- Email: [email protected]
Organization and export performance
Economics Letters 146 (2016) 130–134
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Organization and export performance Grigorios Spanos Aix-Marseille University (Aix-Marseille School of Economics), CNRS & EHESS, France
highlights • • • • •
Present new facts on the internal organization of firms and export performance. In the aggregate there is an ordering of the distribution of organizations. The number of layers in firms is positively correlated with export performance. Firms with a greater number of layers export more products to more destinations. Introducing organization into heterogeneous firm models can explain these facts.
article
info
Article history: Received 13 May 2016 Received in revised form 24 July 2016 Accepted 26 July 2016 Available online 30 July 2016
abstract This paper presents new facts on firms’ internal organization and their export performance. I find an ordering of the distribution of organizations and, both across and within firms, firms’ number of layers is positively correlated with their export performance. © 2016 Elsevier B.V. All rights reserved.
JEL classification: F14 L23 Keywords: Trade Firms Margins Exports Organization
1. Introduction Recent research in international trade has shown that firms’ internal organization is important for understanding the outcomes of firms and workers in the economy. In particular, Caliendo and Rossi-Hansberg (2012) show that exporters are on average more productive because they have a greater number of layers. Despite its relevance, empirical evidence on the organization of firms remains limited in international trade. The existing literature has mainly focused on firm size and productivity to examine firms’ export performance, characteristics that are related to organization.1
E-mail address: [email protected]. 1 I broadly define export performance to include the decision to export, as well the number destination markets served, the number of products sold, and the value of exports. http://dx.doi.org/10.1016/j.econlet.2016.07.031 0165-1765/© 2016 Elsevier B.V. All rights reserved.
This paper fills this gap and examines how organization, the number of layers in firms, relates to French manufacturing firms’ export performance over the years 2000–2006. I first examine the distribution of organizations, the percent of firms producing with a given number of layers. I find there is a ranking of distributions. More precisely, the distribution of organizations of exporters, of firms that export to many destinations and export many products, first-order stochastically dominates the distribution of non-exporters, and of firms that export to a few countries and a few products. Second, using regression analysis and controlling for the characteristics of firms, I find, both across and within firms, the number of layers in firms is positively correlated with export performance. Exporters, and firms that sell more products to more destinations, and at a greater value, have a greater number of layers. Overall, these results provide further evidence of the importance of studying organization in order to understand firms’ outcomes in the economy.
G. Spanos / Economics Letters 146 (2016) 130–134
131
Table 1 Descriptive statistics. 2000
2003
2006
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
Organization Exporters Non-exporters
2.98 2.09
0.88 0.95
2.96 2.04
0.88 0.94
2.95 2.01
0.87 0.93
Size Exporters Non-exporters
56.23 23.37
146.53 506.76
59.98 23.44
167.47 510.13
61.75 21.77
177.18 221.74
ln productivity Exporters Non-exporters
3.04 2.89
0.66 0.70
2.25 2.12
0.68 0.69
2.31 2.19
0.69 0.71
Notes: Sample descriptive statistics for the years 2000, 2003 and 2006.
