International Fragmentation of Production and Trade Volatility: An Analysis for the European Countries

In the last years, world exports have displayed very strong fluctuations, much higher than GDP fluctuations. Some recent works tried to understand the causes of these fluctuations, and in particular if these strong swings in world trade flows are related to the growing weight of intermediates in trade and to the phenomenon of international fragmentation of production (IFP) or international outsourcing. This is the issue addressed in this paper. We analyze the influence of international fragmentation of production and trade in intermediate goods on the fluctuations of trade flows for the EU countries. After assessing the relevance of international trade in intermediate goods at the sector level using indices computed from the input-output tables of the EU countries, we test whether a significant relation exists between the extent of international fragmentation of production measured by this proxy and the volatility of trade flows, and the possible implications of these results.


Introduction
A distinctive feature of the recent international economic crisis has been the dramatic effect it displayed on international trade. The trade collapse of 2009, amounting to a fall of 12% in volume according to the WTO data, was subject of a number of analysis and concerns, as it appeared by far as the largest drop since World War II. Quite remarkable and to some extent unexpected was also the fast rebound registered in 2010, when trade increased by 14.5% (WTO, 2011), suggesting a very strong pro-cyclicality of international trade. Even before the crisis, in the past decade world exports typically displayed fluctuations much stronger than GDP fluctuations, and even if the crisis has amplified this behavior, this might not be a short-term cyclical effect.
Some recent works tried to understand the causes of trade fluctuations (Eaton et al., 2011), and whether the amplified swings in world trade flows are related to structural changes in the composition of trade, in particular to the growing weight of intermediate inputs in trade and to the phenomenon of international fragmentation of production, IFP (Eichengreen, 2009;Yi, 2009, Bridgman, 2011. The relevant role of intermediate inputs for macroeconomic changes in industrialized and industrializing economies has already been highlighted in the literature (Ciccone, 2002;Jones 2011). Intermediate inputs play an increasingly relevant role also in international trade thanks to the diffusion of international fragmentation of production (Feenstra, 1998;Hummels et al., 2001), spreading the different phases of the production chain of a good across many countries, and giving rise to international exchanges of the good at various stages of production across the countries involved. This organization of production can affect world trade volatility through a number of channels.
First of all, it is possible to show that intermediate goods can generate a sort of multiplier in a simple macroeconomic framework, amplifying the variations of many variables, including trade flows. Furthermore, from the accounting point of view, there can be a statistical effect due to the "double counting" of trade values when a good crosses borders many times during different production phases. This can increase the growth of trade over time, but it can also amplify the fall in the low phases of the business cycle (Yi, 2003;Escaith, 2009). Another possible channel of additional volatility, suggested by some open macroeconomic models, arises because international fragmentation of production can increase the synchronicity of the business cycles of the countries involved in the different phases of the production process by linking demand and production cycles between these countries (see for example Kose and Yi, 2001;Burstein et al. 2008;Arkolakis and Ramanarayanan, 2009;di Giovanni and Levchenko, 2009). In presence of IFP, a demand shock for final goods in a country is a demand shock for intermediate goods or components in another, and the resulting effect on inventories of internationally traded components can also affect the supply chain in third countries (Alessandria et al, 2010). Such links can therefore amplify the volatility of the resulting trade flows.
But other models also suggest that the presence of international production chains can have a stabilizing role on international trade. If the costs of setting up a trade relationship to organize an international production network are relevant and sunk for most firms (Jones et al., 2005;Antràs and Helpman, 2005), the presence of such costs can make the trade relationships related to IFP quite stable over time, and less volatile than the traditional trade flows of final goods (Escaith et al., 2011).
The relationship between IFP and trade fluctuations suggested by the theory is open to empirical testing. In this paper we want to analyze the influence of international fragmentation of production on the fluctuations of trade flows for the EU countries. As shown in Figure 1, also for the EU27 export flows are strongly pro-cyclical, and the recent international crisis sharply increased trade volatility. EU27 exports slowed down remarkably in 2008, and eventually dropped by more than 16% in value in 2009, bouncing back almost entirely in 2010. Furthermore, the EU member countries are on average quite open to international trade and are all deeply connected -even if to different extentsto international networks of production and involved in IFP among them and with third countries. Therefore, this group of countries can represent a good sample to look for the existence of a relationship between IFP and trade volatility.
In the following sections, we first present a simplified theoretical framework linking intermediate inputs and macroeconomic fluctuations. We then display some evidence on the extent of volatility of trade flows at the sector level for 20 EU countries, 2 and we assess the relevance of international fragmentation of production in the manufacturing sectors of those countries, using as a proxy some indices computed from the input-output tables. We then test whether a significant relation exists between international trade in intermediate inputs and the volatility of trade flows in the same sectors, and the possible implications of these results.

