Crisis-Proof Services: Why Trade in Services Did Not Suffer ∗
Transcription
Crisis-Proof Services: Why Trade in Services Did Not Suffer ∗
Crisis-Proof Services: Why Trade in Services Did Not Suffer During the 2008-2009 Crisis∗ Andrea Ariu† October 17, 2013 Abstract During the 2008-2009 crisis trade in goods experienced the deepest decline ever recorded. Surprisingly, trade in services lived through the crisis unharmed and some service categories carelessly stuck to their growth paths. In this paper, we investigate the reasons behind the different reaction of services and goods trade using firm-country-product exports for Belgium. First, we decompose the change in exports during the crisis into the different margins, then we investigate econometrically the possible reasons behind the different behavior of the two types of trade. Our analysis shows that both goods and services trade reacted qualitatively similar during the crisis: both the 3% drop in exports of services and the 27% fall in exports of goods were mostly driven by changes in the average quantities exported per market and product, while we do not observe within-firm reallocations in terms of number of destination countries and number of products. The main reason for the different quantitative reaction of services exports is that they are, unlike exports of goods, immune to negative income shocks. We observe that to a one percent drop in GDP growth, it is associated an increase in exports of services equal to 0.05% of the decrease of exports of goods. Keywords: Trade Collapse, Belgium, Services and Goods Trade. JEL Classification: F10, F14, L80. ∗ Financial help under the Globalisation Investment and Trade in Services (GIST) project, funded by the EU 7th Framework Programme (ITN-2008-211429), is gratefully acknowledged. This work was carried out while I was an intern at the National Bank of Belgium, I thank for the hospitality and the support provided. All views expressed in this paper, as well as the errors, are my own solely and do not necessarily reflect the views of the National Bank of Belgium. This paper has greatly benefited from the discussions with Anca Cristea, Chiara Farronato, Martina Lawless, Kalina Manova, Giordano Mion, Mathieu Parenti, Alberto Russo and Ilke Van Beveren as well as from the suggestions received from participants at the International Trade Workshop at Stanford University, GIST Conference in Alghero, ETSG at the University of Leuven, EDP Jamboree at the European University Institute, RIEF conference at Universit´e Sorbonne, the Belgian Trade Day at the University of Antwerp, LETC conference at the University of Ljubljana, ITSG conference at the University of Ancona, Doctoral Workshop at University of Namur, CompNet Conference at the European Central Bank. † FNRS and IRES, Universit´e catholique de Louvain, Belgium e-mail: [email protected] 1 1 Introduction Between the third quarter of 2008 and the second quarter of 2009, trade in goods experienced the steepest decline ever recorded, with both exports and imports unexpectedly falling four times more than income (Freund, 2009). The fall was very severe, highly synchronized across countries and mostly concentrated in the category of durable goods (Baldwin, 2009). In this period of economic turmoil, trade in services barely reacted to the crisis. The most important services markedly continued growing without hesitation and only the category of transport services registered negative figures (Borchert and Mattoo, 2009; Francois and Woerz, 2009). This peculiar resilience is also unpredicted, since most of the studies analyzing trade in services at micro level1 suggest that trade in services shares the same characteristics of trade in goods. Despite this intriguing incongruity, while a large amount of research has attempted to understand the causes of the “Great Trade Collapse” (Baldwin, 2009) for trade in goods,2 the distinctive resilience of trade in services did not get the attention of the international trade literature. In this paper, we analyze the peculiar response of trade in services during the trade crisis of 2008-2009 to understand why trade in services reacted differently with respect to trade in goods. We benefit from a unique micro-level dataset with firm-countryproduct3 exports of goods and services for Belgium and we are able to provide evidence on the reaction of firms to the crisis. The analysis is divided into three sections. In the first, we perform a descriptive comparison of goods and services trade by decomposing changes in Belgian exports into changes in the extensive and the intensive margins, where the former refers to changes in the average number of destination countries per firm and the average number of products exported per firm-country and the latter to the average exports per firm, country and product. Keeping a descriptive spirit, in the second section we use a diff-in-diff approach similar to Behrens et al. (2011) in which we use the first semesters of 2007 and 2008 as the pre-treatment period and the first semesters of 2008 and 2009 as post-treatment period and we explore the differential post-treatment effect of firm, country and product covariates separately for goods and services trade.4 In the third part of the analysis, we perform a triple-difference analysis 1 Breinlich and Criscuolo (2011) for the UK, Kelle and Kleinert (2010) for Germany, Walter and Dell’mour (2010) for Austria, Gaulier et al. (2011) for France, Federico and Tosti (2012) for Italy and Ariu (2012) for Belgium. 2 See Baldwin (2009) for a review. 3 For the sake of expositional clarity, we use the expression “product” also when we refer to a service. 4 This analysis is made separately for trade in goods and trade in services. For both types of trade we run a different regression in which the change in exports between the first semesters of 2007 and 2008 represent the “normal” period and first semesters of 2008 and 2009 the “treatment” period. 2 in which we compare changes in exports of services in the pre and post period with changes in exports of goods in the pre and post period by using only firms that export both goods and services. The goal is to understand the magnitude and statistical significance of the different role played by demand and financial factors for services and goods exports during the crisis. Moreover, we check the robustness of the results by applying the triple-difference strategy to matched “mono-exporters” of goods and services. Our results reveal that, besides the difference in the magnitude of the response to the crisis, there are some qualitative similarities. Both the 3% drop in exports of services and the 27% fall in exports of goods were mostly driven by changes in the average quantities exported per market and product, while we do not find evidence of withinfirm reallocations in terms of number of destination countries and number of products. This means that, both for goods and services, firms did not drop destination countries or products and they only adjusted the values shipped per destination and product. We do not observe relevant heterogeneity in destination countries but an important differential response across different products and firms. In particular, the small decline in services was mostly driven by transport services, while business, financial and telecommunication services continued growing along their trends. For trade in goods, all the product categories dropped, but the fall was mostly concentrated on the durable and investment goods. In terms of firm heterogeneity, we observe that for exporters of services the non-multinational, non foreign-owned, smaller and more financially exposed exporters of services suffered relatively more from the crisis, while for exporters of goods we do not find any relevant difference. Focusing on the determinants of the crisis, we show that the different behavior of goods and services trade is mainly due to a different elasticity to income. On the one hand, the evolution over time of trade in goods is heavily related to changes in GDP of partner countries and demand shocks magnify the reaction. On the other hand, changes over time in trade in services are not related to changes in GDP in partner countries, and dramatic negative shocks as the 2008-2009 crisis do not interfere with their growth paths. By applying the triple-difference strategy explained above, we find that the reaction of trade in services during the “Great Trade Collapse” was statistically different from that of trade in goods: a one percent decrease in GDP growth in partner countries is associated to an increase in trade in services which is 0.05% of the magnitude of the decrease of trade in goods. This means that, if we suppose the Belgian drop in exports of goods (26,681 million euros) being entirely driven by demand, 3 we should observe an increase of 133 millions euros in exports of services, which would mean an increase of about 0.6%. The same result is confirmed both qualitatively and quantitatively if we perform propensity score matching and compare exports of firms exporting only services and firms exporting only goods. This feature of services