This paper contributes to a growing literature on the organization of firms in international trade (Caliendo and Rossi-Hansberg, 2012, Caliendo et al., 2012, and Friedrich, 2015) and complements Caliendo et al. (2012), who find in French manufacturing sectors exporters tend to have a greater number of layers than nonexporters. This study provides additional empirical facts on firms’ organization, and examines how organization relates to export performance. Moreover the results from this study can be rationalized by introducing organization into existing models of trade, for example in the multi-country model of Chaney (2008) or the multi-product model of Bernard et al. (2011). 2. Data description The analysis is conducted over the periods 2000–2006 and uses data from three French sources. The first source is annual transactions level data provided by the French Customs Agency, and is used to obtain information on firms’ export performance. For each year and firm, the data report the value and quantity of exports by product (cn8 level) and destination. The second source, which is used to measure firms’ size and to observe their internal organization, is the Déclarations Annuelles des Données Sociales (DADS), provided by the French National Statistical Institute for Statistics and Economic Studies (INSEE). For each year, the DADS is an exhaustive cross-section of all workers who earn a positive wage in mainland France. In a given year, for every firm there is information on its industry, its employees, and their occupation (cs-occupational codes). The third source is balance sheet data from the Fichier Complet Unifié de Suse (FICUS), also provided by INSEE. For each year and firm, the data report firms’ sales, capital stock, and value added. Along with information from the DADS, I use the information from FICUS to estimate productivity using the method of Levinsohn and Petrin (2003). 2.1. Construction of layers Further, I use the first-digit of the occupational codes from the DADS to construct firms’ organization, as in Caliendo et al. (2015). With the occupational codes one can observe at most four distinct layers in firms. Layer 1, the lowest layer in firms, contains ordinary workers. Layer 2, contains their supervisors, while layer 3 contains senior managers, and layer 4 is composed of owners or CEOS.2
2 Caliendo et al. (2015) show that this approach of classifying workers into layers is consistent with models of hierarchical organization of Garicano (2000). Further, Caliendo et al. (2015) categorize firms by their number of layers of management. I depart from the nomenclature and categorize firms by the total number of layers in their organization.
Fig. 1. Kernel density of the value of exports.
In the analysis I retain only firms that operate in manufacturing sectors, that contain at least one employee in layer 1, and that only operate during consecutive years. Because not every layer is present in firms, there are four different types of organizations in the data: one-layer, two-layer, three-layer and four-layer organizations. The final sample contains 406,149 firm–year observations and consists of 89,495 unique firms. Table 1 reports summary statistics for the years 2000, 2003 and 2006. As is expected, across all years exporters are on average bigger, they are more productive, and on average they have an additional layer than non-exporting firms. In addition, for the year 2000, Figs. 1, 2(a) and (b) plot the kernel density of the value of exports, the number of destinations and the number of products, separately for firms with the same organization. These figures show that firms with a greater number of layers on average sell a greater value, to more destinations and they sell more products. 3. Stochastic dominance results To begin, I examine the distribution of firms operating with a given number of layers, the distribution of organizations. I first group firms by their export status. The top part of Table 2 presents summary statistics for the year 2000. In general exporters have a greater number of layers in their organization. Roughly 32.94% of non-exporting firms are one-layer firms, 32.96% are two-layer firms, 25.54% are three-layer firms, and 8.51% are four-layer firms, while the percent of exporters producing with one, two, three and four layers in their organization is 7.08, 18.81, 42.78, and 31.32, respectively. The top panel of Table 3 presents results from a nonparametric comparison of the distribution of organizations, the Mann–Whitney U test. The null hypothesis is that the distributions are equal, and is rejected at the one percent level. The third column indicates that the distribution of organizations of exporting firms first-order stochastically dominates the distribution of nonexporting firms. An exporter chosen at random is 74.1% more likely to have a greater number of layers than a random non-exporting firm. These findings indicate that there is a ranking of distributions, and the ranking is consistent with the model of Caliendo and RossiHansberg (2012).3 I now examine whether firms that export to many destinations (or many products) produce with a greater number of layers than
3 Although they do not use the Mann–Whitney U test to compare distributions, Caliendo et al. (2012) present similar findings between exporting and nonexporting firms.
132
G. Spanos / Economics Letters 146 (2016) 130–134
Table 2 Distribution of organizations. Number of firms
One Layers
Two Layers
Three Layers
Four Layers
All firms Exporters Non-exporters
58,056 14,971 43,085
0.2629 0.0708 0.3294
0.2931 0.1881 0.3296
0.2999 0.4278 0.2554
0.1439 0.3132 0.0851
Number of destinations Only 1 At least 2 1–10 10–20 20–40 40 and +
5,275 9,696 12,028 1,763 932 248
0.1344 0.0362 0.0858 0.0136 0.0032 0.0000
0.2815 0.1373 0.2214 0.0692 0.0321 0.0040
0.3905 0.4481 0.4235 0.4651 0.4130 0.4233
0.1935 0.3783 0.2690 0.4520 0.5515 0.5725
Number of products Only 1 At least 2 1–5 5–10 10–20 20–40 40 and +
4,483 10,488 9,534 2,803 1,655 729 250
0.1231 0.0484 0.0975 0.0328 0.0181 0.0109 0.0000
0.2647 0.1554 0.2413 0.1341 0.0719 0.0288 0.0000
0.3995 0.4399 0.4164 0.4527 0.4471 0.4389 0.4240
0.2125 0.3562 0.2447 0.3803 0.4628 0.5212 0.5760
Notes: Distribution of organizations for the year 2000. Table 3 Distribution of organizations. Group A
vs.
