Abstract
This paper directly compares the size and performance of direct, indirect and non-exporters using firm-level data from 69 countries. The authors find that direct exporters are on average larger and outperform both indirect exporters and non-exporters. Furthermore, they analyse the relationship between source-specific variables, particularly trade costs, and the differences across exporting statuses. The authors find that higher trade barriers significantly lower differences between exporters and non-exporters but raise those between direct and indirect exporters. Similarly, domestic market size and skill endowments affect the differences across firm types.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Findings indicate that wholesalers account for 20 % of total exports in France (Crozet et al. 2013), 22 % of total exports in China (Ahn et al. 2011), 11 % in Italy (Bernard et al. 2011) and 8 % in the US (Bernard et al. 2010). In addition, there is a small literature on indirect importers (e.g. Blum et al. 2010).
We specifically avoid referring to sales/employment as productivity because this measure does not adequately deal with the role of other inputs into the production process. See Pavcnik (2002) for discussion using this measure as productivity. Gorodnichenko and Schnitzer (2013) calculate a measure of productivity for Eastern European countries from the World Bank’s data, however, as they acknowledge, missing information (particularly for capital) severely reduces the number of observations.
Addressing a different question with Swedish data, Akerman (2010) considers the use of intermediaries as it compares to a firm’s product scope, finding that firms which use intermediaries export more products.
A detailed theoretic treatment of this discussion is available on request.
As our data do not provide destination information, our focus on exporting country characteristics is also practical. In unreported, but available results, we also considered exporter democracy which was not often significant and, when it was, suggested smaller differences in size and sales between exporters and non-exporters.
In Namini (2014), where the second factor is endogenously determined, an increase in the exogenous factor results in these two effects exactly offsetting each other, yielding no change in average productivities.
In addition to manufacturing firms, defined as those in industries 15–37 using the ISIC 3.1 Rev. Classification, the surveys include data on services and retail/wholesale firms. Such firms do report indirect exports, albeit at a lower rate, because their industry classification is according to their primary activity and does not preclude them from also producing intangibles. When these firms are included, the results are qualitatively similar. These results are available on request. The specific dataset is entitled “Standardized data” and was downloaded, Dec. 2011. http://www.enterprisesurveys.org/.
There are a few exceptions to this, mainly in Latin America, where surveys have taken place twice. As it is not clear what proportion of firms have been interviewed twice within these countries, we use only the largest survey round for each country. However, this does not cause any major changes to the key results.
Specifically, it uses strata on firm size (with three categories: <20 employees, 20–99 employees, and 100+ employees).
Sales values are reported in dollars only for Ecuador. The Consumer price index for Chile (2006) came from the IMF, Economic Outlook Database Sep. 2011 (Downloaded Dec. 2011).
See the “Appendix” for a detailed description of the process of data cleaning.
Among the group of exporters that make use of both modes, 28 % of total sales are, on average, exported via direct channels, compared to an average of 22 % via indirect channels. This is again part of our rationale for classifying these firms as direct exporters. As an alternative, we defined direct exporters are firms that exclusively export directly. Despite this, the results were relatively unchanged. Finally, we examined estimates where we used three categories: exporters, indirect only, and direct only. These indicated that direct exporters are larger than mixed which are larger than indirect exporters. However, there was no significant performance difference between indirect and mixed exporters. These results are available on request.
The measures take into account all procedures required to get the goods from their warehouse into containers and transported to their port of exit, but exclude ocean transport. See the World Bank (2012a) for details.
Bernard et al. (2011) use the principal component of the equivalent measures for importing, as they have data on the destination of exports across a sample of Italian exporters, enabling them to look at destination-specific fixed costs of exporting.
See http://www.un.org/en/development/desa/policy/cdp/ldc/ldc_definitions.shtml. Note that we use the negative of this so that a higher value implies a more remote country and therefore higher trade barriers. Also, recognize that remoteness may affect other factors beyond trade, such as technology flows, that affect firms.
When using average education of those 25 and older, the results were qualitatively identical.
This list is driven both by data availability and the discussion in the existing literature. Although information such as R&D is available for some countries or firms, missing data prohibits their use in our analysis.
In unreported results, we also applied propensity score matching methods. This confirmed the significant differences between non-exporters and either group of exporters. However, there was not a significant difference in performance between direct and indirect exporters, perhaps due to the small number (and therefore imprecise matches) of indirect exporters.
One reason to not do so is that, particularly in a logged specification, it raises endogeneity concerns. Also note that because of the logs, to obtain the results for e.g. sales with employment as a control, one need only rearrange the results from the performance (where the dependent variable is ln(sales) − ln(employment)) specification.
See http://www.enterprisesurveys.org/methodology for discussion on the survey stratification.