is more important, with respect to the export of durable goods than to consumables and is mostly accounted by the exports of business services. These results can be explained by the fact that, from a demand point of view, services represent essential inputs for the production process, they are not storable and they are not directly related to the production size. Therefore, their flow must be continuous, it cannot be stored and it cannot easily be modified following changes in the size of the production process. These three characteristics make services look like a fixed cost that firms have to pay in order to ensure the continuity of the production process. Therefore, shocks at the demand level have little impact on the international flows of services. To date, Borchert and Mattoo (2009) is the only paper analyzing trade in services during the crisis. Using aggregate US trade data, they are the first to show that services did not collapse during the crisis of 2008-2009. Then, using the descriptive evidence of Indian exporters in the IT sector, they argue that services did not suffer from the 2008-2009 crisis because their demand is less cyclical and they rely less on external capital. The main contribution of this paper is to go beyond the descriptive evidence available and provide a micro-econometric analysis of the determinants of the different response of trade in goods and trade in services trade during the “Great Trade Collapse”. With our unique data, we are able to test their hypotheses based on the descriptive evidence drawn from a particular set of Indian firms. Secondly, this is the first paper to describe the behavior of trade in services during the crisis at micro level by using a unique dataset on firm-country-product exports from Belgium during the 20082009 crisis. As highlighted by Behrens et al. (2011) and Bricongne et al. (2012) in the analysis of Belgian and French exporters of goods during the crisis, the big advantage of using data at the firm level is that we are able to disentangle the effects of the crisis by looking at within-firm reallocations in terms of changes in the number of products, destination countries and average exports per country and product. Thirdly, this paper builds on the large literature analyzing the effect of macroeconomic shocks on trade. Bernard et al. (2009) for the Asian crisis dissect the fall in exports of goods of US firms into the different margins, concluding, as in our case, that most of the action was on the intensive margin. Amiti and Weinstein (2011) and Iacovone and Zavacka (2009) show that exporters of goods relying more on external capital 4 are those that suffer the most in times of financial crisis. With respect to the recent “Great Trade Collapse”, the emerging consensus points to both demand and supply shocks as the main drivers behind the sudden fall. From the demand side, Behrens et al. (2011), Bricongne et al. (2012) and Eaton et al. (2011) provide evidence of a disproportionate fall in demand for “postponable” goods, such as consumer durables and investment goods. As pointed out by Alessandria et al. (2011), this pushed firms to intensively use inventories and to stop the provision of intermediates, thus reinforcing the negative effect on trade. Moreover, since “postponable” goods constitute a small part of countries’ GDP, but a large share of international trade, this demand shock had dramatic consequences for trade in goods, but relatively little impact on GDP (Francois and Woerz, 2009; Levchenko et al., 2010). From the supply side, Chor and Manova (2012) and Auboin (2009) argue that the financial sector difficulties led to a severe credit crunch that prevented firms from getting enough funds to continue operating in the export markets for goods. As highlighted by Bems et al. (2011), Levchenko et al. (2010) and Altomonte et al. (2012), the interruption of a link in an international production chain can cause the destruction of the entire chain, thus having magnified effects on trade flows. Finally, Evenett (2009) and Jacks et al. (2011) argue that protectionism measures played a further negative role in the collapse. Our paper offers to this literature a new perspective, by adding the service dimension to the analysis of the collapse and by showing that services are immune to sudden negative income shocks. Moreover, it shows that that the different reaction was not due to firm-level differences, since the same divergence is also observed for firms that export both services and goods. Finally, this paper has important policy implications. Our analysis demonstrates the relative stability of services exports during major crises, therefore, countries specializing in the international commerce of services can more easily overcome difficult economic periods and enjoy relatively higher stability. Moreover, it provides further motives for service trade liberalization in the current Doha negotiations. The paper is organized as follows: in section 2 we describe the data; in section 3 we show the descriptive statistics of the crisis; in section 4 we present our diff-in-diff analysis; in section 5 we develop the triple-diff approach and section 6 outlines some concluding remarks and future research directions. 5 2 Data Description The bulk of the dataset used in this paper is composed of three different datasets provided by the National Bank of Belgium (NBB henceforth) concerning trade in services, trade in goods and firm-level accounts. Data on trade in services come from the NBB Trade in Services dataset used to compile the balance of payments and covers the period from 2006 to 2010. The dataset is formed using different surveys conducted by the NBB5 and contains information about trade in services at the firm-destination-product level, so for any Belgian firm present in the dataset we have, depending on the survey, monthly or quarterly information on export values per type of product and destination country. Service products are listed in Table 2 and countries are classified using ISO 2-digit codes. We exclude from the analysis “services to affiliates” (code H7000) because this does not contain information on which specific service is traded and “goods included in the construction services” (code E0002) because this category does not strictly represent trade in services. The dataset captures more or less 60% of total exports of services by Belgium and about 40% of Belgian exporters. The survey nature of the dataset rules out any analysis of entry and exit patterns in foreign markets.6 Therefore, the analysis of this paper will be focused only to the firms that we observe continuously during the period of analysis.7 This means that we are not able to make any analysis on across-firms adjustments, but we can still explore the service and product margins, and thus within-firm adjustments during the crisis. This limitation is not too serious, since entry and exit account for less then 2% of total exports (Ariu, 2012) for both goods and services and since entry and exit has been shown to be a marginal channel of adjustment for firms during the crisis (Behrens et al., 2011; Bricongne et al., 2012).8 Moreover, despite this constraint, this is the only dataset available that can allow the analysis of trade in services at the micro level during the 2008-2009 crisis. Information on trade in goods is taken from the NBB Trade in Goods Dataset, which contains exports and imports of goods performed by Belgian firms at the firm5 For more information on the surveys see Table 1. The main problem is the fact that when a firm enters in the dataset, it is kept for some years even if after a few years it does not meet the thresholds to get into it. Moreover, even by excluding those firms by checking the fulfillment of the criteria, it would provide an idea of the entry and exit into/from the survey but it is questionable whether this would also be representative of entry and exit into/from export markets. 7 These continuing firms account for about 96% of exports and imports present in the surveys, so we can be confident that the data covers the bulk of Belgian trade 8 Bernard et al. (2009) show that during the Asian crisis too the extensive margin was a minor adjustment channel and all the action was concentrated on the intensive margin. 