Group B
Null Hypothesis: distributions are equal
Probability Distribution Group A > Group B
Exporters
vs.
Non-exporters
0.0000
0.741
Number of destinations At least 2 10–20 20–40 40 and +
vs. vs. vs. vs.
Only 1 1–10 10–20 20–40
0.0000 0.0000 0.0000 0.3377
0.656 0.647 0.559 0.517
Number of products At least 2 5–10 10–20 20–40 40 and +
vs. vs. vs. vs. vs.
Only 1 1–5 5–10 10–20 20–40
0.0000 0.0000 0.0000 0.0005 0.0530
0.620 0.612 0.558 0.540 0.536
Notes: Mann–Whitney U test of the distribution of organizations by export status, by number of destinations and by the number of products for the year 2000. Table 4 Cross-section regression results.
Organization ln prod Industry FE R-Squared Observations
Exporting
ln number of destinations
ln number of products
ln density of exports
ln average value
0.146 (0.007)***
0.364 (0.019)***
0.301 (0.014)***
−0.228
0.422 (0.021)***
0.013 (0.004)***
0.132 (0.017)***
0.106 (0.017)***
−0.081 (0.010)***
0.068 (0.029)**
Yes 0.116 58,056
Yes 0.103 14,971
Yes 0.074 14,971
Yes 0.090 14,971
Yes 0.053 14,971
(0.010)***
Notes: OLS Cross-sectional regression results with industry fixed effects for the year 2000. Standard errors are robust. Industries correspond to the 4-digit Nace Rev 1.1 manufacturing industries. *** Significant at the 1% level. ** Significant at the 5% level. Table 5 Panel regressions results.
Organization ln prod Year FE Firm FE R-Squared Observations
Exporting
ln number of destinations
ln number of products
ln density of exports
ln average value
0.0067 (0.0009)***
0.0304 (0.0037)***
0.0336 (0.0044)***
−0.0226
0.0335 (0.0072)***
0.0147 (0.0015)***
0.0694 (0.0060)***
0.0855 (0.0068)***
−0.0483 (0.0047)***
0.2060 (0.0111)***
Yes Yes 0.001 406,149
Yes Yes 0.007 102,959
Yes Yes 0.005 102,959
Yes Yes 0.004 102,959
Yes Yes 0.009 102,959
(0.0031)***
Notes: OLS Panel regression results with firm and year fixed effects for the periods 2000–2006. Standard errors are robust. *** Significant at the 1% level. ** Significant at the 5% level.
G. Spanos / Economics Letters 146 (2016) 130–134
(a) Destinations.
133
(b) Products. Fig. 2. Kernel density of the number of destinations and products.
firms selling to fewer destinations (or fewer products).4 The middle panel of Table 2 groups exporters by the number of destinations. Firms that export to more destinations have more layers. This is also confirmed in the middle panel of Table 3, which presents results from the Mann–Whitney U test. In nearly every case the null hypothesis is rejected, and the test indicates the distribution of organizations of firms exporting to more destinations first-order stochastically dominates the distribution of firms selling to fewer markets. For example, a firm exporting to at least 10 and at most 20 destinations chosen at random is 65.7% more likely to have a greater number of layers than a random firm exporting to at most 10 destinations. Finally, the bottom panels of Tables 2 and 3 group exporters by the number of products sold. The results are similar. Firms that export more products have more layers.