Mengistae and Patillo (2004) suggest that potential learning effects may be greater for direct than indirect exporters, as the former group are directly in contact with the purchaser and therefore likely to receive more and better information, which in turn may translate into lower costs or product improvements.
Thus, the coefficient for this in column one is for the difference between the average exporter, regardless of mode, and the average non-exporter.
The percentage differences are calculated as: 100*(exp(β) − 1).
McCann (2013) uses data collected over three waves of surveys, while we just have a single survey round per country and restrict ourselves to manufacturing firms.
The increase in the share of direct exporters from the second to third quartiles is driven by Argentina.
In unreported results, we plot the estimated country fixed effects against the trade cost index. As expected, higher trade costs are associated with lower performance for all firm types. These are available on request.
To conserve space, we do not report the results of the control variables, the significance and magnitude of which are comparable to Table 7. The full results are available on request.
With the exception of the performance results including ln(emp), the net effect of an increase in GDP or population for direct exporters is still positive and significant. In column (4), the net impact is negative and significant, meaning a decrease in the differences between direct and non-exporters.
Note that since both GDP and Population are in logs, we exclude the log of GDP per capita.
Industry code is determined by the 2-digit sector code as oppose to the 4-digit code of the firm’s main product. The authors of the survey recommend using the latter, as this is more precise measure. However, a relatively large share of this variable is missing, we use the sector codes.
If the variables are standardized the co-variance matrix is used instead (Vyas and Kumaranayake 2006).
This is computed using the factor command with the pcf option in Stata, which makes use of the correlation matrix, rendering prior standardization of the variables unnecessary.
References
Abel-Koch, J. (2013). Who uses intermediaries in international trade? Evidence from firm-level survey data. The World Economy, 36(8), 1041–1064.
Ahn, J., Khandelwal, A. K., & Wei, S. (2011). The role of intermediaries in facilitating trade. Journal of International Economics, 84(1), 73–85.
Akerman, A. (2010). Wholesalers and economies of scope in international trade. Research papers in Economics, Department of Economics, Stockholm University 2010:1.
Bai, X., Krishna, K., & Ma, H. (2014). How you export matters: Export mode, learning, and productivity in China. Mimeo.
Barro, R., & Lee, J. (2013). A new data set of educational attainment in the world, 1950–2010. Journal of Development Economics, 104(2), 184–198.
Bauer, C., Davies, R. B., & Haufler, A. (2014). Economic integration and the optimal corporate tax structure with heterogeneous firms. Journal of Public Economics, 110(C), 42–56.
Bernard, A.B., Grazzi, M., & Tomasi, C. (2011). Intermediaries in international trade: Direct versus indirect modes of export (Working Paper). University of Trento, Department of Economics 16/2010.
Bernard, A. B., Jensen, J. B., Redding, S. J., & Schott, P. K. (2010). Wholesalers and retailers in US Trade. American Economic Review, 100(2), 408–413.
Biglaiser, G. (1993). Middlemen as experts. RAND Journal of Economics, 24(2), 212–223.
Blum, B. S., Claro, S., & Horstman, I. (2010). Facts and figures on intermediated trade. American Economic Review: Papers and Proceedings, 100(2), 419–423.
CEPII (2013). Geography databases. http://www.cepii.fr/CEPII/en/bdd_modele/bdd.asp. Accessed August 2013.
Crozet, M., Lalanne, G., & Poncet, S. (2013). Wholesalers in international trade. European Economic Review, 58(1), 1–17.
De Loecker, J. (2013). Detecting learning by exporting. American Economic Journal: Microeconomics, 5(3), 1–21.
Felbermayr, G. J., & Jung, B. (2011). Trade intermediation and the organization of exporters. Review of International Economics, 19(4), 634–648.
Gorodnichenko, Y., & Schnitzer, M. (2013). Financial constraints and innovation: Why poor countries don’t catch up. Journal of the European Economics Association, 11(5), 1115–1152.
McCann, F. (2013). Indirect exporters. Journal of Industry, Competition, and Trade, 13(4), 519–535.
Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6), 1695–1725.
Melitz, M. J., & Ottaviano, G. J. (2008). Market size, trade, and productivity. Review of Economic Studies, 75(1), 295–316.
Mengistae, T., & Patillo, C. (2004). Export orientation and productivity in Sub-Saharan Africa. IMF Staff Papers, 51(2), 327–353.
Namini, J. E. (2014). The short and long-run impact of globalization if firms differ in factor input ratios. Journal of Economic Dynamics and Control, 38(1), 37–64.
Pavcnik, N. (2002). Trade liberalization, exit and productivity improvements: evidence from Chilean plants. The Review of Economic Studies, 69(1), 245–276.
