Volume 13 Issue 1 Issue 1 - 2 Article 3 12-21-2011 Firm size and the extensive margin Balázs Muraközy László Halpern Follow this and additional works at: https://www.ebrjournal.net/home Recommended Citation Muraközy, B., & Halpern, L. (2011). Firm size and the extensive margin. Economic and Business Review, 13(1). https://doi.org/10.15458/2335-4216.1218 This Original Article is brought to you for free and open access by Economic and Business Review. It has been accepted for inclusion in Economic and Business Review by an authorized editor of Economic and Business Review. 27 ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 2011 | 27–50 FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS1 lászló hAlPern2 bAlázs mUrAKözY3 AbstrAct: In this paper we rely on firm-product-destination level data to analyze Hun- garian trade expansion between 1992 and 2003. We decompose trade growth to the number of firms, the number of markets and products per firm, and analyze these dimensions by firm size. We also distinguish between new firms and continuing exporters. The results sug- gest that the majority of small exporters exit exporting after a few years, but the survivors grow very quickly in every dimension. Firm dynamics across size categories is intensive. Large exporters grow slowly, and macro shocks, destination market and product heteroge- neity strongly affect their performance Keywords: export, extensive margin, firm size, transaction level data, Hungary Jel: F12, L25 1. introdUction Recent models of international trade building on firm-level heterogeneity emphasize the role of the extensive margin, i.e. the change in the number of exporting firms or exported products resulting from trade liberalization (Melitz, 2003). On the empirical side, Ber- nard et al. (2007) estimated the relationship between gravity variables and both the ex- tensive and intensive margins (export volume per product per firm) of US exports. Their results show that both the number of firms and the average number of exported products per firm are increasing in the partner country’s GDP, but strongly decreasing in distance, while the intensive margin is increasing both in GDP and distance. Mayer and Ottaviano (2007) decomposed trade volume to a number of different margins for European coun- tries. They showed that when explaining exports of a country, variation in the number of exporting firms is the most important predictor of exports across destination countries, followed by the number of exported products. 1 This paper is produced as part of the ‘European Firms in a Global Economy: Internal policies for external competitiveness (EFIGE)’, a collaborative project funded by the European Commission’s Seventh Framework Programme (contract number 225551). It is part of the ’Center for Firms in the Global Economy (CEFIG)’ network, too. The authors thank Emília Csiffáry for excellent research assistance. 2 Institute of Economics HAS, CEPR, CEU. 3 Institute of Economics HAS ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201128 Eaton et al. (2008) use transaction-level data from Colombia to estimate export dy- namics. They find that nearly half of exporting firms are new exporters every year, and that most of these firms exit in a few years. As these firms are usually very small, year-to-year changes in aggregate export volume are dominated by the sales of large and stable exporters. However, a few firms from every cohort of new exporters expand rapidly. The aim of this paper is to decompose Hungarian export growth to the extensive and the intensive margins using firm-product-destination level data between 1992 and 2003. The detailed nature of our dataset makes possible to analyze different dimen- sions of the extensive margin. First, total exports can be decomposed to the number of firms (firm-extensive margin) and export volume per firm. Export volume per firm can be further decomposed to within-firm extensive margin and within-firm intensive margin. The extensiveness of firm-level export activities is reflected by the number of export markets served by a firm on average (destination extensive margin) and the number of products exported by the firm (product extensive margin). For a full decom- position, however, one also needs the number of firm-country pairs the firm exports to (within-firm extensive margin). An important contribution of this paper is the analysis of all these margins of exporting, distinguishing between the margins of new and con- tinuing firms as well. Our second important contribution is that we decompose trade growth and its margins by firm size to compare the export growth of different firms. In this exercise we define firm size in terms of exports rather than employment or revenue, as our main interest lies in heterogeneity by export volume. This exercise uncovers some characteristic dif- ferences between small and large exporters. Small exporters are very likely to exit, but surviving small exporters grow quickly on average. As a result, while new exporters do not add too much to export volume in the short run, their contribution to the aggregate trade volume becomes very large in a longer term. Such stylized facts may help explain firm-level exporting decisions, and suggest that different policy approaches may be ap- propriate for small and large exporters. The effect of firm size is also analyzed by Eaton et al. (2008) for Colombia. Our results are comparable to that study, and we show some interesting differences between the two countries. In this time period Hungary was an interesting place to address these questions. At the beginning of our sample, Hungarian trade was still declining as a result of the collapse of the former Soviet market and transitional recession. At about 1994-1995, following macroeconomic reforms and restructuring, Hungary began to integrate strongly into the EU single market, and started its period of strong export-led growth. At the end of the period Hungarian growth was slowing down again as a result of misguided policies, parallel with the full integration of the country into the EU, becoming a full member on 1st May 2004. L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 29 2. dAtA The data used for our empirical analysis were obtained from the Customs Statistics. The dataset consists of all Hungarian exports between 1992 and 2003. One observation in the database is the export of product i by firm j to country k in year t.4 The product dimension of the dataset is highly disaggregated; it is broken down to 6-digit Harmonized System (HS) level. We define a product as a 6-digit category, although using more aggregated (4-digit) categories does not change our results. "Motor cars and vehi- cles for transporting persons" is an example for a 4-digit category, while "Other vehicles, spark-ignition