tion–product. Unsurprisingly, density decreases with firm productivity. Although productive firms sell a greater number of products to more destinations, they do not sell every product to every destination. Table 4 reports firms with a greater number of layers are more likely to export and besides density, the remaining margins are also increasing with the number of layers in firms. For example, an additional layer is associated with a 14.6% increase in the probability of exporting and a 42.2% increase in the average exports per destination and product. Hence firms’ internal organization is related to their export performance. 4.2. Panel regression results I now turn to the panel dimension of the data to examine how, within firms, changes in organization relate to firms’ export performance. I estimate the equation:
4. Regressions results 4.1. Cross-section regressions results
Yjt = β0 + β1 ORGjt + β2 ln φjt + βjt Xjt + ϵjt ,
I now turn to regression analysis. I decompose the value of exports into the following margins: the number of destinations served, the number of products sold, firms’ density, and the average value per destination–product, as in Bernard et al. (2011). I first use the cross-sectional dimension of the data to examine how firms’ export performance relates to their organization, and estimate the following equation:
where Yjt , ORGjt and φjt are as previously defined, and Xjt contains firm and year fixed effects. Results are reported in Table 5. The point estimates of productivity are again as expected. Furthermore the results with respect to organization lead to the same conclusion. A withinfirm increase in the number of layers is positively associated with export performance. For example, adding a layer in a firm is associated with a 0.67% in the probability of exporting and a 3.36% increase in the number of products sold.
Yj = β0 + β1 ORGj + β2 ln φj + βj Xj + ϵj ,
(1)
where the dependent variable Yj is either an indicator for the export status of firm j, or it is equal to the margins described above. The variable ORGj measures the total number of layers in firm j, φj is the estimated productivity of firm j, and Xj controls for the major industry of the firm.5 In all regressions standard errors are robust. Table 4 reports results for the year 2000. The point estimates with respect to firm productivity are as expected. Productive firms are more likely to export, they export more products to more countries, and at a greater value per destina-
4 This result follows from the fact that firms selling to many destinations on average produce a greater scale, and so they would tend to have more layers. 5 Here and in subsequent regressions, I do not control for firm size. The reason is the following. In one-factor heterogeneous firm models, if one controls for productivity and organization, there is no remaining variation in the data.
(2)
5. Conclusion This study has examined how firms’ internal organization relates to their export performance. I find there is a systematic ordering of the distribution of organizations across the different margins of trade, and that both across and within firms, the number of layers in firms is positively correlated with firms’ export performance. These findings further confirm the importance of studying organization to understand the outcomes of firms and workers in the economy. Acknowledgments I thank Linas Tarasonis, Brian Tavares, Marc Sangnier, PierrePhilippe Combes, and Tanguy van Ypersele.
134
G. Spanos / Economics Letters 146 (2016) 130–134
References Bernard, Andrew B., Redding, Stephen J., Schott, Peter K., 2011. Multi-product firms and trade liberalization. Quart. J. Econ. 126 (3), 1271–1318. Caliendo, Lorenzo, Monte, Ferdinando, Rossi-Hansberg, Esteban, 2012. Anatomy of French production hierarchies. NBER Working Paper No. 18259. Caliendo, Lorenzo, Monte, Ferdinando, Rossi-Hansberg, Esteban, 2015. Anatomy of French production hierarchies. J. Polit. Econ. 123 (4), 809–852.
Caliendo, Lorenzo, Rossi-Hansberg, Esteban, 2012. The impact of trade on organization and productivity. Quart. J. Econ. 127 (3), 1393–1467. Chaney, Thomas., 2008. Distorted gravity: The intensive and extensive margins of international trade. Amer. Econ. Rev. 98 (4), 1707–1721. Friedrich, Benjamin., 2015. Trade shocks, firm hierarchies and wage inequality. Working Paper. Garicano, Luis., 2000. Hierarchies and the organization of knowledge in production. J. Polit. Econ. 108 (5), 874–904. Levinsohn, James, Petrin, Kamil, 2003. Estimating production functions using inputs to control for unobservables. Rev. Econom. Stud. 70 (2), 317–342.
