Firm export dynamics and the geography of trade
Introduction
Recent years have seen an important development in empirical research in international economics toward the analysis of firm-level data to gain a better understanding of the processes underlying international trade, with classic contributions including the work of Roberts and Tybout (1997) and Bernard and Jensen, 1995, Bernard and Jensen, 2004). This literature has provided substantial insights into the characteristics of exporting firms. For instance, exporting firms tend to more productive than non-exporting firms and the exporting process is very persistent, so that firms rarely change their status as either exporters or non-exporters.
This evidence has suggested a process in which individual firms face substantial barriers to engaging in trading activity, so that only the most productive can afford to do so. The new empirical findings in turn motivated the development of a new range of theoretical trade models to explain these facts such as Melitz (2003) and Eaton, Kortum and Kramarz (2008). Firm heterogeneity in productivity is the foundation of this new generation of trade model and, combined with different types of trade costs, results in productivity thresholds that allow only the best firms to become exporters.
An important limitation of the empirical evidence to date is that most of the existing datasets report information on whether a firm exports and, if so, how much it exports. However, information on the destination of exports has generally been missing from these studies. One important exception to this rule has been the work of Eaton, Kortum and Kramarz (2004) on the export destinations of French exporters. Using a cross-section of data from 1986, they find great heterogeneity in firms' involvement in exporting. Most firms sell only in the domestic market; amongst exporters, the modal firm exports to only a single foreign market; and a small fraction of firms export to a large number of markets.
This paper adds to this literature by making use of a unique survey data of Irish firms. Like the Eaton–Kortum–Kramarz data, this survey contains detailed information on firm exports and destinations. However, unlike their data, this survey is a panel dataset that follows firms over a period of five years. This dataset allows us to better understand the dynamic processes underlying the patterns of firm exporting. We examine how the facts which emerge from the data fit with the predictions of existing models.
In particular, we begin by presenting a simple model based on Melitz (2003) with heterogeneous firms differing in their productivity endowment and having to face both fixed and variable trade costs to export. The main predictions of the model are discussed and the paper then proceeds to examine to what extent these predictions appear to hold in the data. Amongst our findings are that firm productivity differences are a factor in explaining a firm's range of export markets but the model's prediction that firms should enter export markets according to a common hierarchy could only be weakly upheld by the data. Firm involvement in individual export markets is found to be much more dynamic than their overall status as exporters. Entry and exit to markets is shown to be a quantifiably important component of overall export flows, with this factor becoming more important for less popular markets. The paper also shows how the patterns of entry and exit into export markets combine to determine the overall firm-level distribution of number of markets entered.
The paper proceeds as follows: Section 2 outlines the model and discusses five main predictions that it leads to. Section 3 explains the data source in more detail. 4 Exporting and productivity, 8 Changes in market coverage then each discuss specific predictions of the model. Section 4 examines the link between exporting activity and firm productivity. Section 5 looks for evidence of a hierarchy in the markets firms export to. Section 6 examines at export growth at the firm level and how important changes in market coverage are for the firm. Section 7 then looks across markets, decomposing export growth into contributions by entrant, exitor and continuing firms. Section 8 examines the dynamics of changing market coverage, using a matrix of transition probabilities to generate predictions about the long-run distribution of export market coverage. The hypothetical distribution from this Markov-chain process turns out to be remarkably consistent with observed distribution of firms across markets. Section 9 concludes.
Section snippets
A model with firm heterogeneity and trade costs
This section presents a simple model of firm export participation which incorporates the key features of firm heterogeneity and both fixed and variable trade costs. This is followed by a discussion of the main implications of the model for firm exporting behaviour. The remainder of the paper will examine how well these implications correspond to the observed patterns in the data.