Petropoulou, D. (2008). Information costs, networks and intermediation in international trade (CEP Discussion paper No. 848). Centre for Economic Performance, London School of Economics and Political Science.
Rauch, J. E., & Watson, J. (2004). Network intermediaries in international trade. Journal of Economics & Management Strategy, 13(1), 69–93.
Tang, H., & Zhang, Y. (2012). Quality differentiation and trade intermediation (Development Working Papers 340). Centro Studi Luca d’Agliano, University of Milano, revised 13 Nov. 2012.
Vyas, S., & Kumaranayake, L. (2006). How to do (or not to do)…Constructing socio-economic status indices: how to use principal components analysis. Health Policy and Planning, 21(6), 459–468.
Wagner, J. (2007). Exports and productivity: A survey of the evidence from firm-level data. The World Economy, 30(1), 60–82.
Wagner, J. (2012). International trade and firm performance: A survey of empirical studies since 2006. Review of World Economics/Weltwirtschaftliches Archiv, 148(2), 235–268.
World Bank (2012a). Doing business. http://www.doingbusiness.org/.
World Bank (2012b). Enterprise surveys. http://www.enterprisesurveys.org/.
World Bank (2012c). World development indicators. http://data.worldbank.org/indicator.
Acknowledgments
The authors thank three anonymous referees, Fergal McCann and the UCD trade group for helpful comments and suggestions. The authors acknowledge that this paper is produced as part of the project “Globalization, Investment and Services Trade (GIST) Marie Curie Initial Training Network (ITN)” funded by the European Commission under its Seventh Framework Programme—Contract No. FP7-PEOPLE-ITN-2008-211429.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
1.1 Data cleaning
The raw dataset was comprised of 26,975 manufacturing firms covering 108 countries. Out of these we drop all observations for Ghana and Venezuela due to unreliable sales values. In addition all observations for Micronesia are dropped, as no consumer price index is available for this country. Following each survey, the interviewer is asked to comment on his/hers perception of the truthfulness of the answers provided. Using this information, we furthermore drop all observations deemed to have answered the questions untruthfully or provided arbitrary and unreliable figures pertaining to e.g. sales values etc. In total this affects 10.5 % of the raw data. After deflating all sales values, reported in local currencies, by the annual country-specific consumer price index and converting all monetary values to US dollars, we impose a lower bound on the sales values included and drop all observations with annual sales less than 100 dollars. Likewise, we drop any observations which report sales values in excess of the country’s GDP. In total these criteria leads to a loss of only 21 observations. For any country in which surveys have taken place twice during the time period, we keep only the largest survey round, as no information is available on the share of firms interviewed more than once.
In order to avoid the analysis being driven by outliers, we furthermore drop any observations with a labour productivity lower than the 25th percentile minus three times the interquartile range (75th percentile − 25th percentile) or higher than the 75th percentile plus three times the interquartile range, for a given country. Defining outliers according to this methodology leads to a loss of a further 1,186 observations. Finally, of the remaining observations we lose an additional 33 % due to missing firm-level information. In total, this left with 16,756 observations covering 69 countries.Footnote 33 We note, however, that when omitting firm-level controls to include additional observations, we found a positive and significant coefficient for both the exporting and direct exporting dummies.
1.2 Construction of trade costs
As Vyas and Kumaranayake (2006) discuss, principal component analysis reduces a number of variables into a smaller set of orthogonal components, where each represents a linear combination of the original variables, weighted by the eigenvectors of the correlation matrix.Footnote 34 The maximum number of components or factors computed is then determined by the number of variables used, with each component accounting for a smaller share of the total variation of the original variables. In panel B of Table 12, we list the full set of factors computed. Column (1) shows the variance extracted by each of these, expressed as the proportion of total variance in the data in column (2).Footnote 35 As we can see, the factor 1 accounts for 65 % of the total variance. As this factor is the only one that accounts for at least as much of the variation as one of the original variables, we follow the Kaiser criterion and discard factors 2 and 3. To see how the remaining factor relates to each of the underlying variables, the factor loadings are displayed in column (1) in panel C. These indicate the weights and correlations between factor 1 and each of the individual variables. As expected, these are all relatively large. Finally, column (2) in panel C displays the part of the variance that is unique to each of the individual variables. With the exception of Documents to export, these are all relatively low, meaning that each variable contributes considerably to the computation of the principal component. Table 13 describes the data.
About this article
Cite this article
Davies, R.B., Jeppesen, T. Export mode, firm heterogeneity, and source country characteristics. Rev World Econ 151, 169–195 (2015). https://doi.org/10.1007/s10290-014-0207-4
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10290-014-0207-4