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Organization and export performance Grigorios Spanos Aix-Marseille University (Aix-Marseille School of Economics), CNRS & EHESS, France
highlights • • • • •
Present new facts on the internal organization of firms and export performance. In the aggregate there is an ordering of the distribution of organizations. The number of layers in firms is positively correlated with export performance. Firms with a greater number of layers export more products to more destinations. Introducing organization into heterogeneous firm models can explain these facts.
article
info
Article history: Received 13 May 2016 Received in revised form 24 July 2016 Accepted 26 July 2016 Available online 30 July 2016
abstract This paper presents new facts on firms’ internal organization and their export performance. I find an ordering of the distribution of organizations and, both across and within firms, firms’ number of layers is positively correlated with their export performance. © 2016 Elsevier B.V. All rights reserved.
JEL classification: F14 L23 Keywords: Trade Firms Margins Exports Organization
1. Introduction Recent research in international trade has shown that firms’ internal organization is important for understanding the outcomes of firms and workers in the economy. In particular, Caliendo and Rossi-Hansberg (2012) show that exporters are on average more productive because they have a greater number of layers. Despite its relevance, empirical evidence on the organization of firms remains limited in international trade. The existing literature has mainly focused on firm size and productivity to examine firms’ export performance, characteristics that are related to organization.1
E-mail address: [email protected]. 1 I broadly define export performance to include the decision to export, as well the number destination markets served, the number of products sold, and the value of exports. http://dx.doi.org/10.1016/j.econlet.2016.07.031 0165-1765/© 2016 Elsevier B.V. All rights reserved.
This paper fills this gap and examines how organization, the number of layers in firms, relates to French manufacturing firms’ export performance over the years 2000–2006. I first examine the distribution of organizations, the percent of firms producing with a given number of layers. I find there is a ranking of distributions. More precisely, the distribution of organizations of exporters, of firms that export to many destinations and export many products, first-order stochastically dominates the distribution of non-exporters, and of firms that export to a few countries and a few products. Second, using regression analysis and controlling for the characteristics of firms, I find, both across and within firms, the number of layers in firms is positively correlated with export performance. Exporters, and firms that sell more products to more destinations, and at a greater value, have a greater number of layers. Overall, these results provide further evidence of the importance of studying organization in order to understand firms’ outcomes in the economy.
G. Spanos / Economics Letters 146 (2016) 130–134
131
Table 1 Descriptive statistics. 2000
2003
2006
Mean
St. Dev.
Mean
St. Dev.
Mean
St. Dev.
Organization Exporters Non-exporters
2.98 2.09
0.88 0.95
2.96 2.04
0.88 0.94
2.95 2.01
0.87 0.93
Size Exporters Non-exporters
56.23 23.37
146.53 506.76
59.98 23.44
167.47 510.13
61.75 21.77
177.18 221.74
ln productivity Exporters Non-exporters
3.04 2.89
0.66 0.70
2.25 2.12
0.68 0.69
2.31 2.19
0.69 0.71
Notes: Sample descriptive statistics for the years 2000, 2003 and 2006.