6 6 destination-product level. The data is collected monthly and comes from the Intrastat (Intra-European) and the Extrastat (Extra-European) declarations. Firms are identified thanks to the VAT number, countries are classified using the ISO 2-digit codes and products are classified using the CN nomenclature at 8-digit level. Data on firm-level accounts come from the Business registry covering the population of firms required to file their (unconsolidated) accounts with the NBB. From this dataset we take information on full-time equivalent employment, turnover, operating profits, equities, liabilities, stocks and purchases of intermediates for the year 2007. Unfortunately, turnover figures comprise both goods and services together and there is no information available to distinguish between the two. This prevents us from analyzing the dynamics of goods and services in the domestic market. Multinational and foreign ownership status of firms are taken from the NBB Survey of Foreign Direct Investments. Finally, we take information on GDP growth in destination countries from the IMF World Economic Outlook database (2012 version)9 and information on daily exchange rates on the 1st of April of each year considered from the Statistical Data Warehouse of the European Central Bank.10 3 The Crisis in Figures As previously mentioned, the crisis hit goods more severely than services. Looking at monthly exports of goods and services in Figure 1, we can see that after September 2008 there is a clear rupture and exports of goods fell by about 30%. For services instead, there is no definite sign of discontinuity and they kept about the same pace. This phenomenon was not only confined to Belgium, but worldwide and significant in terms of magnitude. In Figure 2, we plot for OECD countries the ratio of quarterly exports of services over exports of goods, with the first quarter of 2006 normalized to one.11 By looking at the average for the whole OECD, represented by the thick red line, it is evident how as from the third quarter of 2008 the ratio of services over goods increased rapidly and significantly, meaning that, while trade in goods collapsed, trade in services held up relatively well on average in all OECD countries. In the rest of the paper the analysis will be focused only on the first semester of each year (as also in Behrens et al. (2011)). This choice is made in order to i) reduce seasonality issues evident from the monthly data; ii) include the maximum amount 9 Available at http://www.imf.org/external/pubs/ft/weo/2012/01/weodata/index.aspx Available at: http://sdw.ecb.europa.eu/ 11 Data come from the OECD database available at http://stats.oecd.org 10 7 of firms (it should be noted that some firms declare exports only quarterly and most of the firms do not export every month so making an analysis on continuing firms at monthly level would reduce the number of firms dramatically); iii) do a clear prepost comparison avoiding the shock present in the statistics of the third and fourth quarters of 2008. The first step to understanding the composition of the changes in Belgium’s exports is to decompose total Belgian exports at time t (where, in this case t = {S12008, S12009}), of trade type y (where y = {Services, Goods}), Xty , into the number of firms ft , the average number of served markets per firm c¯t , the average number of exported products per market-firm p¯t and the average exports per firmmarket-product (service) x¯t : Xty = f y ∗ c¯yt ∗ p¯yt ∗ x¯yt . By taking the ratio between the first semester of 2008 and the first semester of 2009, we can thus break down the change in total exports, ∆X y = y X2009 , y X2008 into the change in the extensive margins (firms-services-markets) and the change in the intensive margin (the average exports per firm-market-service): ∆X y = ∆f y ∗ ∆c¯y ∗ ∆p¯y ∗ ∆x¯y (1) Since we focus only on continuing firms, the change in the number of firms, ∆f , is equal to one.12 Looking at Table 3, we can appreciate that the change in Belgian exports between the first two semesters of 2008 and 2009 is -26.81% for goods and only -3.13% for services. Even if these falls differ dramatically in quantitative terms, qualitatively they are both generated almost entirely by a reduction in the quantities exported per market and product. This means that Belgian firms, both for goods and services trade, did not significantly leave destination markets or reduce the number of products provided for each market, but they only adjusted the values exported per market and product. Therefore, this huge difference in the reaction of the average quantities exported per market and product suggests that the intensive margin is the key to understanding the different reaction of services and goods trade. By dividing Belgian exports into the different product categories, we can appreciate in Table 4 a wide heterogeneity across products, both for services and goods. Services related to transport experienced a drop commensurate to that of goods. On the other hand, business and telecommunication services continued their sustained growth. Therefore, besides transport services, the other services did not suffer from the crisis and they continued their normal growth paths. If we consider also that financial, insurance and business services represent more than 50% of Belgian exports, this is a quite 12 Which means that it does not contribute to explaining the growth of exports. 8 important result that can have relevant policy implications. For exports of goods, all product categories experienced a decline, yet the bulk of the collapse is accounted for the intermediates and durable goods. By decomposing Belgian exports into EU and non-EU and to OECD and non-OECD in Table 5, we see a mixed country pattern. Intra-EU and extra-OECD exports of services experienced a more important drop than non-EU and OECD. This is because most of the extra-OECD trade is represented by transport of services, which we have seen is the only service category that collapsed. For exports of goods instead, the fall is similar in all the country categories. To discern differences across firms, in Table 6 we divide exports following the multinational and foreign ownership status, size and financial situation of the exporter. The numbers say that, for services, the non-multinational and non foreign-owned firms were hit by the crisis, while multinational and foreign owned registered positive figures. However, these declines are much smaller than those for goods, for which we do not observe any heterogeneity following the multinational and foreign ownership status. Finally, by defining a firm as big if it has the full-time equivalent employment higher than the median exporter in the same industry and as financially exposed if it has higher external financial dependency than the median exporter in the same industry, we can see from Table 7 that there is no heterogeneity for firms exporting goods, while small firms exporting services suffered more from the crisis than big ones. Summing up the descriptive evidence on the crisis in Belgium, it looks like exports of services did not suffer as much as goods exports. Both service and goods exporters kept the same number of destinations and products per destination, adjusting only at the intensive margin, although they did so with very different magnitudes. We find that there is no particular pattern looking at the partner countries for both goods and services, but there is great heterogeneity looking at the different product types. In particular, transport services dropped similarly to trade in goods, while professional, financial and telecommunication services continued growing at a very high pace. For trade in goods, we find that the decrease is mostly due to a reduction in the intermediates and durable goods, while other types of goods declined more smoothly. Finally, we observe an important heterogeneous response of firms, based on ownership and multinational status, size and financial situation for exports of services, but not for exports of goods. Firm, country and product dimensions provide together important information on the nature of the crisis, therefore we are going to take them into consideration in our empirical strategy. 9 4 Diff-in-Diff of the Crisis To understand which factors led to a different response of exports of services with respect to exports of goods, we use a type of diff-in-diff approach similar to Behrens et al. (2011), in which the change in the logged exports to a particular market c, of a particular product p, by a Belgian firm f between S12007 and S12008, and S12008 and S12009, ∆Xfy,tcp = logXfy,t+1 − logXfy,tcp , is regressed, separately for goods and services cp (remember that y = {Services, Goods}), against the treatment dummy T t a vector containing firm, country and product characteristics, Zfy,tcp , and the interaction of this vector with the treatment dummy, Zfy,tcp ∗ T t . ∆Xfy,tcp = α + β00 T t + β10 Zfy,tcp + β20t Zfy,tcp ∗ T t + tf cp (2) In this specification, β00 represents the treatment specific effect, β10 the contribution of the firm, country and service characteristics in normal times and β202009 the contribution of these same variables during the crisis in 2009. Since we do not have services characteristics, Zfy,tcp contains service or product dummies13 to capture heterogeneity across goods and across services. Given that our variables of interest vary along three dimensions (firm, product and country), we use the multi-level clustering procedure developed by Cameron et al. (2011) to correct standard errors. Finally, to alleviate endogeneity issues of our firm-level variables, we use the balance sheet data from 2007 only and trade data from 2006 for computing export and import to turnover ratios. In the first specification, we use dummy variables for all our firm-level variables indicating whether a firm is above or below the median among all exporters in the same industry. In this way, the interaction between the firm-level variable and the crisis dummy would tell what happened in terms of export growth, for example during 2009, to a firm that in 2007 was among the most productive. In a second specification, we also make use of our firm-level covariates in levels. Our independent variables aim to capture the heterogeneity observed in the descriptive statistics, the supply and demand features assumed by Borchert and Mattoo (2009) as the cause of the resilience of trade in services and the different determinants of the fall of trade in goods. We use the size, productivity and multinational and foreign ownership status of the firm to control for heterogeneity across firms, different variables that measure their exposure to external finance14 , variables capturing the involvement in global value chains, the importance of 13 Please note that in order to have the same level of disaggregation between services and goods, we use the CN classification at the 2-digit level. 