engine of a cylinder capacity not exceeding 1,500 cc" is an example of a 6-digit category. Note that in most cases (like in the car example) further disaggregation of the data would not reduce potential quality differences within each category to zero. As a consequence, during the following analysis we define a product as a 2-digit category because further disaggregation would yield too much similarity between categories. The dataset includes both export values and quantities at this highly disaggregated level, thus unit values are calculated as the ratio of these two variables. The customs database can be merged with balance sheet data, which includes industry identifier of the firm. This data also includes main financial indicators and the number of employees. We drop exports of individual entrepreneurs and individuals limiting our analysis to proper firms. In particular, we focus on manufacturing firms, as most trade theories are more easily applied in case of these firms than agricultural firms or whole- salers and retailers. In terms of numbers such exporters are the majority: in the customs datasets there are 79,348 exporters, from which only 13,540 are proper manufacturing firms. These manufacturing firms, however, were responsible for 86.8 percent of Hungar- ian exports in 2003. We have calculated the results for these other sectors as well, and the figures show that trade is even more dynamic in these sectors compared to manufactur- ing. However, the qualitative patterns are similar. Finally, to reduce noise, exports below US$ 2000 will be disregarded. One motivation is that these smaller export shipments follow different patterns than what is supposed by standard trade theories (Békés and Muraközy, 2011). 3. methodologY In terms of methodology, we follow Eaton et al. (2008) in our baseline tables, and extend the approach in order to find more patterns, especially related to the product dimension of our dataset. The basic cross-sectional decomposition of total export volume in year t to country n, Xn(t) has two components, the number of firms and the average export per firm: lnXn(t) = lnNn(t) + lnX − n(t), 4 A more detailed description of our data can be found in Békés et al. (2011). ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201130 where Nn(t) is the number of exporting firms, and X − n(t) is the average export revenue of these firms. As we are more interested in the dynamic rather than the cross-sectional role of the ex- tensive margin, following Eaton et al. (2008) we further decompose export growth into the share of continuers, entrants and exiting firms: where XnHU(t) denotes Hungarian exports to country n in year t, and xn(j,t) is the export of firm j to country n. CNnt−1,t represents (pairwise) continuers that exported both in t−1 and t, ENnt−1,t denotes (pairwise) entrants, which did not export t−1 but exported in t, and EXnt−1,t is (pairwise) exiting firms, which exported in t−1 but did not in t. NENnt−1,t and NEXnt−1,t represents the number of entrants and exiting firms, respectively. The left-hand side of the equation measures the growth of Hungarian exports to country n in year t. The first line of the right-hand side is the contribution of pairwise continuer firms. It is decomposed into two terms. The first represents the share of these firms in year t, while the second one is the export growth of these firms. The second line shows the contribution of pairwise entrants. The first term in this line is the potential contribution of entrants, assuming that these new firms had the same aver- age export volume as those of the average firm in t−1. The second term shows the size difference between year t entrants and the year t−1 average firm. The third line represents the contribution of exiting firms. Similarly to the entrants, it is composed of two terms: (i) what would be the contribution of exiting firms if they had the same average export volume as those of the average firm in t−1, and (ii) the term correcting for the difference in export revenue. XnHU(t) − XnHU(t − 1) [XnHU(t − 1) + XnHU(t)] 2 = [xn(j,t − 1) + xn(j,t − 1)] 2∑j∈CNnt−1,t [xnHU(t − 1) + xnHU(t)] 2 )( ∑j∈CNnt−1,t [xn(j,t) − xn(j,t − 1)] )(∑j∈CNnt−1,t [xn(j,t − 1) + xn(j,t)] + NEN n t−1,txn (t − 1) [XnHU(t − 1) + XnHU(t)] 2 + ∑j∈ENnt−1,t [xn(j,t) − xn(j,t − 1)] [XnHU(t − 1) + XnHU(t)] 2 − NEX n t−1,txn (t − 1) [XnHU(t − 1) + XnHU(t)] 2 − ∑j∈ENnt−1,t [xn(j,t) − xn(j,t − 1)] [XnHU(t − 1) + XnHU(t)] 2 L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 31 4. decomPosition oF hUngAriAn exPort groWth Figure 1 provides a view on the role of extensive margin in Hungarian trade. It shows the relationship between (ln) total export volume and the (ln) number of firms for each destination in 2003. Similarly to Eaton at al (2008), it shows a strong positive and log- linear relationship between the two variables. The slope of the line is 0.56, suggesting that doubling the market size is associated with 56 % more exporting firms. Figure 1: Total export by destination and number of firms, 2003 Table 1 shows the number, the total export revenue and the average export revenue for entering, continuing, exiting and single-year firms in each year. Note that single-year exporters are excluded both from the set of entering and exiting firms. These measures of export performance increased steadily during the period under study. The number of exporters nearly doubled from about 3,000 to nearly 5,800. Total trade volume increased sixfold, and as a result, the growth in average export by firm increased from about US$ 2 million to US$ 6.4 million. The extensive growth was more important until 2001, after which the number of exporting firms stabilized, but export volume per firm still grew fast. Also, the Russian crisis in 1999 led to a decrease in the number of exporters and slowed down the growth of total exports. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201132 The table indicates rapid dynamics: a large share of firms enter and exit exporting every year. On average, firm entry was larger to a great extent than exit in seven years, but there is an interesting pattern from 1999. In that year, reflecting the Russian crisis, firm entry was relatively low, while exit remained about the same, so the latter almost outweighed the former. In the following year, namely 2000, firm entry rose significantly and exit declined considerably as well, so the gap between the two became very large. For the last two years, firm exit rose relatively high, while entry dropped. Entering and exiting firms are 5-10 times smaller than the average continuing firm. In some years, however, with the entry or exit of large firms, the average entering and exiting firm size becomes very large. It is also clear, that the overwhelming majority of exports is realized by pairwise continuing firms, and they are responsible for the majority of year-to-year export growth as well. Table 1: Entering, exiting, continuing and single-year exporters, 1992-2003 Number of firms Entering Continuing Exiting Single-year Total Year (t) 1992 - - - - 3,068 1993 949 1,636 438 343 3,366 1994 804 2,119 466 335 3,724 1995 673 2,478 445 414 4,010 1996 782 2,688 463 351 4,284 1997 792 3,024 446 381 4,643 1998 719 3,296 520 472 5,007 1999 578 3,483 532 379 4,972 2000 879 3,693 368 342 5,282 2001 689 3,848 724 478 5,739 2002 765 3,785 752 519 5,821 2003 - - - - 5,792 Total Value of exports (million US$) Entering Continuing Exiting Single-year Total Year (t) 1992 - - - - 6,348 1993 546 4,810 281 45 5,678 1994 525 6,760 174 23 7,481 1995 324 8,860 377 65 9,625 1996 1,070 11,300 273 67 12,740 1997 416 15,600 255 31 16,328 1998 315 19,600 319 27 20,222 1999 327 19,600 2,390 25 22,336 2000 1,710 21,200 485 83 23,491 2001 415 25,500 1,500 21 27,470 2002 331 29,200 563 33 30,079 2003 - - - - 36,856 L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 33 Exports per firm (thousand US$) Entering Continuing Exiting Single-year Total Year (t) 1992 - - - - 2,069 1993 575 2,938 641 132 1,687 1994 653 3,190 372 70 2,009 1995 481 3,575 847 157 2,400 1996 1,362 4,217 589 191 2,974 1997 526 5,167 571 81 3,517 1998 439 5,935 613 56 4,039 1999 565 5,626 4,490 66 4,492 2000 1,941 5,745 1,318 242 4,447 2001 602 6,636 2,067 45 4,786 2002 433 7,702 749 63 5,167 2003 - - - - 6,363 Table 2 presents the decomposition of export growth for total Hungarian exports. The first column shows total growth. Between 1992 and 1993 exports were still declining with 11 % as a consequence of collapsing eastern export markets and strong domestic transformational recession. From 1994 onwards, however, massive export growth be- gan, with reaching its peak in 1996, increasing by 28% (in US$ terms). Column 2 shows the share of continuers. In 1993, this was 85%, small relative to 97-98% at the end of the period. The latter number suggests that pairwise continuers are responsible for the overwhelming majority of export volume. This does not seem to be unique in Hungary; Eaton et al. (2008) reports very similar values for Colombia. Except in 1993, continuers were able to increase robustly their exports in every year, contributing with nearly 100% to total export growth in each year. In Colombia, in 3 of 9 years continuers’ export de- creased significantly, together with total exports. Analyzing the number of entering and exiting firms, some characteristic patterns emerge. First, the number of entering firms was 7-8 percentage points larger than that of exiting firms until 2001, showing a very large increase in terms of the firm-extensive margin. In line with Table 1, the number of entrants and exiting firms was about equal in 2002 and 2003. In terms of absolute values, as can be expected, the share of enter- ing firms decreased to a large extent: from 44% to about 20%. For exiting firms this measure was smaller during our sample period, but decreased in a similar way. Third, relative to Colombia the share of entering and exiting firms is low: Eaton et al. (2008) reports numbers between 35 and 45 % for most years. Both entering and exiting firms are smaller than the average firm in the previous year. Here, however, it can be impor- tant that we restrict our attention to proper firms and drop export transactions below US$ 2000. During the whole decade, a somewhat different picture emerges. First, firms exporting through the whole period were responsible for the majority of trade growth. Second, a very large number of firms entered during this long period, which exported much above ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201134 the average on 2003. Third, exiting firms were about as productive as the average firm, thus exit did not contribute significantly to export growth. We draw two main conclusions from this dynamic decomposition. First, in every year there is very significant entry and exit from exporting. Second, nearly 100 percent of year-to-year export growth comes from the intensive margin, the increasing exports of pairwise continuing firms. Table 2: Contribution of pairwise entry and exit to the growth of total manufacturing exports between t-1 and t (%) Contribution of pairwise continuers Contribution of pairwise gross entry Contribution of pairwise gross exit Year (t) Growth of exports Continuers’ share in t-1 exports Growth of exports by continuers Added number of firms Exports of entering firms relative to the average Dropped number of firms Exports of exiting firms relative to the average 1993 -11 85 0 44 -35 -34 13 1994 27 93 26 29 -21 -20 15 1995 25 97 24 26 -21 -19 17 1996 28 93 23 24 -14 -18 14 1997 25 97 25 24 -21 -17 14 1998 21 98 21 23 -21 -16 14 1999 10 98 10 18 -17 -19 17 2000 5 91 9 24 -16 -18 7 2001 16 98 16 20 -19 -12 10 2002 9 97 14 21 -20 -20 15 2003 20 97 18 19 -15 -20 18 1992-2003 141 30 112 47 77 -21 5 5. exPorter size And exPort groWth In this section, we analyze the relationship between exporter size and export growth with different approaches. First, for continuers and exiting firms, we repeat the decom- position exercise for each size quintile to see how firm size affects differences in the de- composition of export growth. We also ask the question, whether patterns are different for different destination markets and products. Second, by presenting a transition matrix and regression results, we ask how frequently firms move between quintiles. As a natu- ral expansion of this, we calculate the trade growth of entrant cohorts, and show their contribution to the total Hungarian export volume in the long run. Third, we investigate L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 35 further the firm-level trade growth by showing how firms expand the number of their export markets and exported products. The effect of exporter size Eaton et al. (2008) emphasizes that Gibrat’s law does not characterize export growth; smaller exporters increase their exports more than proportionally. Following their work, we decompose the export growth of continuing exporters by quintiles. Table 3 shows these results. Firms are classified to quintiles based on their export volume in year t−1. Note that this calculation is based on firms exporting in year t−1, so entry is not taken into account. This has the advantage that there is no composition effect. The table shows export growth of continuers, export growth corrected by exiting firms, and total export volume of firms in the different quintiles (averaged over year t and t−1). Several patterns can be seen in this table. First, as shown for example by Bernard et al. (2007) and Mayer and Ottaviano (2007), export revenues are highly skewed: in 2003 the contribution