This paper contributes to a growing literature on the organization of firms in international trade (Caliendo and Rossi-Hansberg, 2012, Caliendo et al., 2012, and Friedrich, 2015) and complements Caliendo et al. (2012), who find in French manufacturing sectors exporters tend to have a greater number of layers than nonexporters. This study provides additional empirical facts on firms’ organization, and examines how organization relates to export performance. Moreover the results from this study can be rationalized by introducing organization into existing models of trade, for example in the multi-country model of Chaney (2008) or the multi-product model of Bernard et al. (2011). 2. Data description The analysis is conducted over the periods 2000–2006 and uses data from three French sources. The first source is annual transactions level data provided by the French Customs Agency, and is used to obtain information on firms’ export performance. For each year and firm, the data report the value and quantity of exports by product (cn8 level) and destination. The second source, which is used to measure firms’ size and to observe their internal organization, is the Déclarations Annuelles des Données Sociales (DADS), provided by the French National Statistical Institute for Statistics and Economic Studies (INSEE). For each year, the DADS is an exhaustive cross-section of all workers who earn a positive wage in mainland France. In a given year, for every firm there is information on its industry, its employees, and their occupation (cs-occupational codes). The third source is balance sheet data from the Fichier Complet Unifié de Suse (FICUS), also provided by INSEE. For each year and firm, the data report firms’ sales, capital stock, and value added. Along with information from the DADS, I use the information from FICUS to estimate productivity using the method of Levinsohn and Petrin (2003). 2.1. Construction of layers Further, I use the first-digit of the occupational codes from the DADS to construct firms’ organization, as in Caliendo et al. (2015). With the occupational codes one can observe at most four distinct layers in firms. Layer 1, the lowest layer in firms, contains ordinary workers. Layer 2, contains their supervisors, while layer 3 contains senior managers, and layer 4 is composed of owners or CEOS.2
2 Caliendo et al. (2015) show that this approach of classifying workers into layers is consistent with models of hierarchical organization of Garicano (2000). Further, Caliendo et al. (2015) categorize firms by their number of layers of management. I depart from the nomenclature and categorize firms by the total number of layers in their organization.
Fig. 1. Kernel density of the value of exports.
In the analysis I retain only firms that operate in manufacturing sectors, that contain at least one employee in layer 1, and that only operate during consecutive years. Because not every layer is present in firms, there are four different types of organizations in the data: one-layer, two-layer, three-layer and four-layer organizations. The final sample contains 406,149 firm–year observations and consists of 89,495 unique firms. Table 1 reports summary statistics for the years 2000, 2003 and 2006. As is expected, across all years exporters are on average bigger, they are more productive, and on average they have an additional layer than non-exporting firms. In addition, for the year 2000, Figs. 1, 2(a) and (b) plot the kernel density of the value of exports, the number of destinations and the number of products, separately for firms with the same organization. These figures show that firms with a greater number of layers on average sell a greater value, to more destinations and they sell more products. 3. Stochastic dominance results To begin, I examine the distribution of firms operating with a given number of layers, the distribution of organizations. I first group firms by their export status. The top part of Table 2 presents summary statistics for the year 2000. In general exporters have a greater number of layers in their organization. Roughly 32.94% of non-exporting firms are one-layer firms, 32.96% are two-layer firms, 25.54% are three-layer firms, and 8.51% are four-layer firms, while the percent of exporters producing with one, two, three and four layers in their organization is 7.08, 18.81, 42.78, and 31.32, respectively. The top panel of Table 3 presents results from a nonparametric comparison of the distribution of organizations, the Mann–Whitney U test. The null hypothesis is that the distributions are equal, and is rejected at the one percent level. The third column indicates that the distribution of organizations of exporting firms first-order stochastically dominates the distribution of nonexporting firms. An exporter chosen at random is 74.1% more likely to have a greater number of layers than a random non-exporting firm. These findings indicate that there is a ranking of distributions, and the ranking is consistent with the model of Caliendo and RossiHansberg (2012).3 I now examine whether firms that export to many destinations (or many products) produce with a greater number of layers than
3 Although they do not use the Mann–Whitney U test to compare distributions, Caliendo et al. (2012) present similar findings between exporting and nonexporting firms.
132
G. Spanos / Economics Letters 146 (2016) 130–134
Table 2 Distribution of organizations. Number of firms
One Layers
Two Layers
Three Layers
Four Layers
All firms Exporters Non-exporters
58,056 14,971 43,085
0.2629 0.0708 0.3294
0.2931 0.1881 0.3296
0.2999 0.4278 0.2554
0.1439 0.3132 0.0851
Number of destinations Only 1 At least 2 1–10 10–20 20–40 40 and +
5,275 9,696 12,028 1,763 932 248
0.1344 0.0362 0.0858 0.0136 0.0032 0.0000
0.2815 0.1373 0.2214 0.0692 0.0321 0.0040
0.3905 0.4481 0.4235 0.4651 0.4130 0.4233
0.1935 0.3783 0.2690 0.4520 0.5515 0.5725
Number of products Only 1 At least 2 1–5 5–10 10–20 20–40 40 and +
4,483 10,488 9,534 2,803 1,655 729 250
0.1231 0.0484 0.0975 0.0328 0.0181 0.0109 0.0000
0.2647 0.1554 0.2413 0.1341 0.0719 0.0288 0.0000
0.3995 0.4399 0.4164 0.4527 0.4471 0.4389 0.4240
0.2125 0.3562 0.2447 0.3803 0.4628 0.5212 0.5760
Notes: Distribution of organizations for the year 2000. Table 3 Distribution of organizations. Group A
vs.