14 In line with the previous literature measuring credit constraints (see for example Manova and Yu 10 stocks and typical demand determinants as GDP growth, exchange rates and dummies for OECD (but not EU) destinations and non-OECD (and non-EU) destinations. The complete list of variables, their description and their source is presented in Table 8. Focusing attention on the left hand side of Table 9, we replicate in a slightly different setting15 the results of Behrens et al. (2011) for trade in goods. Our estimates point to a weak influence of supply factors in explaining the reaction of goods during the crisis, while demand determinants played a very important role. In particular, GDP growth coefficient is positive and significant, both before and during the crisis, which means that exports of goods follow the economic cycle, and important negative shocks magnify the reaction. As pointed out by Behrens et al. (2011), the drop in demand is the most important factor in explaining the “Great Trade Collapse” and it accounts for about two-thirds of the fall. With respect to the global value chain, we observe that firms having higher exports of goods over intermediates suffered more from the crisis, while those more involved in imports of goods were less affected. Finally, our variables capturing the financial situation of the firm point to a weak incidence of credit constraints in the explanation of the fall. Switching to the right-hand part of Table 9, we clearly see that demand and all the other factors that characterized the huge drop for trade in goods did not play any role both before and during the crisis for trade in services. This means that the growth of trade in services does not depend significantly on demand and supply determinants, and it proves to be immune to shocks at the supply and demand level. By performing the same regressions using continuous firm-level variables (Table 10), we get the same qualitative results both for goods and services exports. What is the reason behind the different reaction of services to the supply and demand determinants? From the demand point of view, services represent essential inputs for the production process: a company without its call center, janitorial, accounting and financial services can hardly continue producing; the work of construction companies is seriously hampered by the lack of architectural and engineering services. At the same time, services are intangible: this means that they cannot be stored and their flow must be continuous, thus forcing firms to buy services to keep the production process going. Finally, services are often not directly related to the size of the firm’s production: the value of audit, advisory and call center services purchased by a firm do not necessarily (2012), Whited (1992), Fazzari and Petersen (1993), Greenaway et al. (2007) and Knight et al. (2010)) we control for the dependence on external finance, the structure of the debts (time horizon and debt types) and the debt burden. 15 The main differences are based on the fact that we use product dummies instead of product characteristics to capture heterogeneity across products. 11 change following a reduction in the sales of a firm. Therefore, services flows cannot be easily modified following changes in the size of the production process. These three characteristics: essentiality, non-storability and size independency of services, make them look like a fixed cost that firms have to pay to ensure the continuity of the production process. Therefore, sudden income shocks do not have direct effects on the international flows of services. From the supply side, our results suggest that services seem to be less affected than goods by financial constraints. Borchert and Mattoo (2009) show that many services are traded over the internet and that they can more easily rely on advance payments. Therefore, they do not need big investments to be produced and traded internationally and in general these characteristics lower their need for external capital. At the same time, it must be noted that services can hardly be used as collateral: first, they are intangible and it is hard to objectively value something that is immaterial; second, they are usually very customized and they have little value outside the specific contract between two parties. Therefore, service exporters might not even be able to ask for external trade capital and thus be less sensitive to credit shocks. 5 Diff-in-Diff-in-Diff of the Crisis The previous section highlighted the different role that supply and demand factors had on services and goods during the crisis. In this section, we delve deeper into the analysis and we test the significance and magnitude of the differences across goods and services by focusing on the income and financial determinants highlighted in the previous section. A possible issue to test the differences in the role of GDP growth and external finance across exports of goods and services can be represented by the fact that firms exporting services might be different from those exporting goods. Our strategy to tackle this problem is to consider only firms that export both services and goods, the “bi-exporters”, and compare changes in exports of services before and during the crisis with changes in exports of goods before and during the crisis for the same firm. In this way, we rule out any difference across goods and services related to both observable and unobservable components of supply. In a certain sense, it is like performing a “perfect matching” since we compare the same firm when exporting services and when exporting goods. Then, to check whether these “bi-exporters” represent a particular set of firms and control the robustness of our results, we focus solely on the “mono-exporters” and for every service exporter we find the closest goods exporter by applying propensity 12 score matching.16 Using these matched “mono-exporters”, we can make the same type of comparison, but we are able to match exporters of services with exporters of goods only on observables. With respect to the previous analysis, we have to drop the product dimension, since we cannot say which good should be matched with a particular service and vice-versa. Therefore, the unit of analysis is represented by the change in exports type y of a firm f in a country c at time t, (∆Xfyct ). The interaction between a dummy indicating the service flow Sf , the treatment dummy for the crisis T t and the GDP growth, GDPct , will provide evidence on the differential impact of GDP on exports of services with respect to exports of goods during the crisis. This triple-difference strategy can be expressed analytically as: ∆Xf ct = α+β00 T t +β10 GDPct +β20 GDPct ∗T t +γ00 Sf +γ10 Sf ∗T t +γ20 Sf ∗GDPct +γ30t Sf ∗GDPct ∗T t +tf c (3) Where β00 , β10 and β20 have the same interpretation as the diff-in-diff strategy we used before. γ00 is the specific treatment effect controlling for differences across goods and services. γ10 captures the different response across goods and services during the crisis. γ20 controls for specific differences in the effect of GDP growth across goods and services. Our variable of interest, γ30t captures the differential effect of GDP growth on export of services (with respect to exports of goods) during (for 2009) and after (for 2010) the crisis. Since the GDP growth is at the country level, we cluster standard errors accordingly. Moreover, to control for firm-level determinants, we use firm-year dummies. Looking at the estimated coefficient of γ302009 in the first column of table 12 (panel a), we can see that the GDP growth had a negative and significant effect on exports of services with respect to exports of goods during the crisis of 2009. This means that the reaction of services’ exports to the income shock was significantly different from that of goods’ exports. Our estimates say that a one percent decrease in income growth is associated with an increase of exports of services equal to 0.05% of the reduction of trade in goods. By using the “matched” mono-exporters, and performing the same analysis, we can see from the first column of Table 12 (panel b) that the results are the same both qualitatively and very similar also in quantitative terms. Therefore, it looks like the different reaction of services with respect to goods is not influenced by the fact that bi-exporters represent a particular category of exporters. As pointed out in the descriptive part of the paper, most of the decline in the exports 16 We present results using Mahalanobis Matching (with replacement) where firms are matched in terms of: size, productivity, capital intensity, average wage, multinational and foreign ownership status. Using other matching functions like Nearest Neighbor or Kernel Matching does not change the results. Table 11 reports the statistics and differences for the control and the treatment group. 