Macroeconomic fluctuations and intermediate goods
The idea that linkages between sectors created by intermediate goods are central to economic performance has been around for nearly a century, but it was kept in the background for a long time. As the need of physical proximity to perform different tasks and different stages of production in many sectors diminished, the practice to outsource segments of production to other industrial sectors has increased and the role of intermediate inputs gained new relevance. With the reduction of barriers to international trade, outsourcing became increasingly international, generating flows of trade in intermediate goods.
It is fairly straight forward to show that linkages between sectors can generate a multiplier similar to other multipliers associated with macroeconomic variables. A recent paper by Jones (2011) presents a simple example showing how such a multiplier can work. Assume gross output Q t is produced using primary factors of production (e.g. K t and L t ) and intermediate goods X t according to the expression: where 0 < α,σ < 1. Gross output can be used for consumption or investment, but also as an intermediate good. In this very simplified framework, it is assumed that a constant share x̄ of gross output is used as an intermediate good, so that gross domestic product in this economy is defined as Y t = (1-x̄ )Q t and X t+1 = x̄ Q t Assuming that also investments are a constant share of GDP, and that they depreciate at a constant rate, in steady state GDP is given by In this framework, changes in the allocation of resources defined by x̄ are amplified by the multiplier given by 1/(1-σ). It is also worth noting that the effect of changes in x̄ on Y depends crucially on the level of σ, which represents the importance of intermediate goods in the aggregate production function.
If the economy opens up and international fragmentation of production becomes possible, and intermediate inputs can be imported, expression (1) can be changed so that intermediate goods used in production become (X t + O t ), where O t represents imported intermediate inputs. Import of intermediate goods can change x, and it might also increase the total amount of intermediate inputs used, also changing the GDP level. Importing intermediate goods makes possible to reach a higher GDP with the same factors' endowment in the short run, as gross output is positively related to the total amount of inputs used in production. 3 The impact of imported inputs on gross output needs not to be the same as the domestic inputs (if for example imported inputs have a different technological content). But the relationship between x̄ and Y in (3) is non-monotonic and reductions in x̄ will increase Y in steady state only up to a point. 4 Changes in the amount of imported inputs can therefore per se give rise to variations in GDP. If we further assume that also imported intermediate inputs are proportional to gross output, we have in this framework a co-cyclical behavior of trade and output driven by the supply-side of the economy, additional to the aggregate demand pro-cyclicity. 5 Therefore, in presence of trade in intermediate goods, changes in the economy can give rise variations in trade flows much stronger than the ones expected considering only demand for final goods.

Trade volatility in the EU countries
The expressions in the previous section representing the aggregate economy can be seen as the (geometric) average of sectoral production functions, using intermediate goods to different extents. Previous results in the literature stress the different cyclical properties of trade flows of different types of goods, and especially of durable and non-durables (Engel and Wang, 2011;Levchenko et al., 2010). For these reasons, it is important to examine the behavior of trade flows at the sector level. Here we consider 22 manufacturing sectors, 6 and indeed the extent of fluctuations of EU trade flows appears quite differentiated across sectors.
In what follows, the volatility of trade flows is measured first of all in terms of standard deviations of trade values at current prices over the observation period, 1999-2010 and in the years of the crisis, 2008-2010. These standard deviations of trade flows are normalized by taking their ratio on GDP standard deviations over the same period. 7 Table 1 reports the average variation of imports and exports of the 20 EU countries considered for all manufacturing sectors.
As expected, volatility in most traditional sectors such as apparel (NACE code 18), footwear (NACE code 19), wood products (NACE code 20) is generally lower than volatility in the more advanced or durable goods sectors, like motor vehicles (NACE code 34) or communication equipment (NACE code 32). Volatility is also high in sectors producing investments or intermediate goods, such as machinery (NACE code 29) and basic metals (NACE code 27). We can also observe that variations in export and imports flows appear correlated, as shown in other works.
Also at the sector level, the volatility of trade flows increased sharply during the international economic crisis. Not only on average, as shown in Table 1, but also for each sector and country, volatility in the 2008-2010 period is higher than the average volatility in the 1999-2010 period, even if also in this respect we can observe differences across sectors. In some sectors we have a collapse of trade at the beginning of 2009, with a drop of trade values on a year-to-year basis over 30%, with a strong rebound in the following quarters, while in other sectors trade remains much more stable, and a few cases even display a modest growth. In order to verify the above-mentioned differences in volatility across sectors and over time, we compared trade flows for countries and sectors computing also a rolling standard deviation measured over four quarters for the entire period, and the deviations from a Hodrick-Prescott trend of each trade flow. These measures confirm the different behavior of the flows over the past decade and after the crisis shock. We can therefore infer that there must be some sector characteristics that affect the degree of volatility and the extent of the reaction to shocks.

