively from the left to right columns. Domestic absorption is an important variable to consider since it measures the overall domestic consumption which, domestic agents including households, government and businesses seek to smooth via import and export channels. The results in Table 9 appear to contradict the conjecture that countries can benefit most in terms of business cycle smoothing if their business cycles are least correlated with their trading 31 partners. To put it another way, being synchronized with trading partners is seemingly associated with less volatility in own business cycles for any component of aggregate demand and also the relative volatility of consumption, after controlling for general country specific characteristics and measures of international trade integration that includes trade openness and geographical diversification. However, in the last column of Table 9, we include a measure of exposure to international shocks, namely the volatility of the terms of trade and this reverses our results. Not only does the IBCSync variable lose its stabilizing status, it now becomes relatively destabilizing for the volatility of output per capita, although with relatively low levels of significance. Overall, the results suggest that controlling for an external shock, it is optimum to trade with countries that are different from us in terms of business cycle behavior. While this new variable captures an important feature of international trade relations, its further study is left for future work. 4.6 Robustness Check A number of alternative approaches were considered throughout our empirical work to make sure the results are not driven by the methodologies applied here. Specifically, our main results, namely geographical diversification's being an economically and statistically significant stabilizing factor on business cycle fluctuations, proved to be robust to at least: 1. The measure of BCV of output; whether being the standard deviation of the growth rate of output per capita or that of the HP-filtered time-series. 2. The choice of the sub-period length. In particular, we divided the whole sample period into two sub-periods and then nine sub-periods – as done in Farshbaf (2012b) – instead of five, and the results follow the same pattern. 3. Control variables; a large set of control and instrumental variables were considered in the regression analyses. 5 Summary and Conclusions This paper empirically investigates the impact of geographical diversification in imports and exports on business cycle volatility of output. Despite being subject to extensive examination in the literature, the BCV-trade relation still has many unexplored gaps of which an analysis of geographical diversification is one. Our empirical analysis follows an IV strategy mostly due to the belief that there is a simultaneity between BCV and trade variables. We showed that for a 32 given level of trade openness, diversifying international trade more evenly across more countries in the sense captured by Herfindahl index for either imports or exports, lowers output volatility significantly, and lowers consumption volatility but this relation is not statistically significant. Further, through the introduction of four new variables, we found that for a given level of trade openness and geographical diversity, trading with economies that are more developed, larger and to some degree more stable are consistently associated with smoother output fluctuations. Finally, once a major external shock is controlled for, being less internationally synchronized, as captured by lower IBCSync, is shown to be associated with lower macroeconomic volatility at home. One important observation from the stylized facts laid out in the Descriptive Statistics section was that although the degree of diversification in international trade has, on average across nations, expanded notably over the last five decades, countries seem to have maintained their commercial relations with relatively similar economies in terms of their trading partners' level of economic development, size