Description of data
The data used in this paper come from a survey of Irish firms undertaken by Enterprise Ireland and Forfás, which is the Irish national policy advisory board for enterprise, trade and technology and operates under the Government Department of Enterprise, Trade and Employment. The focus of survey is on Irish-owned and predominantly exporting firms. Of the 751 firms in the sample, 83% are exporters, compared to 44% of all Irish-owned firms (Central Statistics Office, 2000–2004). As such the
Exporting and productivity
The key assumption of Melitz-style models is that firms differ in their productivity. In order to enter any export market, firms must have high enough productivity to be able to cover the additional fixed and variable costs associated with operating in that market. Previous research has shown exporters to be more productive than non-exporters; here we extend this finding by showing that productivity can also be loosely linked to extent of market coverage.
We begin by looking at some simple
Destination hierarchy?
In the Melitz model, firms differ only by their productivity levels, and thus have different levels of unit costs. If trade barriers are identical for all firms, then these cost differentials will determine which export markets can be profitably entered. Specifically, each market will have a cost threshold. If the firm is efficient enough to enter the k-th market, then by definition it is efficient enough to be exporting to all markets with a higher cost threshold than k's. Data on which
Firm dynamics
This section looks at the implications of the model for how an individual firm's exports change, both within and across markets. The productivity cut-off derived earlier showed that firms will tend to enter low-GDP high-trade cost markets later. This implies that firms will tend to progressively sell less in each additional market as they move down the hierarchy and begin to enter these less popular markets. In addition, as productivity increases firms will tend to increase their sales more in
Market dynamics
We now turn from examining the number of markets that firms export to, and focus instead on the dynamics of entry and exit viewed from the viewpoint of specific markets. The Melitz–Chaney model implies that entering and exiting firms should be more marginal than existing exporters, as they will have productivity levels very close to the threshold for participating in the market. Section 7.1 looks at how prevalent entry and exit are across markets. Section 7.2 then presents a decomposition of
Changes in market coverage
Finally, we examine how individual firms change their number of export markets over time. According to the model outlined earlier, changes in number of markets are the result of either changes in firm productivity or changes in export market characteristics such as GDP or trade costs. The model suggests that the productivity cut-offs represent significant barriers to moving into new markets and therefore one would expect changes in market coverage to be gradual. Because the cut-offs are
Conclusions
The empirical literature on firm exporting has been restricted by a lack of data, most notably in relation to the breakdown by destination markets. This paper uses a survey of Irish firms over a five-year period, which contains detailed information on exports to over fifty markets. This allows us to document some new facts on the distribution of firms' export markets and on changes in market coverage.
These new facts largely correspond with a set of predictions drawn from a simple version of a
References (17)
- et al.
Exceptional exporter performance: cause, effect, or both?
Journal of International Economics
(1999) - et al.
Cities and growth: theory and evidence from France and Japan
Regional Science and Urban Economics
(1997) Empirical cross-section dynamics in economic growth
European Economic Review
(1993)- et al.
Exporters, jobs and wages in US manufacturing: 1976–1987
Brookings Papers on Economic Activity: Microeconomics
(1995) - et al.
Entry, expansion and intensity in the US export boom, 1987–1992
Review of International Economics
(2004) - et al.
Importers, exporters and multinationals: a portrait of firms in the U.S. that trade goods
Census of Industrial Production. Dublin
(2000–2004)Distorted gravity: the intensive and extensive margins of international trade
American Economic Review
(2008)
Cited by (123)
Technical barriers to trade and export performance: Comparing exiting and staying firms
2023, Economic ModellingWhy do some firms stop exporting?
2023, International Business ReviewMarket entry and the dynamics of export product quality: Evidence from Chinese manufacturing firms
2023, Economic Analysis and PolicyThird-country effects of regional trade agreements: A firm-level analysis
2023, Journal of International EconomicsHow can export improve firms’ energy efficiency? The role of innovation investment
2021, Structural Change and Economic DynamicsDynamic spatial panel data models with common shocks
2021, Journal of Econometrics
- 1
Thanks to Forfás for provision of the anonymous data and to Ali Ugur for help with queries on the data. I would also like to thank Karl Whelan for many useful comments and Jonathan Eaton and Samuel Kortum whose suggestions on previous work motivated much of this paper. The views expressed in this paper are the author's own, and do not necessarily reflect the views of the Central Bank and Financial Services Authority of Ireland or the ESCB.