of the first quintile was US$ 49.4 million, while it was above US$ 36 billion for the largest quintile.5 Also, the results show that total export in the first quintile was changing little during more than 10 years, while it increased nearly fivefold for the largest quintile. The increasing skewness of the exporter size distribution can also be observed in Colombia, although at a lower degree. The widening gap between the average first and fifth quintile firm can be interpreted as a confirmation of the intra-industry reallocation prediction of heterogeneous firm theories: as a result of trade liberalization, more productive firms are able to expand their export sales rapidly. While it is a possible explanation, the entry of multinational firms may have played a much more important role in practice. The composition of the largest quintile changed radically between 1992 and 2003. In 1992, firms in the top quintile were mainly state-owned post-socialist giants, but after 1996, the overwhelming majority consisted of multinational affiliates. The numbers for trade growth strongly reject Gibrat’s law. While the smallest exporters increased their exports well above 100 percent each year, the export growth of largest firms was between 5 and 15 percent in most years. One cannot see a clear trend in export growth for most quintiles, except for the two largest. In 1993, continuing firms suffered a decrease at these quintiles reflecting the loss of their main export markets. In more recent years, however, continuers in the top quintiles were able to increase their export volume in a stable way. Smaller firms, on the other hand, increased robustly their export volumes even in the early years. This comparison shows the duality of the economy: for- mer state-owned firms struggled for survival, while dynamic small new exporters were able to rapidly enter foreign markets in this phase of transition. The effect of exit seems to be qualitatively unimportant for most quintiles and years. However, it proved quite important for the largest firms in the beginning of the period, when large exporters disappeared or were radically restructured. 5 The table also shows that outliers may affect strongly the yearly results. The rapid growth of a few foreign- owned firms in 1996, for example, leads to very large increase in the first two quintiles. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201136 Ta bl e 3: E xp or t g ro w th b y qu in til es o f v al ue o f e xp or ts in y ea r t -1 , c on tin ui ng a nd ex iti ng m an uf ac tu ri ng fi rm s Qu in til e 1 Qu in til e 2 Qu in til e 3 Qu in til e 4 Qu in til e 5 Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t -1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t -1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt G row th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt G row th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g Fir ms (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g Fir ms (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ye ar (t ) 19 93 16 0 15 8 32 76 70 51 40 33 15 6 15 1 49 0 -5 -2 8 4,9 90 19 94 17 2 17 1 46 12 6 12 4 91 76 74 20 5 48 45 55 4 18 12 5,4 10 19 95 15 2 15 1 37 82 80 86 53 51 25 7 35 33 80 2 20 17 7,1 80 19 96 19 5 19 4 40 8 15 0 14 9 25 4 36 35 28 3 51 49 10 50 7 2 8, 62 0 19 97 14 1 13 9 39 10 3 10 2 13 0 38 37 29 4 16 15 95 3 23 21 12 ,90 0 19 98 12 9 12 7 34 87 86 11 9 48 47 32 4 22 21 10 40 20 18 16 ,6 00 19 99 12 5 12 3 36 53 52 10 0 21 20 30 2 7 6 10 80 10 8 19 ,6 00 20 00 94 91 23 59 57 97 30 29 31 9 13 12 10 60 7 -5 20 ,5 00 20 01 11 1 11 0 24 64 63 89 18 17 28 1 18 17 11 00 16 14 23 ,8 00 20 02 11 0 10 7 26 52 50 82 27 25 28 0 13 12 10 90 13 7 27 ,10 0 20 03 10 8 10 5 27 68 66 10 2 36 35 30 7 26 25 11 70 17 16 31 ,10 0 An nu al Av er ag e 13 6 13 4 67 84 82 10 9 39 37 27 3 24 21 94 4 13 8 16 ,2 00 L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 37 These average patterns are similar for different destination markets, as can be inferred from Table 4. Small firms were able to increase their exports rapidly, well above 125% per year on each of the 10 most important destination markets (in terms of the number of exporting firms). The largest continuing firms were also able to increase their export volume on average to each country. In the top quintile the effect of multinationals is ob- vious: the largest trade growth can be observed to countries of large multinationals being present in Hungary: Germany, the Netherlands and France. We have also decomposed export growth by the industry of the firm (Table 5). The firm size distribution is very different in different industries. In the textiles the exporter rev- enue in the first quintile is very large compared to other industries, but this is not the case for the largest quintile. As a result, skewness is low in textiles compared to other industries. Chemicals is the other extreme of the skewness distribution, where the small- est quintile exported US$ 4 million, while the export volume of the largest quintiles was more than US$ 2 billion on average. Theoretically skewness of the export distribution should be related to Pareto-k parameter of firm productivity distribution. Melitz and Ottaviano (2008, p 45.) calculate this for different industries in Italy and France. The estimated Pareto-k is low in textiles in both countries, which is in line with our results. The Pareto-k of chemicals is also low, however, especially in France, which would predict a relatively low skewness of export distribution in this industry. The large skewness of Hungarian chemical exports can be explained by the fact that a few very large pharma- ceutical firms are operating in Hungary, affecting strongly the size distribution. In terms of export growth, there is no evident difference across industries in the lower quintiles: there is a rapid growth for smaller firms. The only exception is the textiles, where the growth of ‘only’ 103 percent is significantly different from the growth rates in other industries, 120-140 percent. Industry differences are more pronounced for larger firms. Large machinery firms (mainly multinational affiliates) were expanding their export vol- ume with a robust 19 percent per year. The slowest growth in the larger quintiles can be observed for food and textiles, where – taking account of exit – export growth of firms exporting in t−1 was -4 and -5 percent, respectively. This suggests that the duality between small and large firms was the most important in these industries: a steady decline in export revenues of large firms was paralleled with strong export growth of small firms. We were interested whether product-level heterogeneity is related to the patterns of export growth. For this, we decomposed export growth by the homogeneity of the product using