13 of goods was accounted by durable goods. At the same time, we outlined the fact that different services reacted differently to the crisis. In order to refine our results, we divide exports of goods and services into the different product categories and we apply the same type of analysis for each of them. We divide exports of goods into the exports of durable (or postponable) goods and exports of non-durable (or consumables). Results in columns 2 and 3 of Table 12 indicate that the resilience of services is more important with respect to durable goods than to consumables. This is not an unexpected result, since most of the decline in goods exports was accounted by durable goods. In columns 4-10 of Table 12, we present the results for exports of the different service categories. As can be seen from the magnitude and significance of the coefficients, most of the findings are accounted by the Business Services. This can be explained by the fact that during the crisis firms might have needed more external support than in usual times. Therefore the demand for services like consultancies, legal, audit, remained high besides the economic downturn. Having found evidence of a significant differential effect of the GDP growth on export of services with respect to exports of goods, we can apply the same type of analysis to check if there is a differential effect of financial constraints on the export of services with respect to goods. In this case, the interaction between a dummy indicating the service flow Sf , the treatment dummy for the crisis T t and the external financial exposure variable17 F INft will provide evidence on the role of credit constraints. Analytically the equation to be estimated is very similar to (3): ∆Xf ct = α+β00 T t +β10 F INct +β20 F INct ∗T t +γ00 Sf +γ10 Sf ∗T t +γ20 Sf ∗F INct +γ30t Sf ∗F INct ∗T t +tf c (4) The only differences are that we cluster standard errors at the firm level and we use country-year dummies to control for demand determinants. The results in Table 13 do not provide as solid results as for the GDP growth. In most of the specifications the coefficient is not significant, so we do not find strong evidence supporting a different role of credit constraints for exports of services relative to exports of goods during the crisis. This result holds both for “bi-exporters” and “mono-exporters”, and differentiating goods and services into their different categories. Moreover, it holds across different variables controlling for the financial situation of the firm.18 17 Measured as investments minus operating profits over investment. As previously mentioned, the results in the tables refer to investments minus operating profits over investment, but they are also available upon request when using the share of financial debts and the share of short-term debts. 18 14 6 Conclusions This paper shows that exports of services did not suffer from the 2008-2009 crisis because they are more immune to income shocks than trade in goods. This difference proves to be statistically significant: a one percent decrease in GDP growth is associated with an increase of trade in services equal to 0.05% of the reduction in exports of goods. This means that had the crisis been totally driven by an income shock, we should have observed an increase of trade in services of 133 million euros, which means a rise of 0.6%. This peculiar resilience of trade in services is more pronounced with respect to durable goods than to consumables and is mostly accounted by the business services category. This means that despite the crisis, demand for services like consultancies, audit and legal remained high. In terms of theoretical models, the resilience of trade in services with respect to goods can be coherent with differences in demand and supply. On the one hand, trade elasticities with respect to income might be higher for goods than for services. On the other hand, trade costs may differ as well. In particular, higher fixed costs for services would result in more productive and resilient services exporters.19 Finally, these results provide strong policy implications: countries specializing their exports towards services can benefit from a lower sensitivity to demand shocks, thus alleviating the consequences of an economic turmoil thanks to the special nature of services. Moreover, they provide further motives for service trade liberalization to Doha trade negotiations round. References Alessandria, G., Kaboski, J. P., and Midrigan, V. (2011). Us trade and inventory dynamics. American Economic Review, 101(3):303–07. Altomonte, C., Mauro, F. D., Ottaviano, G. I. P., Rungi, A., and Vicard, V. (2012). Global value chains during the great trade collapse: A bullwhip effect? CEP Discussion Papers dp1131, Centre for Economic Performance, LSE. Amiti, M. and Weinstein, D. E. (2011). Exports and financial shocks. The Quarterly Journal of Economics, 126(4):1841–1877. 19 In the appendix we show how these differences can be interpreted in terms of existing theoretical models, leaving to future research the challenge of building a new framework to analyze services and goods trade together. 15 Ariu, A. (2012). Services versus goods trade: Are they the same? Working Paper Research 237, National Bank of Belgium. Auboin, M. (2009). Restoring trade finance: What the g20 can do. In The Collapse of Global Trade, Murky Protectionism, and the Crisis. Richard Edward Baldwin Simon Evenett, London, UK. Baldwin, R. E. (2009). The great trade collapse: Causes, consequences and prospects. Technical report, CEPR: London, UK. Bas, M. (2012). Technology adoption, export status, and skill upgrading: Theory and evidence. Review of International Economics, 20(2):315–331. Behrens, K., Corcos, G., and Mion, G. (2011). Trade crisis? what trade crisis? The Review of Economics and Statistics, forthcoming. Bems, R., Johnson, R. C., and Yi, K.-M. (2011). Vertical linkages and the collapse of global trade. American Economic Review, 101(3):308–12. Bernard, A. B., Jensen, B. J., Redding, S. J., and Shott, P. K. (2009). The Margins of US Trade. American Economic Review, 99(2):487–93. Borchert, I. and Mattoo, A. (2009). The crisis-resilience of services trade. The Service Industries Journal, 30(13):2115–2136. Breinlich, H. and Criscuolo, C. (2011). International Trade in Services: a Portrait of Importers and Exporters. Journal of International Economics, Elsevier, 84(2):188– 206. Bricongne, J.-C., Fontagn, L., Gaulier, G., Taglioni, D., and Vicard, V. (2012). Firms and the global crisis: French exports in the turmoil. Journal of International Economics, 87(1):134 – 146. Bustos, P. (2011). Trade liberalization, exports, and technology upgrading: Evidence on the impact of mercosur on argentinian firms. American Economic Review, 101(1):304–40. Cameron, A. C., Gelbach, J. B., and Miller, D. L. (2011). Robust inference with multiway clustering. Journal of Business & Economic Statistics, 29(2):238–249. 16 Caron, J., Fally, T., and Markusen, J. R. (2012). Skill premium and trade puzzles: a solution linking production and preferences. NBER Working Papers 18131, National Bureau of Economic Research, Inc. Chao, H., Kim, S., and Manne, A. S. (1982). Computation of equilibria for nonlinear economies: Two experimental models. Journal of Policy Modeling, 4(1):23 – 43. Chor, D. and Manova, K. (2012). Off the cliff and back? credit conditions and international trade during the global financial crisis. Journal of International Economics, 87(1):117–133. Eaton, J., Kortum, S., Neiman, B., and Romalis, J. (2011). Trade and the global recession. NBER Working Papers 16666, National Bureau of Economic Research, Inc. Evenett, S. J. (2009). What can be learned from crisis-era protectionism? an initial assessment. CEPR Discussion Papers 7494, C.E.P.R. Discussion Papers. Fazzari, S. M. and Petersen, B. C. (1993). Working capital and fixed investment: New evidence on financing constraints. RAND Journal of Economics, 24(3):328–342. Federico, S. and Tosti, E. (2012). Exporters and importers of services: firm-level evidence on italy. Temi di discussione (Economic working papers) 877, Bank of Italy, Economic Research and International Relations Area. Fieler, A. C. (2011). Nonhomotheticity and bilateral trade: Evidence and a quantitative explanation. Econometrica, 79(4):1069–1101. Francois, J. F. and Woerz, J. (2009). Follow the bouncing ball – trade and the great recession redux. In Baldwin, R. E., editor, The Great Trade Collapse: Causes, Consequences and Prospects. CEPR, UK. Freund, C. (2009). The trade response to global downturns : historical evidence. Policy Research Working Paper Series 5015, The World Bank. Gaulier, G., Mirza, D., and Milet, E. (2011). French Firms in International Trade in Services. Economie et Statistique, 435(1):125–147. Greenaway, D., Guariglia, A., and Kneller, R. (2007). Financial factors and exporting decisions. Journal of International Economics, 73(2):377–395. 