Group B
Null Hypothesis: distributions are equal
Probability Distribution Group A > Group B
Exporters
vs.
Non-exporters
0.0000
0.741
Number of destinations At least 2 10–20 20–40 40 and +
vs. vs. vs. vs.
Only 1 1–10 10–20 20–40
0.0000 0.0000 0.0000 0.3377
0.656 0.647 0.559 0.517
Number of products At least 2 5–10 10–20 20–40 40 and +
vs. vs. vs. vs. vs.
Only 1 1–5 5–10 10–20 20–40
0.0000 0.0000 0.0000 0.0005 0.0530
0.620 0.612 0.558 0.540 0.536
Notes: Mann–Whitney U test of the distribution of organizations by export status, by number of destinations and by the number of products for the year 2000. Table 4 Cross-section regression results.
Organization ln prod Industry FE R-Squared Observations
Exporting
ln number of destinations
ln number of products
ln density of exports
ln average value
0.146 (0.007)***
0.364 (0.019)***
0.301 (0.014)***
−0.228
0.422 (0.021)***
0.013 (0.004)***
0.132 (0.017)***
0.106 (0.017)***
−0.081 (0.010)***
0.068 (0.029)**
Yes 0.116 58,056
Yes 0.103 14,971
Yes 0.074 14,971
Yes 0.090 14,971
Yes 0.053 14,971
(0.010)***
Notes: OLS Cross-sectional regression results with industry fixed effects for the year 2000. Standard errors are robust. Industries correspond to the 4-digit Nace Rev 1.1 manufacturing industries. *** Significant at the 1% level. ** Significant at the 5% level. Table 5 Panel regressions results.
Organization ln prod Year FE Firm FE R-Squared Observations
Exporting
ln number of destinations
ln number of products
ln density of exports
ln average value
0.0067 (0.0009)***
0.0304 (0.0037)***
0.0336 (0.0044)***
−0.0226
0.0335 (0.0072)***
0.0147 (0.0015)***
0.0694 (0.0060)***
0.0855 (0.0068)***
−0.0483 (0.0047)***
0.2060 (0.0111)***
Yes Yes 0.001 406,149
Yes Yes 0.007 102,959
Yes Yes 0.005 102,959
Yes Yes 0.004 102,959
Yes Yes 0.009 102,959
(0.0031)***
Notes: OLS Panel regression results with firm and year fixed effects for the periods 2000–2006. Standard errors are robust. *** Significant at the 1% level. ** Significant at the 5% level.
G. Spanos / Economics Letters 146 (2016) 130–134
(a) Destinations.
133
(b) Products. Fig. 2. Kernel density of the number of destinations and products.
firms selling to fewer destinations (or fewer products).4 The middle panel of Table 2 groups exporters by the number of destinations. Firms that export to more destinations have more layers. This is also confirmed in the middle panel of Table 3, which presents results from the Mann–Whitney U test. In nearly every case the null hypothesis is rejected, and the test indicates the distribution of organizations of firms exporting to more destinations first-order stochastically dominates the distribution of firms selling to fewer markets. For example, a firm exporting to at least 10 and at most 20 destinations chosen at random is 65.7% more likely to have a greater number of layers than a random firm exporting to at most 10 destinations. Finally, the bottom panels of Tables 2 and 3 group exporters by the number of products sold. The results are similar. Firms that export more products have more layers.