of their economy and their stability, over time. This is further confirmed by the fact that the most industrialized nations have maintained a rather high share in the total trade volume of other nations. In future research, we plan to study the structural mechanisms through which geographical diversification per se and trading with more advanced, larger and more stable economies reduce business cycle volatility by extending a theoretical framework for studying diversification and macroeconomic volatility. In addition, we plan to examine the connection between geographical diversification and sectoral concentration, for example to examine how trade patterns change when a country diversifies trade in favor of more advanced countries as opposed to less developed ones. Finally we plan to analyze the consumption smoothing anomaly found for most countries in our sample; namely, that consumption volatility exceeds output volatility. 33 References Backus, D.K., Kehoe, P.J. and Kydland, F.E. 1992. “International Real Business Cycles”. Journal of Political Economy 100 (4), 745–775. Barlevy, G., 2004. “The Cost of Business Cycles Under Endogenous Growth.” American Economic Review. 94 (4), 964–990. Bacchetta, Jansen, Piermartini and Amurgo-Pacheco, 2007. “Export Diversification As an Absorber of External Shocks.” (Preliminary Draft). Blanchard, Olivier J. and John Simon, 2002. “The Long and Large Decline in US Output Volatility,” forthcoming in the Brookings Papers on Economics Activity. Buch, C.M., Döpke, J. and Pierdzioch, Ch. 2005. “Financial Openness and Business Cycle Volatility.” Journal of International Money and Finance, 24(5): 744–65. Cavallo, E., 2008. “Output Volatility and Openness to Trade: A Reassessment”, Economía, Vol. 9, No. 1, Fall, 105-152. Di Giovanni, J., Levchenko, A., 2009. “Trade Openness and Volatility.” The Review of Economics & Statistics, Aug., Vol. 91 (3), 558-585. Easterly, W., Islam, R. and Stiglitz, J.E. 2001 “Shaken and Stirred: Explaining Growth Volatility.” Annual World Bank Conference on Development Economics, ed. by Boris Pleskovic and Nicholas Stern (Washington: World Bank). Farshbaf, Arian, 2012a. “Geographical Diversification in International Trade and Its Determinants”. Dissertation chapter. Department of Economics, USC. Farshbaf, Arian, 2012b. “Business Cycle Volatility and Economic Globalization”. Dissertation chapter. Department of Economics, USC. Fatás, A., and Mihov, I. 2001. “Government Size and Automatic Stabilizers: International and Intranational evidence”, Journal of International Economics, Volume 55, Issue 1, Oct., 3‐28. 34 Ferreira da Silva, Gisele, 2002. “The Impact of Financial System Development on Business Cycles Volatility: Cross-Country Evidence.” Journal of Macroeconomics, Vol. 24, Issue 2, June, 233-253. Furceria, and Karras, 2007. “Country Size and Business Cycle Volatility: Scale Really Matters.” Journal of the Japanese and International Economies, Volume 21, Issue 4, Dec., 424‐434. Hodrick, R., and Prescott, E.C. (HP),1997. “Postwar U.S. Business Cycles: An Empirical Investigation” Journal of Money, Credit, and Banking, 29 (1), 1–16. Karras, G., and E Song, 1996, “Sources of Business-Cycle Volatility: An Exploratory Study on a Sample of OECD Countries.” Journal of Macroeconomics, Vol. 18, No. 4, 621-37. Kose, M. Ayhan, Prasad, E., and Terrones, M.E. (KPT) 2006. “How Do Trade and Financial Integration Affect the Relationship between Growth and Volatility?” Journal of International Economics 69(1): 176–202. Kose, Otrok and Whiteman, 2003. “International Business Cycles: World, Region, and Country‐Specific Factors”, The American Economic Review, Vol. 93, No. 4 (Nov., 2003), 1216‐1239. Krusell, P., Smith, A., 1999. “On the Welfare Effects of Eliminating Business Cycles.” Rev. Econ. Dynam. 2 (1), 245–272. Lane, Philip R., and Giani Maria Milesi-Ferretti. 2001. “The External Wealth Of Nations: Measures Of Foreign Assets And Liabilities For Industrial And Developing Nations.” Journal of International Economics 55:263–94. Lucas Jr., R.E., 1987. “Models of Business Cycles.” Blackwell, New York. Malik, A. and J. Temple, 2009.