International fragmentation of production and trade in intermediate goods in the EU countries
Many sectors' characteristics can influence the observed differences in volatility, both because of the organization of production and because of demand patterns. As mentioned, we focus on the role of IFP and on trade in intermediate goods related to IFP, putting aside the possible role of other factors.
There is not a uniquely identified indicator to measure the extent of IFP in any given sector. The very definition of IFP often slightly changes from one work and context to another, let alone any precise measurement of this phenomenon. Here we use the data available in the input-output (I-O) tables provided by Eurostat for the year 2000 and 2005 to build two different indices of intermediate goods trade. The first is the so-called index of narrow offshoring, commonly used in the literature to measure the share of imported intermediate inputs belonging to sector i and employed for production in the same sector. 8 This index should provide an approximate measure of the amount of production of a given good that takes place abroad and is then embodied in the final good through the use of imported inputs, and therefore it is used as a proxy of IFP. The index for country c and manufacturing sector i has the following expression: where a ii is the cell of the I-O matrix for imports measuring the value of imported intermediate inputs i employed in the same sector i, and at the denominator we have the total value of all intermediate inputs used in production in sector i.
The second index that we use measures instead the weight of imports of intermediate goods on the total value of imports of a given sector. This index is obtained as the ratio between the sum by row of the import I-O table and the total value of all imports in the same sector: InterImp ic = ∑ j (import interm. inputs j ) i /tot Imp i This index should instead capture the share of import flows in each sector that is directly affected by the production cycle rather than by final demand.
In Tables 2 and 3 we report the indices for the year 2005, the last year for which import I-O tables are available for most EU countries. The indices aim at measuring some structural characteristics of the economies and manufacturing sectors of our sample, and although they may change over long time periods, they remain relatively stable over a few years. Both indices should capture the relevance of intermediate goods trade at the sector and country level, but the correlation between the two indices across sectors and countries is quite low, equal to 0.15 for the entire sample. The two indices in fact present different features of trade in intermediate goods. The narrow offshoring index can reveal some characteristics of the organization of production, as it measures the weight of imported inputs from the point of view of producing sectors. Instead the intermediate import ratio index should capture the relative weight of two components of the import demand, one coming from firms, demanding intermediate inputs for production, and the rest coming from final consumers demanding final goods. Even if the narrow offshoring index appears more appropriate to capture the extent of IFP, which is determined by producers and firms' choices, in terms of explaining the volatility of trade flows also the other index more focused on demand could provide some indications. Looking at Table 3, reporting the weighted average indices for each sector, we observe that sectoral characteristics are quite different also with respect to the relevance of trade in intermediates. In the adopted classification, some sectors produce a very limited amount of final goods, and therefore present a very high ratio of imported intermediate inputs (e.g. basic metals, NACE code 27). But also the narrow offshoring index displays large differences across sectors (see also Figure 2, where the dispersion of all the narrow offshoring indices is pictured). Generally, the indices for the EU countries are in line with the existing evidence suggesting that some sectors (e.g. office machinery, NACE code 30 or communication equipment, NACE code 32) are highly involved in IFP.