the liberal classification of Rauch (1999). The results are shown in table 6. There is a very large difference in terms of skewness by homogeneity. The ratio of total exports of the smallest and largest quintile is 180 for homogeneous products, 337 for reference priced goods and 457 for differentiated goods. Productivity distribution in the homogeneous goods industries has a lower skewness parameter, reflecting smaller firm size differences. Similarly to previously examined dimensions of heterogeneity, export growth differenc- es by product homogeneity are less obvious for smaller firms than for larger ones. In the bottom quintile average export growth of firms already exporting in year t−1 is 123% per ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201138 Ta bl e 4: E xp or t g ro w th b y qu in til es o f v al ue o f e xp or ts in y ea r t -1 , c on tin ui ng a nd ex iti ng m an uf ac tu ri ng fi rm s Te n m os t p op ul ar d es tin at io ns . A nn ua l A ve ra ge 19 92 -2 00 3 Qu in til e 1 Qu in til e 2 Qu in til e 3 Qu in til e 4 Qu in til e 5 Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt G row th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nti nu ing Fir ms (% ) Ex po rt Gr ow th Co nti nu ing - Ex itin g F irm s (% ) Me an To tal ex po rts be tw ee n t-1 an d t (m illi on US $) Ge rm an y 13 9 13 7 27 89 87 72 31 29 13 6 19 16 41 3 12 6 5,9 10 Au str ia 13 6 13 2 7 83 79 15 42 37 32 26 21 95 8 1 1,3 80 Ro m an ia 14 2 13 6 3 98 92 5 64 56 11 35 27 28 8 -2 23 5 Ita ly 15 0 14 7 8 88 83 17 40 36 39 17 11 10 3 7 -1 87 0 Slo va kia 13 8 13 2 3 80 73 4 51 45 8 27 21 20 15 7 19 3 Fr an ce 14 3 13 9 5 90 86 12 42 37 23 22 16 60 16 8 78 3 Sw itz er lan d 12 6 12 0 2 74 68 5 46 39 9 14 7 19 2 -6 17 2 Cz ec h R ep ub lic 12 8 12 3 2 90 85 5 50 45 10 32 28 29 9 2 27 5 Th e N et he rla nd s 14 4 14 0 4 85 80 8 47 41 15 31 23 52 16 3 62 2 Po lan d 14 4 14 0 3 10 2 96 7 60 55 13 34 27 35 12 5 31 2 L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 39 Ta bl e 5: E xp or t g ro w th b y qu in til es o f v al ue o f e xp or ts in y ea r t -1 , c on tin ui ng a nd ex iti ng m an uf ac tu ri ng fi rm s M an uf ac tu rin g ca te go rie s. An nu al A ve ra ge 19 92 -2 00 3 Qu in til e 1 Qu in til e 2 Qu in til e 3 Qu in til e 4 Qu in til e 5 Ma nu fac tu rin g c ate go ry Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s ( %) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Fo od an d T ob acc o 12 9 12 7 6 67 64 18 35 31 55 15 9 17 0 4 -4 1,1 40 Tex tile s 10 3 10 2 14 35 34 43 16 14 11 2 6 2 25 8 3 -5 1,2 70 Wo od , p ap er an d p rin tin g 12 2 11 7 2 66 61 4 29 24 10 14 10 30 12 6 41 7 Ch em ica l in du str y 13 2 13 0 4 90 88 9 30 28 20 22 20 82 11 7 2, 21 0 Oth er no n-m eta llic pr od uc ts 11 7 11 3 1 50 47 2 46 44 7 13 10 27 5 1 22 8 Me tal pr od uc ts 12 1 11 8 4 59 56 11 38 34 30 22 17 79 7 0 1,0 20 Ma chi ne ry 14 0 13 8 42 10 6 10 5 41 51 49 83 39 37 35 8 19 14 9,5 10 Oth er ma nu fac tur ing 12 6 12 3 2 42 39 3 21 17 9 13 7 29 7 -3 16 4 ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201140 Ta bl e 6: E xp or t g ro w th b y qu in til es o f v al ue o f e xp or ts in y ea r t -1 , c on tin ui ng a nd ex iti ng m an uf ac tu ri ng fi rm s Ra uc h cla ss ifi ca tio n. A nn ua l A ve ra ge 19 92 -2 00 3 Qu in til e 1 Qu in til e 2 Qu in til e 3 Qu in til e 4 Qu in til e 5 Pr od uc t t yp e Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ex po rt Gr ow th Co nt inu ing Fir ms (% ) Ex po rt Gr ow th Co nt inu ing - Ex iti ng Fi rm s (% ) Me an To ta l ex po rts be tw ee n t-1 an d t (m illi on US $) Ho m og en ou s 13 0 12 3 3 79 71 5 47 41 15 25 18 59 2 -11 49 0 Re fe re nc e- pr ice d 14 5 14 0 7 11 0 10 6 13 53 49 28 35 32 12 1 6 1 22 50 Di ffe re nc iat ed 13 7 13 5 24 95 93 81 44 42 16 3 23 21 57 0 15 9 11 10 0 To ta l 13 5 13 4 42 85 84 11 9 41 39 27 5 24 21 92 7 13 7 15 70 0 L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 41 year for homogeneous goods and it is 137% for differentiated goods. In the top quintile, on the other hand, average growth was 2% for homogeneous, 6% for reference priced and 15% for differentiated goods in the largest quintile. Also, the difference in growth rates explained by exit was more than twice as much for homogeneous than for differentiated goods. These results document the fundamental restructuring in Hungarian trade: the declining importance of homogeneous goods was driven by a low growth of exports by continuing firms and significant exit of large homogeneous-goods exporting firms. This was, however, paralleled by the strong growth of small homogeneous good exporters. Firm dynamics across quintiles Let us see how individual firms increase their exports, and move across quintiles. Table 7 presents the transition matrix for the quintiles of export: what is the probability that a firm in quintile i at t−1 will be in quintile j at year t? We also include a non-exporting category, consisting firms which exported for at least one year between 1992 and 2003, but did not export in that year. The matrix is an average of the transition probabilities in all sample years, thus it shows average yearly probabilities. The most obvious characteristic of the matrix is its persistence: firms are likely to remain in the quintile where they are. It is not surprising, that the two most persistent quintiles are the top quintile and the non-exporting category. Generally, persistence decreases with firm size, which can be explained by the large probability that firms exit from ex- porting altogether: this is 48% for firms in the bottom quintile, 26% for firms in quin- tile two, and there is even a probability of 7% that the largest exporters quit the export market every year. Interestingly these exiting probabilities are even larger in Colombia, where there is a 76% probability that firms in the first quintile stop exporting, and this probability is 10% for firms in the top quintile. ‘Upward mobility’ is present too. Small exporters in the first quintile face a 24% prob- ability to move up to a larger quintile, compared to 28% probability of staying in quintile 1. It is less likely that larger exporters move up, but its probability is still significant: for example, firms in the third quintile move up with a probability of 19%. Table 7: Transition matrix for the quintiles of exports to which a firm belongs Initial quintile (x) Final quintile (y) Non-exporting 1 2 3 4 5 Non-exporting 0.87 0.48 0.26 0.15 0.09 0.07 1 0.06 0.28 0.15 0.04 0.01 0.00 2 0.03 0.17 0.36 0.14 0.02 0.00 3 0.02 0.05 0.19 0.48 0.12 0.01 4 0.01 0.01 0.03 0.17 0.63 0.08 5 0.01 0.00 0.01 0.02 0.13 0.84 Notes: Conditional probability of transiting from quintile of exports x in t-1 to quintile y in t ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201142 We checked the robustness of these