“The Geography of Output Volatility”, Journal of Development Economics Volume 90, Issue 2, Nov., 163‐178. Mendoza, E., 2000. “Terms-of-Trade, Uncertainty, and Economic Growth..” Journal of Development Economics 54 (2), 323–356. 35 Prasad, Rogoff, Wei and Kose. 2007. “Financial Globalization, Growth and Volatility in Developing Countries.” Globalization and Poverty, Harrison, A. (ed.). Chicago, IL: University of Chicago Press, 2007. Ramey, G., and V.A. Ramey, 1995. “Cross-Country Evidence on the Link Between Volatility and Growth.” American Economic Review, Vol. 85, 1138-51. Rose, Andrew K. 2006 “Size Really Doesn’t Matter: In Search of a National Scale Effect.” Journal of the Japanese and International Economies 20 (4): 482-507. Rodrik, Dani, 1998. “Why Do More Open Economies Have Bigger Governments?”, Journal of Political Economy 106(5): 997–1032. Sørensen, B. E. and Yosha, O., 1998 “International Risk Sharing and European Monetary Unification.” Journal of International Economics, Aug., 45(2), 211-38. Sutherland, A., 1996, “Financial Market Integration and Macroeconomic Volatility.” Scandinavian Journal of Economics, Vol. 98, pp. 521-39. Data Sources Bilateral Trade Data: Source: Dyadic Trade Dataset for the Correlates of War Project Description: Integrating bilateral trade data from previous data projects and supplemented with additional sources when missing. Based on: Majority from IMF's Direction of Trade Statistics. Supplemented with data from Barbieri, Keshk & Pollins (“Trading Data: Evaluating our Assumptions and Coding Rules.”) & Barbieri’s International Trade Dataset. Notes: All entries are in current U.S. Dollars. For detailed description of data construction procedures refer to: Correlates of War Project Trade Data Set Codebook, Version 2.0. Online: http://correlatesofwar.org. Country-Specific Economic Data: Source: Penn World Table (PWT) and World Bank's World Development Indicators (WDI) Based on: World Bank's WDI Description: PWT's real per capita GDP along with the shares of consumption, investment and government expenditures in it were used to construct time series for each aggregate along with the totals using population data from the same source. Units and other observations: Base years are different for different nations. 36 Financial Openness: Source: "The External Wealth of Nations (EWN) Mark II: Revised and extended estimates of foreign assets and liabilities, 1970–2004" by P.R. Lane & G.M. Miles-Ferretti Based on: IMF's International Financial Statistics dataset Description: As defined by Prasad et al (2007), Financial openness is the gross stock of capital flows to GDP. Gross stock being equal to the accumulated sum of capital inflows and outflows, each including foreign direct investment (FDI), portfolio investment and bank lending –as constructed by Lane & G.M. Miles-Ferretti. Geographical Variables: Source: Compiled by Andrew Rose, available at: http://faculty.haas.berkeley.edu/arose/recres.htm Description: The data contains many geographical and cultural variables and is frequently used in the literature for instrumenting international trade variables, including the language dummies, distance from open seas or rivers, distance from central world cities, landlocked binary variable among others. Polity: Source: Polity IV Project: Political Regime Characteristics and Transitions Description: The Polity variable captures the authority characteristics of a state and was initially collected under direction of T.R. Gurr. This is a composite index constructed upon various political characteristics like competitiveness of executive recruitment, openness of executive recruitment, constraints on and competitiveness of the executive power among others. Notes: This variable can take on values between -10 (full autocracy) and 10 (full democracy). 