17 Hanoch, G. (1975). Production and demand models with direct or indirect implicit additivity. Econometrica, 43(3):395–419. Iacovone, L. and Zavacka, V. (2009). Banking crises and exports : lessons from the past. Policy Research Working Paper Series 5016, The World Bank. Jacks, D. S., Meissner, C. M., and Novy, D. (2011). Trade booms, trade busts, and trade costs. Journal of International Economics, 83(2):185–201. Kelle, M. and Kleinert, J. (2010). German firms in service trade. Applied Economics Quarterly (formerly: Konjunkturpolitik), Duncker & Humblot, Berlin, 56(1):51–72. Knight, J., Ding, S., and Guariglia, A. (2010). Investment and financing constraints in china: does working capital management make a difference? Economics Series Working Papers 521, University of Oxford, Department of Economics. Levchenko, A. A., Lewis, L. T., and Tesar, L. L. (2010). The collapse of international trade during the 200809 crisis: In search of the smoking gun. IMF Economic Review, 58(2):214–253. Manova, K. and Yu, Z. (2012). Firms and credit constraints along the value-added chain: Processing trade in china. NBER Working Papers 18561, National Bureau of Economic Research, Inc. Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6):1695–1725. Walter, P. and Dell’mour, R. (2010). Firm-Level Analysis of International Trade in Services. IFC Working Papers No.4. Whited, T. M. (1992). Debt, liquidity constraints, and corporate investment: Evidence from panel data. Journal of Finance, 47(4):1425–60. 18 7 Appendix In this appendix we discuss how the differences between services and goods trade can be thought within existing theoretical frameworks. We do not intend to present a new framework or a complete and fully-fledged model, but rather to depict how the interaction between services and goods trade highlighted in this paper can be rationalized into existing models. Suppose consumers can choose to consume a continuum of two products: goods G, denoted by jG ∈ [0, 1] and services S, denoted by jS ∈ [0, 1]. They can choose the quantities of goods jG , q(jG ) and services jS , q(jG ) to maximize the following constant relative income elasticity (CRIE) utility function: αS σS σS − 1 Z 1 q(jS ) σS −1 σS djS + αG 0 σG σG − 1 Z 1 q(jG ) σG −1 σG djG (5) 0 These non-homothetic preferences were introduced by Hanoch (1975) and Chao et al. (1982) and they have been recently used for trade models by Fieler (2011) and Caron et al. (2012). αS > 0 and αG > 0 represent the weights of, respectively, services and goods (with αS + αG = 1) and σS and σG can be interpreted as the income elasticity of demand of goods and services.20 To understand the reaction of services and goods to income changes, it is useful to take the ratio between the total expenditure of services and goods: xS = λσG −σS xG αS (PS )1−σS αG (PG )1−σG As explained in Fieler (2011), the term in parentheses sets the level of captures how xS xG xS xG xS xG (6) xS , xG while λσG −σS varies with income changes. In particular, if σS > σG , the ratio is decreasing in λ and hence increasing in income. If instead σS < σG the ratio is increasing in λ and hence decreasing in income. This second case is indeed representative of the results we get from the empirical analysis. Changes in income have a strong effect on goods, while they have very little influence on services. Therefore, a negative income shock leads to increase the ratio xS , xG which is exactly in line with our empirical evidence. From the supply side, let us suppose an environment similar to Melitz (2003) in which firms are heterogeneous in their productivity ϕ and pay a fixed costs to produce and export. Assume that the fixed cost of exporting services, fS is higher than the fixed cost of exporting goods fG .21 At the same time, assume that trade is subject to iceberg 20 See Fieler (2011) for a complete discussion of this. Please note that if σ = σS = σG we are back to the usual CES environment. 21 This choice would motivated by the evidence that introducing services in foreign countries is 19 costs that are higher for goods than for services τG > τS .22 Given this setting, it is possible to map the productivity of the firm and the export choice. In Figure 3, we plot the profits, π and the productivity, ϕ, of the firms choosing different export strategies. We can see that the least productive firms (those below ϕD ) will stay in the domestic market and produce services or goods; those in the middle (between ϕD and ϕG ) export goods and the most productive ones (those above ϕS ) export services. Bearing this framework in mind, the consequences of a decrease in income are straightforward: as σS < σG the demand for services decreases less than the demand for goods and firms exporting goods are more seriously hit by the income shock; at the same time, firms exporting services are more productive and thus more resilient to shocks. The same mechanisms can also be extended for firms exporting both services and goods (and thus for the analysis of within-firm adjustments).23 subject to higher costs than for goods and fewer (but more productive) firms are able to enter foreign markets for services (Breinlich and Criscuolo, 2011; Ariu, 2012). 22 Services are immaterial and they do not physically travel around the world (for example, many services are traded over the internet, thus enjoying negligible transportation costs). Moreover, services are not subject to tariffs, so once the fixed cost of exporting is paid, firms can freely trade their desired amounts. Therefore, it is plausible to think variable trade costs being lower than for goods, which instead are subject to physical transportation and the payment of tariffs. 23 Suppose that firms can invest a fixed sum fB to add to its exports also the other type. This investment represents the cost of adapting the production process to the other type of product and it leads to increase the productivity of the firm to γϕ > ϕ. This means that by exporting both goods and services firms can more efficiently use their internal resources and decrease their marginal costs. Using this simple strategy, inspired by the works of Bustos (2011) and Bas (2012), the most productive firms export both goods and services and an income shock hurts more goods than services even within the same firm. 20 Figure 1: Belgian Monthly Exports 2006-2010 Table 1: Overview of the NBB Surveys on Trade in Services Survey Name F01DGS F01CDC F01MER F03MER F03CMS Frequency M M M/Q M/Q M/Q/A Mandatory declaration criteria e10M annual Intrastat or e5M monthly Intrastat or e1M annual Extrastat List of firms in Law n 187 of 30.12.1982 e10M annual Intrastat or e5M monthly Intrastat or e1M annual Extrastat All firms declaring e10M Intrastat e1000 annual Intrastat or e5000 monthly Intrastat or e1000 annual Extrastat F13CON F23CON F03TRP F03AVS F02BRO F02CCI F02TRA M/Q M/Q M/Q/A Q M/Q M M/Q All firms All firms All firms All firms All firms with more than 10 employees All payments with credit cards not included in the other surveys Travel Agencies with more than 10M annual turnover Services Targed All All Merchanting Merchanting All except Transport, Merchanting, Telecommunications, Insurance, Construction, Energy Construction Energy, Recycling, Real Estate Transport Telecommunications Insurance All Travel Note: This table presents the different surveys that constitute the NBB Trade in Services dataset from 2006 onwards, their frequency (M stands for monthly, Q for quarterly and A for annual), the criteria to make the declaration mandatory and the services targeted by the each survey. All values are in Euros. 