results with regression analysis. We estimated how firm size, i.e. the initial quartile of the firm relates to the probability of export growth. For comparability with the previous table, we define export growth as a tran- sition to a higher quintile. We model the probability of this 'upward' transition be- tween periods t and t+1 with the firm’s quintile in t and labor productivity6 (turnover/ employees), its 2-digit industry classification and year dummies. We only consider firms exporting in t and exclude firms in quintile 5 in year t, as they cannot move up to another quintile. We estimate a probit model, and report the marginal effects at the sample mean. The results of the regression analysis are in line with the descriptive patterns. Upward mobility is more important for smaller firms, and its probability declines monotonically with firm size. As expected, more productive firms are also more likely to ‘jump up’ to a higher quintile. Table 8: Probability of export growth All firms Continuers Quintile 2 -0.020*** (0.007) -0.024*** (0.007) -0.162*** (0.009) -0.164*** (0.009) Quintile 3 -0.056*** (0.007) -0.064*** (0.007) -0.258*** (0.008) -0.263*** (0.008) Quintile 4 -0.136*** (0.006) -0.155*** (0.006) -0.391*** (0.008) -0.409*** (0.008) Turnover/employees 0.273*** (0.081) 0.358*** (0.082) 0.424*** (0.109) 0.567*** (0.112) 2-digit industry dummies no yes no yes Observations 31,185 31,176 23,948 23,942 Pseudo R-squared 0.0276 0.0379 0.0925 0.102 Log-likelihood -16,797 -16,614 -13,554 -13,401 Note: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Earlier results showed that small exporters grow very fast, but their contribution is quite small in the year of their entry. It is very natural to ask: how much these firms contribute to aggregate export growth in the long run? Following Eaton et al. (2008), we analyse the evolution of different cohorts of exporters in Table 9. The first cohort, firms already exporting in 1992, includes all firms which started exporting in that year or earlier. Similarly to the transition matrix, the table shows the extensive churning of exporters over time: from the cohort entering in 1993, only about 34% exported in 2003. Compared to Colombia, however, churning is relatively low: there only 8% of the cohort entering 6 We have chosen labour productivity because it is easy to interpret. Including TFP estimated with different methods (OLS, fixed effects, Olley-Pakes) did not change the results significantly. L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 43 1997 continued exporting until 2005. Surviving firms, on the other hand, were able to increase their export volume massively. Exporters entering in 1993 and 1994 exported a similar amount on average than firms already exporting in 1992 (and possibly much earlier). These numbers show the dominance of the firm extensive margin in the long run. First, the total export of firms already exporting in 1992 increased by only 56% com- pared to the 480% increase in total export volume. These firms contributed only by 27% to total exports in 2003, and the remaining 73% was realized by firms that start- ed exporting after 1992.7 As a comparison, the contribution of firms already export- ing in 1996 was 76.5% to total exports in 2003. Structural change and rapid trade liberalization in Hungary led to an export growth mainly driven by the entry of new exporters, the firm-level intensive margin. Also, this is not only a result of very early entering firms. Firms entering after 1994 contributed 42% to total export volume in 2003. The table provides information about exporter survival, too. Exit was very frequent in the long run: only 28% of exporters in the 1992 cohort exported continuously in the whole period. Interestingly, however, there are no large differences in shorter term sur- vival across different phases of transition. Table 9: Firms by initial export year cohorts, 1992-2003 Number of firms Year (t) First year of report between 1992 and 2003 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Total 1992 3,068 - - - - - - - - - - - 3,068 1993 2,074 1,292 - - - - - - - - - - 3,366 1994 1,742 949 1,033 - - - - - - - - - 3,724 1995 1,544 797 736 933 - - - - - - - - 4,010 1996 1,413 733 632 590 916 - - - - - - - 4,284 1997 1,312 664 565 503 636 963 - - - - - - 4,643 1998 1,243 620 545 493 519 661 926 - - - - - 5,007 1999 1,161 580 509 420 473 538 580 711 - - - - 4,972 2000 1,092 560 475 382 418 499 507 433 916 - - - 5,282 2001 1,025 540 470 388 401 493 481 402 689 850 - - 5,739 2002 940 497 419 348 356 444 424 345 538 521 989 - 5,821 2003 866 441 389 306 324 400 360 293 456 395 619 943 5,792 7 This is not only a characteristic of manufacturing firms. The result for all exporters is strikingly similar: 26.7 percent. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201144 Value of exports (million US$) First year of report between 1992 and 2003 Year (t) 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Total 1992 6,350 - - - - - - - - - - - 6,348 1993 5,090 591 - - - - - - - - - - 5,678 1994 5,630 1,320 529 - - - - - - - - - 7,481 1995 6,480 1,700 1,090 352 - - - - - - - - 9625 1996 7,060 1,990 1,420 1,690 578 - - - - - - - 12,740 1997 8,060 2,270 2,180 2,460 966 403 - - - - - - 16,328 1998 8,600 2,410 3,970 2,900 1,200 839 305 - - - - - 20,222 1999 8,530 2,360 4,930 3,050 1,180 1,290 697 298 - - - - 22,336 2000 7,850 2,360 5,480 2,460 973 1,050 826 898 1,600 - - - 23,491 2001 8,050 3,030 4,930 2,930 789 1,110 1,000 1,230 3,990 418 - - 27,470 2002 8,320 4,800 5,160 2,620 842 1,160 1,050 974 4,030 787 329 - 30,079 2003 9,950 5,050 6,420 1,660 1,110 1,490 1,480 1,300 5,470 1,010 638 1,270 36,856 Exports per firm (million US$) First year of report between 1992 and 2003 Year (t) 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Total 1992 2,069 - - - - - - - - - - - 2,069 1993 2,453 458 - - - - - - - - - - 1,687 1994 3,234 1,389 513 - - - - - - - - - 2,009 1995 4,196 2,135 1,485 378 - - - - - - - - 2,400 1996 4,997 2,710 2,255 2,865 631 - - - - - - - 2,974 1997 6,140 3,413 3,859 4,884 1,519 418 - - - - - - 3,517 1998 6,918 3,895 7,285 5,879 2,306 1,269 329 - - - - - 4,039 1999 7,349 4,065 9,688 7,258 2,503 2,395 1,201 419 - - - - 4,492 2000 7,186 4,222 11,500 6,444 2,329 2,099 1,629 2,074 1,743 - - - 4,447 2001 7,851 5,604 10,500 7,561 1,967 2,251 2,082 3,059 5,790 491 - - 4,786 2002 8,855 9,665 12,300 7,524 2,365 2,617 2,477 2,822 7,488 1,511 333 - 5,167 2003 11,500 11,400 16,500 5,410 3,415 3,733 4,112 4,453 12,000 2,568 1,030 1,350 6,363 Within-firm extensive margins In the previous section we analyzed how the total export volume of continuing firms changed. In this section we decompose these firms’ export growth to see how they ex- tended the number of their export markets and the export products. We will categorize firms