37 Appendix Figure A1: Herfindahl Index of Exports for Geographical Diversification in International Trade. Select Countries: Unites States, Great Britain, Italy, Mexico, Chile, Brazil, Japan, Korea, China, Kuwait, Sudan and South Africa. USA .07 GBR .1 .08 x hix .09 ITA .1 .06 .09 .05 .08 .07 .04 .06 .03 1960 1970 1980 1990 2000 .07 .06 1960 1970 1980 1990 2000 MEX 1960 1970 1980 1990 2000 CHL BRZ .25 .6 .2 .2 .15 .15 .4 .1 .1 .3 .05 .05 .5 x .25 hix .7 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 JPN 1960 1970 1980 1990 2000 KOR CHN .18 .5 .4 .16 .4 .3 .3 .2 .2 .1 x hix .14 .12 .1 .08 .1 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 KWT .2 x hix .15 .1 .05 1960 1970 1980 1990 2000 year 0 1960 1970 1980 1990 2000 SDN ZAR .4 .5 .3 .4 .2 .3 .1 .2 0 .1 1960 1970 1980 1990 2000 year 1960 1970 1980 1990 2000 year 38 Table A1: Correlations of Various Measures of BCV Correlations among Volatility Measures for Output GDPPC.cyc GDP.cyc GDP.cyc 1 GDP.gr GDPPC.cyc 0.963 1 GDP.gr 0.873 0.900 1 GDPPC.gr 0.871 0.906 0.993 GDPPC.gr 1 Correlations among Volatility Measures for Consumption Cons.cyc ConsPC.cyc Cons.gr ConsPC.gr Cons.cyc 1 0.946 0.884 0.841 ConsPC.cyc Cons.gr ConsPC.gr 1 0.877 0.905 1 0.953 1 Notes: In terms of notations in this table, the ".cyc" suffix corresponds to volatility measures based on the standard deviation of the HP-filtered time series and ".gr" corresponds to one based on the standard deviation of the growth rates. 39 Table A2: Trade Shares with the U.S. and G-7 Countries, five-year period averages Full Sample of 133 Countries - Trade Shares 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average US share / EX 0.16 0.16 0.17 0.17 0.17 0.17 US share / IM 0.20 0.16 0.14 0.15 0.15 0.16 G7 share / EX 0.54 0.51 0.49 0.47 0.41 0.48 G7 share / IM 0.57 0.51 0.46 0.45 0.40 0.47 Sample of 33 Latin America Countries - Trade Shares 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average US share / EX 0.39 0.35 0.36 0.36 0.36 0.36 US share / IM 0.45 0.35 0.31 0.35 0.40 0.37 G7 share / EX 0.65 0.62 0.58 0.58 0.53 0.59 G7 share / IM 0.66 0.61 0.51 0.54 0.54 0.57 Sample of 28 OECD Countries - Trade Shares 1962-1970 1971-1979 1980-1988 1989-1997 1998-2006 Average US share / EX 0.14 0.16 0.13 0.14 0.14 0.14 US share / IM 0.19 0.15 0.13 0.13 0.14 0.15 G7 share / EX 0.48 0.46 0.43 0.46 0.40 0.45 G7 share / IM 0.47 0.45 0.40 0.42 0.41 0.43 Notes: Each sub-period stretches over 9 years following the starting year. IM and EX refer to shares in total imports and exports, respectively. 40 Table A3: List of Countries in the Full Sample country code country code country code 1 Albania ALB 28 China CHN 55 Haiti HTI 2 Algeria DZA 29 Colombia COL 56 Honduras HND 3 Angola AGO 30 Congo, Dem. Rep. ZAR 57 Hungary HUN 4 Argentina ARG 31 Costa Rica CRI 58 Iceland ISL 5 Australia AUS 32 Cote d`Ivoire CIV 59 India IND 6 Austria AUT 33 Cuba CUB 60 Indonesia IDN 7 Bahamas BHS 34 Denmark DNK 61 Iran IRN 8 Bahrain BHR 35 Djibouti DJI 62 Iraq IRQ 9 Bangladesh BGD 36 Dominican Rep. DOM 63 Ireland IRL 10 Barbados BRB 37 Ecuador ECU 64 Israel ISR 11 Belgium BEL 38 Egypt EGY 65 Italy ITA 12 Belize BLZ 39 El Salvador SLV 66 Jamaica JAM 13 Benin BEN 40 Equatorial Guinea GNQ 67 Japan JPN 14 Bhutan BTN 41 Estonia EST 68 Jordan JOR 15 Bolivia BOL 42 Ethiopia ETH 69 Kenya KEN 16 Botswana BWA 43 Fiji FJI 70 Korea KOR 17 Brazil BRZ 44 Finland FIN 71 Kuwait KWT 18 Brunei BRN 45 France FRA 72 Lebanon LBN 19 Bulgaria BGR 46 Gabon GAB 73 Liberia LBR 20 Burkina Faso BFA 47 Gambia, The GMB 74 Libya LBY 21 Burundi BDI 48 Germany GER 75 Luxembourg LUX 22 Cambodia KHM 49 Ghana GHA 76 Macedonia MKD 23 Cameroon CMR 50 Greece GRC 77 Madagascar MDG 24 Canada CAN 51 Grenada GRD 78 Malawi MWI 25 Centr. African Rep. CAF 52 Guatemala GTM 79 Malaysia MYS 26 Chad TCD 53 Guinea-Bissau GNB 80 Maldives MDV 27 Chile CHL 54 Guyana GUY 81 Mali MLI 41 country code country code 82 Mexico MEX 109 Singapore SGP 83 Moldova MDA 110 South Africa ZAF 84 Mongolia MNG 111 Spain ESP 85 Morocco MAR 112 Sri Lanka LKA 86 Mozambique MOZ 113 Sudan SDN 87 Namibia NAM 114 Suriname SUR 88 Nepal NPL 115 Sweden SWE 89 Netherlands NLD 116 Switzerland CHE 90 New Zealand NZL 117 Syria SYR 91 Nicaragua NIC 118 Tanzania TZA 92 Niger NER 119 Thailand THA 93 Nigeria NGA 120 Togo TGO 94 Norway NOR 121 Trinidad &Tobago TTO 95 Oman OMN 122 Tunisia