21 Figure 2: Ratio of Quarterly Exports of Services over Exports of Goods for OECD Countries, Q1 2006=1 Source: OECD.Stat Extracts Figure 3: Profits and Productivity π S πexp G πexp πD ϕD f ϕG ϕS fG fS 22 ϕ Table 2: Service Categories G0001 G0002 G1000 G5000 G6000 G7000 H0000 H1000 H1100 H1500 H2000 H3000 H4000 H5000 H5101 H5102 H6100 H6200 H6300 H6400 H6500 H7000* H8000 H9000 H9100 H9200 BUSINESS SERVICES IT services IT services Repair services for IT Information services Commission Fees Franchising services Patent services Services for intangible investments Other commissions Technical and Business Services Legal services Accounting services Consulting services Marketing and opinion polling services Research and development services Architectural services Waste treatment services Agricultural and mining services Local services Secretarial services Security services Translation services Photographic services Janitorial services Services to affiliates Audio-visual services Educational services Health service Other cultural and personal services B1101 B1102 B1103 B1200 B2500 B3500 B4000 B4500 Passenger transport by different means of transportation Freight transport by different means of transportation Agency services for transport of passenger (different means of transportation) Satellite launch services and freight transportation in the space Passenger Auxiliary services to transport Auxiliary services to different transport modes Any rental with crew C0000 C0301 C0302 C0303 C0304 C0401 C0402 C1301 C2301 C9000 C9100 SERVICES TO NON RESIDENTS Agency Services Conference fees Car rental services Excursion fees Cruises Agency services for concerts Agency services for sport and cultural events Agency services for tourism Car dealers services Repair services Teaching services Healthcare services D0001 D0002 D1000 B0001 B0002 B0003 B2001 B2002 B2003 B3000 B0101 B0102 B2101 B2102 B2103 B3100 B0201 B0202 B2201 B2202 B2203 B3200 B0301 B0302 B2301 B2302 B2303 B3300 B0401 B0402 B2401 B3400 B0500 B1000 TRANSPORT SERVICES Sea Transport Freight Boat rental for freight transportation Transport of floating devices Passenger Boat rental for passenger transportation Agency services for transport of passengers Auxiliary services to sea transport Air Transport Freight Plane rental for freight transportation Passenger Plane rental for passenger transportation Agency services for transport of passengers Auxiliary services to air transport Rail Transport Freight Train rental for freight transport Passenger Train rental for passenger transport Agency services for transport of passengers Auxiliary services to rail transport Road Transport Freight Truck Rentals for freight transport Passenger Bus rentals for passenger transport Agency services for transport of passengers Auxiliary services to road transport Internal Waters Transport Freight Boat rental for freight transport Boat rental for passenger transport Auxiliary services to internal waters transport Other Transport Freight Freight transport by pipes and electricity transport TELECOMMUNICATION SERVICES Postal and Communication services Postal services Communication services Telecommunication services Telecommunication services CONSTRUCTION SERVICES E0001 E0002* E0003 E0100 E0200 E0301 E0302 E5300 E5400 F1000 F1100 F2000 F2100 F3000 F3100 F4000 F5001 F5002 F6001 F6002 F6003 F6004 F6005 F6006 F6301 F6302 F6303 F6304 F6305 F6306 Construction services abroad for less than one year Goods included in the construction services Infrastructure reparation Construction services abroad for less than one year (sub-tendered) Construction services in Belgium for a foreign company (sub-tendered) Construction services in Belgium for a sub-tendered foreign company (sub-tendered) Purchase of construction services (only for imports) Construction and installation services in Belgium (for a foreign company) Construction and installation services abroad FINANCIAL AND INSURANCE SERVICES Insurance Services Insurance premium against theft, damage and loss of freight Freight insurance Other insurances Indemnities related to other insurance Re-insurance premium Indemnities for re-insurance Fees, consultancies and other services related to insurance and re-insurance Financial Services Financial services from postal companies Financial services Financial services related to transport of passenger Financial services related to transport of freight Operational leasing Other operational leasing Real estate services to international organizations Real estate services to non residents Financial leasing for passenger transport Financial leasing for freight transport Financial leasing for IT Other financial leasing Financial leasing for real estate services to international institutions Other financial leasing Note: This table presents the different service categories present in the different surveys of the NBB. Codes marked with an asterisk (*) are excluded from the analysis. 23 Table 3: Change in the Margins of Belgian Exports (2008S1-2009S1) Panel a: Exports Period 2008S1 Total 21,757 Services 2009S1 (∆-1)% 2008S1 Goods 2009S1 (∆-1)% 21,075 -3.13% 99,534 72,853 -26.81% 2,107 11.37 1.55 -0.33% 2.00% 12,964 8.58 3.72 12,964 8.46 3.79 -1.41% 1.83% 0.57 -4.72% 0.24 0.18 -27.09% Extensive Margins: Firms 2,107 Countries 11.41 Products 1.52 Intensive Margin: Average Sales 0.60 Note: This table presents the decomposition of the growth rate of Belgian exports between the first semester of 2008 and the first semester 2009 into the extensive margin (average number of export markets per firm and average number of product per market-firm) and the intensive margin (average exports per firm, market and product). Table 4: Change in the Margins of Belgian Exports (2008S1-2009S1) by Product and Service Type Total % Change Panel a: Services Goods Transport Passenger Transport Auxiliary Services for Transport Service to non-Residents Telecommunication Services Construction Services Financial and Insurance Services Business Services Panel b: Goods Intermediates Capital Goods Consumer Durables Consumer non Durables Energy Other Extensive Margins Countries Services Intensive Margin -22.25 -1.98 -10.62 -0.34 11.66 -0.79 21.49 4.90 -3.77 1.94 -4.21 -0.14 5.13 -2.77 1.59 -0.23 -0.33 1.80 1.99 1.29 -1.47 -0.82 0.26 2.11 -18.94 -5.55 -8.52 -1.47 7.80 2.87 19.27 2.97 -31.24 -23.64 -38.23 -7.74 -44.47 -25.51 -0.61 -1.62 -4.21 0.17 -3.94 -1.84 1.51 1.87 1.99 0.36 0.04 0.28 -31.85 -23.81 -36.00 -8.22 -42.22 -24.33 Note: This table presents the decomposition of the growth rate of Belgian exports between the first semester of 2008 and the first semester 2009 for EU and non-EU countries, type of service exported, multinational status of the firm and type of firm. 24 Table 5: Change in the Margins of Belgian Exports (2008S1-2009S1) by Country of Destination Total % Change Panel a: Services EU -4.30 non-EU -1.14 OECD -1.05 non-OECD -14.85 Panel b: Goods EU -26.73 non-EU -27.27 OECD -26.64 non-OECD -27.75 Extensive Margins Countries Services Intensive Margin -1.08 0.50 -0.84 -0.09 1.94 2.17 1.74 2.80 -5.09 -3.73 -1.92 -17.09 -3.15 -0.76 -1.92 -0.59 1.54 4.14 2.09 1.58 -25.50 -29.63 -26.74 -28.45 Note: This table presents the decomposition of the growth rate of Belgian exports between the first semester of 2008 and the first semester 2009 for EU, non-EU, OECD and non-OECD countries. Table 6: Change in the Margins of Belgian Exports (2008S1-2009S1) by Ownership Status Total % Change Panel a: Services MNE non-MNE Foreign Owned Non-Foreign Owned Panel b: Goods MNE non-MNE Foreign Owned Non-Foreign Owned Extensive Margins Countries Services Intensive Margin 7.38 -8.54 3.17 -8.64 0.63 -0.60 0.86 -0.88 2.36 1.80 3.89 0.89 4.24 -8.54 -1.53 -8.64 -29.77 -25.04 -30.32 -22.98 -1.28 -1.44 -2.04 -1.27 2.44 1.65 4.53 0.99 -30.55 -25.19 -31.96 -22.75 Note: This table presents the decomposition of the growth rate of Belgian exports between the first semester of 2008 and the first semester 2009 by ownership status of the firms. 25 Table 7: Change in the Margins of Belgian Exports (2008S1-2009S1) by Firm Characteristics Total % Change Panel a: Services Big Small Financially exposed Financially non-exposed Panel b: Goods Big Small Financially exposed Financially non-exposed Extensive Margins Countries Services Intensive Margin -0.27 -22.65 -1.32 -3.07 -0.38 -1.13 0.25 -0.74 2.60 -0.57 0.97 3.20 -2.42 -21.32 -2.51 -5.37 -27.08 -23.98 -29.68 -23.82 -1.85 -0.48 -1.94 -0.90 2.85 1.28 1.84 3.36 -27.76 -24.58 -29.58 -25.63 Note: This table presents the decomposition of the growth rate of Belgian exports between the first semester of 2008 and the first semester 2009 by size and debt structure of the firms (we define a firm as big if the number of full time equivalent employees is above the median and as financially exposed if the share of financial debts of the firm is above the median). 26 Table 8: Description of the Variables 27 Variable Name Trade Variables: Export of Services Export of Goods Firm-level variables: Dsize Dproductivity Dintermediate share Dshare exp sales Dshare imp interm Dvalue added chain Dext f in dep Dshare debts over liab Dshare debts due af ter one year Dshare f in debt Dshare stock F OR MNE Nace codes Country-level variables: OECD non − EU non − OECD non − EU Exchange rate change GDP growth Description Source 2007-2010 monthly exports of services by firm, service, country 2007-2010 monthly exports of goods by firm, service, country NBB Trade in Services Dataset NBB Trade in Goods Dataset Log of firm size, measured in terms of full-time equivalent employment Log of Value added per worker Share of intermediates over turnover Share of exports over turnover Share of imports over intermediates Exports times imports over turnover Investments minus operating profits over investments Ratio of debts over total liabilities Share of debts due after one year Share of financial debt Ratio of stock over turnover Dummy indicating foreign ownership Dummy indicating a multinational firm NACE rev 1.1 2-digit industry NBB NBB NBB NBB NBB NBB NBB NBB NBB NBB NBB NBB NBB NBB Dummy for countries belonging to the OECD (in 2008) but not to the EU Dummy for countries belonging neither to the OECD nor to the EU % change in the daily exchange rate with the euro between at