according to the number of their export markets/exported products in year t−1, and calculate the growth in the number of these variables for each group separately. Finally, we calculate the within-firm extensive margin, i.e. the number of destination- product pairs the firms export to, and decompose it in a similar way. Table 10 presents how the most important firm-level variables were related to the firm- level extensive margin in 2003. These variables are the real value added per employee, the capital to labor ratio and the number of employees in relative terms compared to the industry average. As a comparison, the table also includes all firms, which did not export in 2003, using balance sheet data. L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 45 The largest differences can be observed in terms of number of employees followed by added value. Also, there is a strict sorting of firms by these variables both in terms of the number of export markets and number of products. In terms of capital to labor ratio, there is only a strict sorting of firms in terms of destination-product pairs, rather than the two components of the within-firm extensive margin. Table 10: Firms’ characteristics and number of destinations, products and destination- product pairs in 2003 Number of destinations Mean(ry) Mean(rk) Mean(rl) Mean(l) Number of firms 0 0.84 1.08 0.38 15.84 8,515 1 1.01 0.82 0.66 27.81 2,600 2-5 1.20 0.86 1.51 68.54 2,172 6-10 1.58 1.11 3.18 162.61 509 11-30 1.89 0.97 6.45 356.89 439 31-50 2.70 1.64 17.67 920.97 57 50+ 3.51 3.51 44.87 2,831.47 15 Total 1.00 1.00 1.00 48.26 14,307 Number of products Mean(ry) Mean(rk) Mean(rl) Mean(l) Number of firms 0 0.84 1.08 0.38 15.84 8,515 1 1.05 0.83 0.69 28.83 2,800 2-5 1.30 0.87 1.90 88.43 2,423 6-10 1.71 1.31 5.69 313.62 446 11-20 1.71 1.05 13.52 768.71 114 21-50 5.22 0.32 53.36 3,679.11 9 Total 1.00 1.00 1.00 48.26 14,307 Number of destination- product pairs Mean(ry) Mean(rk) Mean(rl) Mean(l) Number of firms 0 0.84 1.08 0.38 15.84 8,515 1 1.00 0.86 0.50 20.84 1,877 2-5 1.12 0.83 1.12 46.95 2,338 6-10 1.43 0.95 2.09 102.66 726 11-50 1.68 1.02 5.21 276.99 767 50+ 3.03 1.59 24.27 1,425.25 84 Total 1.00 1.00 1.00 48.26 14,307 Note: ry, rk and rl are value added per employee, capital to labor ratio and number of employees in relative terms, i.e. ratio of firm i to sector j. l is number of employees. Table 11 shows how continuing firms exporting to different numbers of destinations increased the number of their export markets in the following year. The average number of export markets per firm does not follow any clear trend, which is surprising given the fundamental changes in the Hungarian trade structure. Like in our earlier tables it can be seen that Gibrat’s law does not seem to apply. Continu- ing firms exporting to only one market export to 1.4 markets in the next year, while the growth is only 0.6 for firms exporting to at least 11 markets. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201146 Also, there is important variation over time. One the one hand, it is not surprising, that at the beginning of the period the number of export markets decreased strongly for large firms. On the other hand, this decrease was less pronounced for firms exporting to 6-10 countries than for firms exporting to more than 10 markets, showing that these large exporters were able to keep their export markets to a larger degree even in such turbulent times. Parallel with this, continuing small firms were very fast in expanding the number of their export markets. After the initial period of transition, the growth of small exporting firms slowed down significantly from 0.59 in 1994 to 0.31 in 2003, suggesting that a number of small export- ers exported steadily to one export market. Cyclical effects seem to be more important for larger firms. In 1999, as the consequence of the Russian crisis, the number of export markets decreased for all firms, except the bottom two categories. Table 11: Expanding exporting activity by number of destinations in year t-1, continuing firms Year (t) Growth: number of destinations Total 1 2 3 - 5 6 - 10 10+ 1993 0.55 0.30 -0.07 -0.74 -0.34 4.07 1994 0.59 0.41 0.36 0.40 0.68 4.35 1995 0.49 0.42 0.23 0.34 0.72 4.61 1996 0.42 0.08 0.17 -0.05 0.19 4.56 1997 0.40 0.27 0.18 0.13 -0.23 4.58 1998 0.33 0.24 0.07 -0.11 -0.29 4.49 1999 0.32 0.10 -0.22 -0.06 -0.32 4.38 2000 0.32 0.21 0.20 0.14 0.18 4.52 2001 0.35 0.08 -0.02 0.11 0.14 4.45 2002 0.28 0.10 -0.06 0.12 -0.12 4.47 2003 0.31 0.16 -0.06 0.15 0.00 4.60 Annual Average 0.40 0.22 0.07 0.04 0.06 4.46 This analysis may be upward biased in the sense that firm exit is ignored. To get a more complete picture we present a transition matrix in Table 12. Most entering firms enter only 1 market in the first year, but about 25% of new exporters start exporting to more than one market. Firms exporting to one market exit with a probability of 45%, and ex- pand to new markets with a probability of 13%. Exporters selling to more markets also exit with relatively high probability: even firms exporting to more than 10 markets exit from exporting with a yearly probability of 3%. Also, the matrix shows a ‘downward’ drift: with the exception of firms exporting to 1 market, all other firms are more likely to move downward then upward, suggesting, in line with Table 11, that large exporters are unlikely to expand the number of their export markets quickly. We also checked whether these patterns differ across product groups classified by heterogeneity. We did not find systematic differences. L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 47 Table 12: Transition matrix for number of destinations Final number of destinations (y) Initial number of destinations (x) 0 1 2 3-5 6-10 10+ 0 0.87 0.31 0.17 0.11 0.09 0.07 1 0.09 0.51 0.25 0.08 0.01 0.00 2 0.02 0.12 0.34 0.17 0.02 0.00 3-5 0.01 0.05 0.21 0.49 0.21 0.01 6-10 0.00 0.00 0.02 0.14 0.53 0.10 10+ 0.00 0.00 0.00 0.01 0.15 0.81 Notes: Conditional probability of transiting from exporting to x destinations in t-1 to y destinations in t The second dimension of the within-firm extensive margin is the number of products ex- ported per firm, which is shown in Table 13. Remember, that these are quite aggregated, 2-digit product categories. The pattern by firm size is quite surprising: only small firms, exporting only 1 product increased the number of exported products every year on av- erage. Firms exporting 2-5 products increased the number of products only in 4 years; larger firms reduced the number of their exported products in all years, with only one exception. For large firms, the decrease is very spectacular in 1993 and 1994, suggesting that restructuring led to a serious reduction in the number of their product lines. Table 13: Expanding exporting activity by number of products