TUN 96 Pakistan PAK 123 Turkey TUR 97 Panama PAN 124 Uganda UGA 98 Paraguay PRY 125 Untd. Arab Emir. ARE 99 Peru PER 126 United Kingdom GBR 100 Philippines PHL 127 United States USA 101 Poland POL 128 Uruguay URY 102 Portugal PRT 129 Venezuela VEN 103 Qatar QAT 130 Vietnam VNM 104 Romania ROM 131 Yemen YEM 105 Rwanda RWA 132 Zambia ZMB 106 Saudi Arabia SAU 133 Zimbabwe ZWE 107 Senegal SEN 108 Sierra Leone SLE Groups of countries as used in various part of this paper: OECD: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States. Latin America: Argentina, Bahamas, Barbados, Belize, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Dominican Republic, Ecuador, El Salvador, Grenada, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Suriname, Trinidad & Tobago, Uruguay, Venezuela. G-7: Canada, France, Germany, Italy, Japan, United Kingdom, and United States. 42 A Quick Reference for Variables and Notations used in this Paper DA NX BCV TO FO OECD G-7 LA WTO/GATT σ (G) σ (M) σ (TOT) σ (RER) σ (r) .m .x HIM, HIX TP TPGDP TPGDPPC TPBCV IBCSync Air Distance English, Spanish, French Landlocked Elevation SEM Domestic Absorption ≡ C + I + G, which is the sum of consumption, investment and government expenditures. Net Exports = Exports - Imports, (in goods and services) Business cycle volatility, of either components of output. Throughout this paper this variable is constructed as the standard deviation of the growth rate of real aggregate (such as GDP or consumption) per capita over the length of each sub-period. Trade openness. Ratio of the sum of exports and imports (of goods and services) to GDP. Financial openness, which equals the ratio of gross stocks of foreign capital assets and liabilities to GDP. Foreign capital includes FDI, portfolio investment and bank lending. Organization for Economic Cooperation and Development. Canada, France, Germany, Italy, Japan, United Kingdom, and United States. Latin America countries. See “List of Countries” for the full list. World Trade Organization/ General Agreement on Tariffs and Trade. Fiscal policy volatility, calculated as the standard deviation of the growth rate of government expenditures in each period. Monetary policy volatility, calculated as the standard deviation of the growth rate of M2 measure of money in each period. Volatility in terms of trade, calculated as the standard deviation of the terms of trade over the length of each sub-period Volatility in real effective exchange rate. Real effective exchange rate (RER) is a “nominal effective exchange rate (a measure of the value of a currency against a weighted average of several foreign currencies) divided by a price deflator or index of costs.” (Source: WB) Volatility is then calculated as the standard deviation of this variable in each sub-period. Standard deviation of the real interest rate in each sub-period as a proxy for the volatility in financial markets and or monetary policy. This suffix indicates that the variable has been constructed using trading partners' import shares in total. This suffix indicates that the variable has been constructed using trading partners' import shares in total. Herfindahl index for geographical diversification. HIM and HIX are respectively calculated based on shares of imports or exports in totals. Variables that capture economic characteristics of trade partners. Including TPGDP, etc. Weighted average GDP of international trade partners. GDP's are from the year 2000. Weighted average GDPPC of international trade partners. GDPPC's are from the year 2000. Weighted average output volatilities of international trading partners. Multilateral correlation of home country's business cycles with respect to those of its international trading partners, weighted by trade shares. Average air distance of the capital of a country to NYC, Tokyo & Rotterdam. Language dummies which indicate the official language of a country. A binary variable that equal 1 if the country does not have access to open waters. Average height of the country's surfaces above the Earth's sea level. Simultaneous equations models. 43