tion–product. Unsurprisingly, density decreases with firm productivity. Although productive firms sell a greater number of products to more destinations, they do not sell every product to every destination. Table 4 reports firms with a greater number of layers are more likely to export and besides density, the remaining margins are also increasing with the number of layers in firms. For example, an additional layer is associated with a 14.6% increase in the probability of exporting and a 42.2% increase in the average exports per destination and product. Hence firms’ internal organization is related to their export performance. 4.2. Panel regression results I now turn to the panel dimension of the data to examine how, within firms, changes in organization relate to firms’ export performance. I estimate the equation:
4. Regressions results 4.1. Cross-section regressions results
Yjt = β0 + β1 ORGjt + β2 ln φjt + βjt Xjt + ϵjt ,
I now turn to regression analysis. I decompose the value of exports into the following margins: the number of destinations served, the number of products sold, firms’ density, and the average value per destination–product, as in Bernard et al. (2011). I first use the cross-sectional dimension of the data to examine how firms’ export performance relates to their organization, and estimate the following equation:
where Yjt , ORGjt and φjt are as previously defined, and Xjt contains firm and year fixed effects. Results are reported in Table 5. The point estimates of productivity are again as expected. Furthermore the results with respect to organization lead to the same conclusion. A withinfirm increase in the number of layers is positively associated with export performance. For example, adding a layer in a firm is associated with a 0.67% in the probability of exporting and a 3.36% increase in the number of products sold.
Yj = β0 + β1 ORGj + β2 ln φj + βj Xj + ϵj ,
(1)
where the dependent variable Yj is either an indicator for the export status of firm j, or it is equal to the margins described above. The variable ORGj measures the total number of layers in firm j, φj is the estimated productivity of firm j, and Xj controls for the major industry of the firm.5 In all regressions standard errors are robust. Table 4 reports results for the year 2000. The point estimates with respect to firm productivity are as expected. Productive firms are more likely to export, they export more products to more countries, and at a greater value per destina-
4 This result follows from the fact that firms selling to many destinations on average produce a greater scale, and so they would tend to have more layers. 5 Here and in subsequent regressions, I do not control for firm size. The reason is the following. In one-factor heterogeneous firm models, if one controls for productivity and organization, there is no remaining variation in the data.
(2)
5. Conclusion This study has examined how firms’ internal organization relates to their export performance. I find there is a systematic ordering of the distribution of organizations across the different margins of trade, and that both across and within firms, the number of layers in firms is positively correlated with firms’ export performance. These findings further confirm the importance of studying organization to understand the outcomes of firms and workers in the economy. Acknowledgments I thank Linas Tarasonis, Brian Tavares, Marc Sangnier, PierrePhilippe Combes, and Tanguy van Ypersele.
134
G. Spanos / Economics Letters 146 (2016) 130–134
References Bernard, Andrew B., Redding, Stephen J., Schott, Peter K., 2011. Multi-product firms and trade liberalization. Quart. J. Econ. 126 (3), 1271–1318. Caliendo, Lorenzo, Monte, Ferdinando, Rossi-Hansberg, Esteban, 2012. Anatomy of French production hierarchies. NBER Working Paper No. 18259. Caliendo, Lorenzo, Monte, Ferdinando, Rossi-Hansberg, Esteban, 2015. Anatomy of French production hierarchies. J. Polit. Econ. 123 (4), 809–852.
Caliendo, Lorenzo, Rossi-Hansberg, Esteban, 2012. The impact of trade on organization and productivity. Quart. J. Econ. 127 (3), 1393–1467. Chaney, Thomas., 2008. Distorted gravity: The intensive and extensive margins of international trade. Amer. Econ. Rev. 98 (4), 1707–1721. Friedrich, Benjamin., 2015. Trade shocks, firm hierarchies and wage inequality. Working Paper. Garicano, Luis., 2000. Hierarchies and the organization of knowledge in production. J. Polit. Econ. 108 (5), 874–904. Levinsohn, James, Petrin, Kamil, 2003. Estimating production functions using inputs to control for unobservables. Rev. Econom. Stud. 70 (2), 317–342.