The relationship between trade volatility and trade in intermediate goods
Is the high share of intermediate goods in international trade -confirmed also for the EU countries -associated with high volatility and instability of trade flows? Some recent empirical evidence produced in particular for the US in the context of the recent international economic crisis seems to confirm this hypothesis, even if the results are not always robust (Yi, 2009;Levchenko et al., 2010).
For our sample of EU countries, the overall simple correlation for the period 1999-2010 between the measures of intermediate goods trade and volatility of trade flows results positive, but not very high, with a value slightly higher than 0.30 both for imports and exports. We also observe a positive and significant correlation between the share of sector's imports over GDP and the volatility both of import and exports. It seems that more openness is associated with higher volatility of trade flows, which is plausible, and it is in line with the existing evidence for other countries (Di Giovanni and Levchenko, 2009).
We test the significance of the relationship between trade in intermediate goods and trade volatility, adding also some control variables, to condition the results on some of the characteristics of the industries involved as suggested by the theoretical framework sketched in Section 2. 9 Results are reported in Tables 4 and 5.
Considering our main variables of interest over the period 1999-2010, we observe that the narrow offshoring index is not always significantly correlated with the measures of trade volatility. When controlling for countries and industries characteristics using fixed effects, the narrow offshoring coefficient is positive but not significant. The variable instead displays a positive and significant coefficient in the regressions where we introduce a dummy variable separating old and new EU members. Trade in intermediate goods does not seem to have the same effect on all countries, because without this distinction the narrow offshoring variable is positive but not significant. The narrow offshoring index is correlated to trade volatility in a different way according to the position that a country holds in the production chain, and it seems to affect especially trade fluctuations the new EU member countries where production phases have been delocalized from old EU members. In fact, the dummy variable EU15 (taking value 1 if the country belongs to the EU15 group of old members) displays a negative sign, indicating that the old EU member countries' trade tends to have generally a lower volatility. 10 The negative effect of the EU15 dummy is preserved also when this variable is interacted with the offshoring index.
In these regressions, also the share of intermediate goods in production is generally positive and significant, both for the volatility of imports and of exports. This gives support to the idea of a multiplier effect generated by intermediate goods.
Considering the ratio between imported intermediate inputs and total imports, this is never significantly correlated with the volatility of imports, but it displays a positive and significant coefficient in relation with export volatility when country dummies are present. This positive correlation could depend from the high content of imported inputs that exports require.
The relationship between trade volatility and trade in intermediate goods is more robust if we consider only the past couple of years, covering the period of the economic and financial crisis. In Table 5 we report the estimates of this relation for volatility over the period 2008-2010, and we observe a higher goodness-of-fit of the regression for this time frame, as well as a higher and more significant coefficients for the narrow offshoring indices than for the entire decade. Also the coefficient of the share of intermediate inputs over production is reinforced.
Much less stable is the sign and the significance of the share of inputs on total imports. The effect on import volatility seems negative, but it is positively correlated with export volatility. Instead in this period, the EU15 dummy variable is never significant, even if maintaining the negative sign, hinting to the fact that some convergence in the production structures might have occurred over time between old and new members of the EU, or that the shock due to the crisis might have blurred some of the differences between groups of EU members.

Concluding remarks
In conclusion, there is some evidence of a relationship between volatility of trade flows and the extent of trade in intermediate goods also for the EU countries, but this does not appear very robust. The relationship seems confirmed for some countries, but this result cannot be generalized for the entire Europe in the past decade. The link between trade volatility and trade in intermediates was stronger in past two years, when the international financial crisis hit Europe: during the crisis, trade flows fluctuations have been stronger (both in terms of falls and rises) in those sectors and countries displaying a higher offshoring index. These results are broadly in line with the theory suggesting that this relationship is contingent on a number of factors.
These results are still quite preliminary, as a longer observation period after the crisis is necessary to check whether the observed correlation between trade in intermediates and trade volatility during the crisis is a spurious one. In fact, the very strong shock brought with the crisis can be the common cause of many observed fluctuations of economic variables in EU countries, without a direct relationship between the variables themselves. But if confirmed, these results have interesting positive and normative implications: the increase of international production linkages through trade in intermediate inputs does not necessarily imply a higher volatility of trade flows and a higher exposure to shocks. Therefore, the increased economic integration of the national economies through production linkages might speed up the transmission of shocks, but it needs not to increase the overall instability of the economic systems.    Note: Standard errors in parenthesis. Coefficients with ***, **, * are significant at 99%, 95% e 90% respectively.