in year t-1, continuing firms Number of exported products Growth of number of exported products Total 1 2 - 5 6 - 10 11+ Year (t) 1993 0.45 -0.24 -1.50 -2.84 2.30 1994 0.48 -0.02 -0.98 -3.16 2.33 1995 0.48 0.08 -0.62 -0.36 2.48 1996 0.46 -0.09 -0.84 -1.41 2.48 1997 0.41 0.04 -0.40 -3.04 2.58 1998 0.40 -0.02 -0.53 -1.11 2.59 1999 0.33 -0.08 -0.47 -1.31 2.59 2000 0.32 0.02 -0.72 -1.05 2.63 2001 0.36 0.17 0.18 -0.88 2.75 2002 0.32 -0.02 -0.37 -0.51 2.76 2003 0.31 0.01 -0.35 -0.80 2.82 Annual Average 0.39 -0.01 -0.60 -1.50 2.57 These results are reinforced by the transition matrix of the number of products, which is shown in Table 14. On average, firms exporting more than one product decreased the number of products exported. For example, firms exporting 6-10 products face 7% prob- ability of exit, 39% probability of reducing their exported product range and only 9% probability of moving to a higher category. These results are in line with the prediction of Bernard et al. (2011) that as a result of trade liberalization firms drop their marginal products and concentrate on their core competencies. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201148 Compared to the transition matrix for the number of destinations, two other differences can be observed. First, the probability that firms exporting at least 20 products exit in the next year is 11%, which is larger than the exit probability of firms exporting to a large number of markets, suggesting that market-specific fixed costs are more important than product-specific ones. Second, the persistence of the number-of-products transition ma- trix is stronger than that of the number-of-markets matrix. Table 14: Transition matrix for number of products a firm sells Final number of products (y) Initial number of products (x) 0 1 2-5 6-10 11-20 21-50 0 0.87 0.30 0.13 0.07 0.08 0.11 1 0.10 0.51 0.18 0.02 0.01 0.00 2-5 0.03 0.18 0.64 0.32 0.06 0.00 6-10 0.00 0.00 0.06 0.50 0.30 0.00 11-20 0.00 0.00 0.00 0.09 0.52 0.30 21-50 0.00 0.00 0.00 0.00 0.03 0.58 Notes: Conditional probability of transiting from exporting x products in t-1 to y products in t Finally, Table 15 shows the growth of the within-firm extensive margin. Its behavior is very similar to the pattern for the number of export markets. On average this measure declined from 26.5 in 1992 to 23.62 in 1999, and increased after it to 25.1. When decom- posed by firm size, the growth is the largest in the bottom quintile, but it is decreasing with the passage of time. Its growth is negative in some years for the largest firms, espe- cially in the beginning of the period and around the Russian crisis. Table 15: Expanding exporting activity by number of destination-product pairs in year t-1, continuing firms Number of destination-product pairs Growth of number of pairs Total 1 - 4 5 - 9 10 - 20 21 - 44 45+ Year (t) 1993 0.61 -0.35 -1.22 -0.76 -2.90 26.54 1994 0.88 0.19 0.98 1.32 -2.09 26.96 1995 0.73 0.31 1.33 2.10 3.55 26.13 1996 0.59 0.29 0.22 0.27 -3.68 24.59 1997 0.53 0.47 0.21 -0.24 0.63 24.66 1998 0.46 0.24 0.18 0.26 -2.32 24.38 1999 0.33 0.02 -0.23 0.56 -4.36 23.62 2000 0.44 0.24 0.39 1.04 1.23 24.43 2001 0.40 0.42 1.36 1.50 3.72 25.35 2002 0.32 0.12 0.04 1.14 2.54 25.15 2003 0.34 0.13 0.29 0.43 1.32 25.14 Annual Average 0.51 0.19 0.32 0.69 -0.21 25.18 L. HALPERN, B. MURAKÖZy | FIRM SIZE AND ExTENSIVE MARGIN: HUNGARIAN ExPORTS 49 6. conclUsions This paper analyzed the Hungarian export growth between 1992 and 2003, concentrat- ing on different dimensions of the extensive margin. One of our main aims was to present stylized facts on the relationship between firm size and elements of trade growth. The estimates show very strong dynamics in terms of entry and exit to exporting. The new entrants, however, are very small, and do not contribute too much to the year-to- year export growth. New entrants are likely to exit in a few years, but surviving new entrants grow quickly. This also means, that their share in total exports increases fast: in 2003, 73 % of export volume was realized by firms which started exporting after 1992. Hungarian exports are not characterized by Gibrat’s law. Small exporters are growing very quickly, why trade growth of larger firms is smaller. When decomposing across des- tination countries, industries and products, we found small differences in growth rates of smaller firms, but significant heterogeneity for larger firms. When analyzing firm-level exports in more detail, we found similar patterns: surviving small firms are likely to export to new markets and introduce new export products. This growth was smaller for larger firms, and macro shocks (transition and the Russian crisis) affected them strongly. Also, firms exporting the largest number of products consistently reduced the number of their exported product lines. Some of our results are comparable with Eaton et al. (2008) for Colombia, with the quali- fication, that we only considered proper firms and the manufacturing sector. The main patterns in the Hungarian data are in line with their findings, but there are some differ- ences in the details. We have found less entry and exit, but stronger growth of surviving firms and an even larger effect of entry in the long run. Also, we found that larger firms are more likely to exit in Hungary than in Colombia. These differences can be explained by the specificities of transition or by the extent of trade liberalization between Hungary and the EU market. reFerences Békés, G. & Muraközy, B. (2011). Temporary trade. Mimeo, available at http://cefig.eu/view/cfigwp/2/ Békés G., Harasztosi, P. & Muraközy, B. (2011). Firms and Products in International Trade: Data and Patterns for Hungary. Economic Systems, 35 (1), 4-24. Bernard, A. B. et al. (2007). Firms in International Trade. Journal of Economic Perspectives, 21 (3), 105-130. Bernard, A. B., Redding, S. & Schott, P. (2011). Multiproduct Firms and Trade Liberalization. Quarterly Jour- nal of Economics, 126 (3), 1271-1318. Eaton, J. et al. (2008). Export Dynamics in Columbia: Transaction-Level Evidence. In Helpman, E., Marin, D. & Verdier, T. (Eds.), The Organization of Firms in a Global Economy (pp. 231-273). Cambridge: Harvard University Press. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 1-2 | 201150 Mayer, T. & Ottaviano, G. I. (2007). The Happy few: the internationalisation of European firms. New facts based on firm-level evidence. Brussels: Bruegel Blueprint Series (Vol. III). Melitz, M. J. (2003). The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productiv- ity. Econometrica, 71 (6), 1695-1725. Melitz, M. J. & Ottaviano, G. I. (2008). Market size, trade and productivity. Review of Economic Studies, 75 (1), 295–316. Rauch, J. (1999). Networks versus markets in international trade. Journal of International Economics, 48 (1), 7-35.