the 1st of april of each year Average annual growth rate of the countrys GDP OECD and European Commission OECD and European Commission European Central Bank Business Registry Business Registry Business Registry Business Registry Business Registry Business Registry Business Registry Business Registry Business Registry Business Registry Business Registry Survey of Foreign Direct Investments Survey of Foreign Direct Investments Crossroads Bank IMF World Economic Outlook Note: All firm characteristics prefixed with a D are dummy variables that take value one if the firm characteristic is above the NACE rev 1.1 2-digit industry median across all trading firms (so both those trading services and those trading goods) and zero otherwise. All firm characteristics prefixed with a C indicate that we use the actual value of the variable. Table 9: DD Regression on continuing firm-country-service triplets, dummy variables β1 Firm Characteristics Dsize Dproductivity Dintermediate share Dshare exp sales Dshare imp interm Dvalue added chain Dext f in dep Dshare debts over liab Dshare debts due af ter one year Dshare f in debt Dshare stock F OR MNE Country Caracteristics: OECD non − EU non − OECD non − EU Exchange rate change GDP growth Constant Service Dummies Industry Dummies Observations R2 Goods β22009 0.0395b (0.017) 0.0105 (0.017) 0.0096 (0.016) -0.0059 (0.016) -0.0356b (0.015) -0.0005 (0.016) -0.0477b (0.023) -0.0278 (0.019) 0.0197 (0.020) 0.0114 (0.021) 0.0098 (0.021) 0.0033 (0.025) 0.0160 (0.029) -0.0287 (0.029) -0.0108 (0.028) -0.0285 (0.025) -0.0520c (0.028) 0.0558b (0.026) -0.0295 (0.028) 0.0351 (0.029) -0.0053 (0.029) 0.0298 (0.024) -0.0509c (0.027) 0.0315 (0.029) -0.0212 (0.043) -0.0383 (0.038) -0.1604a 0.2848a (0.030) (0.056) -0.0773b 0.1088c (0.034) (0.060) -0.2851a -0.1463 (0.085) (0.134) 0.0101c 0.0139c (0.005) (0.008) -0.1129 (0.176) Yes Yes 650,570 0.0147 Services β1 β22009 -0.0584 (0.207) 0.0986 (0.097) -0.0197 (0.125) 0.0481 (0.105) -0.0998 (0.108) 0.0075 (0.136) -0.0637 (0.080) -0.0876 (0.097) -0.1467 (0.091) -0.0273 (0.103) -0.0892 (0.095) -0.0509 (0.126) 0.0381 (0.079) 0.5499 (0.409) -0.0082 (0.157) 0.0575 (0.186) 0.0283 (0.147) 0.2219 (0.162) -0.0661 (0.208) 0.2205 (0.135) 0.0374 (0.157) 0.2921b (0.140) 0.0034 (0.189) 0.2339 (0.150) 0.0589 (0.172) -0.1199 (0.140) 0.0486 -0.1359 (0.087) (0.155) 0.1760c -0.1961 (0.095) (0.138) 0.1552 -0.2725 (0.369) (0.526) 0.0142 -0.0075 (0.017) (0.027) 0.1076 (0.408) Yes Yes 23,249 0.0529 Note: This table presents the estimated coefficients for the variable of interest in this study (the complete table is available upon request) β1 refers to the estimated effects in normal time and β22009 refers to the estimated effect of the same variables during the crisis in 2009. On the left side estimates for exports of goods and on the right for exports of services. Multi-level clustered standard errors in parentheses (at the firm, product and country level). a p<0.01, b p<0.05, c p<0.1. 28 Table 10: DD Regression on continuing firm-country-service triplets, continuous variables β1 Firm Characteristics Csize Cproductivity Cintermediate share Cshare exp sales Cshare imp interm Cvalue added chain Cext f in dep Cshare debts over liab Cshare debts due af ter one year Cshare f in debt Cshare stock F OR MNE Country Caracteristics: OECD non − EU non − OECD non − EU Exchange rate change GDP growth Constant Service Dummies Industry Dummies Observations R2 Goods β22009 0.0260a (0.008) 0.0333b (0.015) 0.0055 (0.005) -0.3424 (0.338) -0.5168 (0.365) 0.0133 (0.009) 0.0000 (0.000) -0.0195 (0.043) 0.0505 (0.043) -0.0236 (0.038) 0.0013 (0.131) -0.0286 (0.029) -0.0146 (0.027) β1 Services β22009 -0.0187c (0.011) -0.0422b (0.019) -0.0039 (0.006) -0.5167 (0.392) -0.1397 (0.583) 0.0011 (0.014) 0.0001 (0.000) -0.0429 (0.058) -0.0474 (0.064) -0.0029 (0.046) 0.0336 (0.140) -0.0085 (0.043) -0.0116 (0.041) -0.0000 (0.000) -0.0090 (0.039) 0.0171 (0.187) -0.2919 (0.358) 0.3992 (0.376) -0.0035 (0.002) 0.0000 (0.000) -0.2010 (0.225) 0.1781 (0.216) -0.2391 (0.181) -0.0582 (0.423) -0.0280 (0.117) 0.0496 (0.091) -0.0000 (0.000) -0.0017 (0.052) -0.2233 (0.289) 0.5367 (0.541) -1.5467b (0.717) 0.0162a (0.004) -0.0000 (0.000) 0.2775 (0.353) -0.1919 (0.379) 0.2993 (0.323) 0.1566 (0.647) -0.0079 (0.159) -0.1949 (0.161) -0.1665a 0.2906a (0.030) (0.056) -0.0824b 0.1124c (0.034) (0.062) -0.2907a -0.1369 (0.089) (0.141) 0.0095c 0.0141c (0.005) (0.008) -0.0404 (0.145) Yes Yes 650,570 0.0143 0.0536 (0.087) 0.1864b (0.091) 0.1650 (0.369) 0.0152 (0.017) 0.0655 (0.371) Yes Yes 23,249 0.0566 -0.1356 (0.156) -0.2175 (0.138) -0.2568 (0.534) -0.0098 (0.026) Note: This table presents the estimated coefficients for the variable of interest in this study (the complete table is available upon request) β1 refers to the estimated effects in normal time and β22009 refers to the estimated effect of the same variables during the crisis in 2009. On the left side estimates for exports of goods and on the right for exports of services. Multi-level clustered standard errors in parentheses (at the firm, service or product and country level). a p<0.01, b p<0.05, c p<0.1. 29 Table 11: Treated and Control Group Characteristics log Size log Productivity log Capital Intensity log Average Wage MNE FOR Control 2.983 (0.0445) -2.302 (0.020) -3.952 (0.043) -2.867 (0.015) 0.052 (0.006) 0.130 (0.009) Treated 3.091 (0.046) -2.350 (0.021) -4.020 (0.052) -2.893 (0.013) 0.055 (0.006) 0.132 (0.009) Diff -0.107 (0.064) 0.048 (0.029) 0.068 (0.067) 0.026 (0.020) -0.002 (0.008) -0.002 (0.013) Note: This table presents the characteristics and the differences of Control and Treated groups, in terms of size, productivity, capital intensity, average wage, multinational status and foreign ownership. In the last column there is the t statistic of a t-test where the null hypothesis is that the difference is equal to zero. b indicates significance of the difference at the 5% level. Table 12: DDD Regression on continuing firm-countries, GDP Growth Panel a: Bi-Exporters (1) All γ32009, GDP Growth -0.0549b (0.027) Firm-Year Dummies Yes Observations 22,162 R2 0.0763 Panel b: Matched Mono Exporters (1) All γ32009, GDP Growth Firm-Year Dummies Observations R2 -0.0466a (0.013) Yes 40,272 0.0717 (2) Durables (3) Non-Durables -0.0802b (0.038) Yes 14,117 0.1093 -0.0427 (0.027) Yes 16,129 0.1063 (2) Durables (3) Non-Durables -0.0560b (0.022) Yes 21,685 0.0945 -0.0334b (0.017) Yes 29,402 0.0857 (4) Business Services -0.0943b (0.040) Yes 12,099 0.0798 (5) Transport Services -0.0617a (0.022) Yes 10,671 0.0798 (6) Services to Non Residents -0.2479 (0.153) Yes 948 0.1562 (7) Telecommunication Services 0.0415 (0.051) Yes 1,329 0.1208 (8) Construction Services 0.1391 (0.118) Yes 1,243 0.1017 (9) Financial Services 0.0432 (0.150) Yes 1,865 0.0874 (10) Other Services -0.0360 (0.088) Yes 2,860 0.0979 (4) Business Services -0.1087c (0.065) Yes 11,908 0.0805 (5) Transport Services -0.0480 (0.032) Yes 18,931 0.0767 (6) Services to Non Residents 0.2993c (0.172) Yes 1,336 0.1013 (7) Telecommunication Services -0.6340c (0.347) Yes 3,498 0.0588 (8) Construction Services -0.0280 (0.044) Yes 2,597 0.0879 (9) Financial Services -0.0039 (0.025) Yes 4,270 0.0506 (10) Other Services -0.0534a (0.019) Yes 2,201 0.1128 Note: This table reports the estimated coefficients for γ3t, GDP Growth from equation 3 for the firms that trade both goods and services (Bi-Exporters) and for the matched firms exporting either goods or services (Mono-Exporters). Clustered standard errors in parentheses (at the country level). a p<0.01, b p<0.05, c p<0.1. Table 13: DDD Regression on continuing firm-countries, Credit Constraints Panel a: Bi-Exporters (1) All (2) Durables γ32009, F IN 0.4557 0.0938 (0.340) (0.408) Country-Year Dummies Yes Yes Observations 18,790 12,127 R2 0.0313 0.0347 Panel b: Matched Mono Exporters (1) (2) All Durables γ32009, F IN Country-Year Dummies Observations R2 0.0275 (0.123) Yes 26,317 0.0331 -0.0031 (0.128) Yes 13,535 0.0180 (3) Non-Durables 0.2772 (0.378) Yes 13,817 0.0327 (3) Non-Durables 0.1837 (0.183) Yes 19,235 0.0413 (4) (5) (6) Business Transport Services to Services Services Non Residents 0.6000 0.1076 2.7639 (0.385) (0.232) (1.719) Yes Yes Yes 10,360 9,006 699 0.0372 0.0597 0.2547 (7) (8) Telecommunication Construction Services Services -0.6523 1.2524 (1.201) (1.099) Yes Yes 1,073 1,201 0.1940 0.1788 (9) (10) Financial Other Services Services 1.7447b 0.5249 (0.738) (0.762) Yes Yes 1,708 2,536 0.1834 0.1169 (4) (5) (6) Business Transport Services to Services Services Non Residents -0.0190 0.1138 -4.1627 (0.246) (0.321) (9.027) Yes Yes Yes 7,182 12,503 1,043 0.0699 0.0644 0.3433 (7) (8) Telecommunication Construction Services Services -0.0056 -0.7648 (0.280) (2.102) Yes Yes 3,079 1,933 0.2400 0.2421 (9) (10) Financial Other Services Services -0.2652 0.9249 (1.637) (2.144) Yes Yes 2,895 2,090 0.1614 0.1779 Note: This table reports the estimated coefficients for γ3t, F IN from equation 4 for the firms that trade both goods and services (Bi-Exporters) and for the matched firms exporting either goods or services (Mono-Exporters). Clustered standard errors in parentheses (at the firm level). a p<0.01, b p<0.05, c p<0.1. 30