Extensive and Intensive Margin Growth and Developing Country Exports

Cong Pham and Will Martin

DECRG

World Bank

Wednesday, March 14, 2007

Extensive and Intensive Margin Growth and Developing Country Exports

A key concern for development economists has been the risk that increasing exports will depress the terms of trade, and hence make rapidly-growing economies worse off—or at least severely diminish the gains from trade. In earlier decades, this trade pessimism was widely regarded as a sufficiently serious problem as to lead to developing countries to focus on import substitution as a path to growth. In more recent decades, an export-oriented approach has been widely seen as a more promising approach to growth.

These two different perspectives have pervasive effects on the appropriate approach to economic development, and on the role of international agencies, particularly the World Bank, but also agencies such as the World Trade Organization (WTO). They determine, for instance, whether the World Bank should encourage and support import-substituting industrialization projects, or adopt the more recently-favored paradigm of a relatively open economic policy using exports as an engine of growth. While the answer to this question was assumed to be known during the era of the Washington consensus, it seems less clear-cut in today’s era that has been characterized as one ofWashington confusion (Rodrik 2006).

One possible escape route from trade pessimism is the possibility that export growth does not consist solely of expansion in the quantities of the same goods. If, instead, export growth comes in large measure through expansion in the number of range of products exported, the outcome of export growth may be much more optimistic. Some recent studies have suggested that export growth may involve substantial increases in the range of varieties exported. This extensive margin growth may-- particularly if consumers prefer a wider variety of imports—help offset the deterioration in the terms of trade associated with traditional, intensive margin growth characterized by expansion of exports of the same products to the same markets. Another, related, escape route from trade pessimism is improvements in the quality of goods exported by developing countries.

Objectives and Goals of the Study

Whether extensive margin growth is sufficiently important as to require an overhaul of the way we approach trade and development policy depends upon its empirical importance, and on some key parameters of production and trade. Some key questions that will be used to guide this study include:

  1. To what extent is export growth in developing countries due to growth at the extensive and the intensive margins?
  2. How do we reconcile apparent differences in the results obtained from cross-section and time series analyses?
  3. To what extent do income growth, variable production costs and trade costs—and policies that influence them— influence the two margins of growth?
  4. Given the answer to questions 1 and 2 what can we infer about the welfare consequences of export growth?
  5. To what extent does the impact of the two margins on the export growth differ by sector (i.e. manufacturing versus agriculture), by the size of importers (i.e. large importers versus small importers) and by market structure (i.e. homogeneous products versus differentiated products)

Literature Review

The move away from import substitution to export promotion, and the more recent proposed move to growth diagnostics (Rodrik 2006) seem to have taken our policy recommendations well beyond our analytical toolkit. Most of our quantitative models are still based on a theoretical framework (the Armington (1969) framework) in which countries can only expand their exports by driving down their prices relative to those supplied by other countries. This contributes to an acute credibility problem for our modeling efforts in the form of low estimated gains from trade reform, and serious implied terms of trade deteriorations for countries which increase their exports. As noted by Krugman (1989), this problem also contaminates our estimates of both the income and demand elasticities, and hence the Armington elasticities which are the most influential determinants of the estimated welfare effects in conventional models of trade reform.

Some models based on the new trade theory of the 1980s (see Krugman 1980) get around this problem by assuming that at least part of the economy produces differentiated products, and that consumers have a love of variety based on the preference structure proposed by Dixit and Stiglitz (1977). Models based on this assumption have been used. However, these models have some features that are widely seen as inconsistent with the facts. In particular, they are seen as unrealistic in specifying consumers as preferring variety without limit. More importantly, they imply that all products will be exported to all markets, when in reality most potential trade flows are zero.

Hummels and Klenow (2005) have recently shown that neither the Armington nor the Krugman model can explain trade outcomes satisfactorily. The number of products exported rises less rapidly than predicted by the Krugman model, but more rapidly than predicted by the Armington model. Building on Feenstra’s classic (1994) study, Hummels and Klenow provided a quantitative framework for dividing trade expansion into an extensive margin involving expansion in the number of products exported and an intensive margin involving expansion in the quantity of each export. This framework has important implications for developing country export performance: if there is a sufficiently strong preference for variety, expansion at the extensive margin may cause the demand curve for a rapidly-growing exporter to shift out rapidly enough to overcome the price-depressing impacts of expanded export volumes.

The econometric analysis of Hummels and Klenow (2005), using cross-sectional approach, suggested that extensive margin growth was a large part of total growth in exports for a large sample of countries. However, it did not attempt to assess the welfare consequences of this phenomenon. Dimaranan, Ianchovichina and Martin (2007) used the Hummels-Klenow parameter estimates and embedded them into a global simulation model to assess their potential economic importance. For an experiment involving higher growth in China and in India, they found that allowing for growth and quality improvements increased the gains to China and to India by about 10 percent of theirinitial estimate. The gains to other developing countries, however, increased by almost 50 percent, reflecting the benefits from increased variety and higher quality of imports from China and India. Hummels and Klenow examined only the effects of higher growth on diversity and product quality. If such phenomena apply to trade growth resulting from trade reform, as well as from economic growth, the 50 percent multiplier on the welfare gains may be more typical than the 10 percent increases observed for China and India.

Adelean (2006) builds on Hummels and Klenow by incorporating a more general Constant-Elasticity-of-Substitution utility function that nests both the Armington and Krugman frameworks. Incorporating one additional parameter in our analysis enables us to determine, rather than impose, the strength of purchasers’ preference for variety. Helpman, Melitz and Rubenstein (2007) have recently made another important analytical advance by specifying a model that makes explicit the possibility of zero trade flows for a particular firm. It specifies a distribution of firm productivity levels of the type observed in firm-level analyses such as Bernard, Eaton, Jensen and Kortum (2003). With this, they use the Melitz (2003) formulation to determine a cut-off level of productivity below which firms—and by extension countries—do not enter international trade. This model provides an analytical framework for supply-side determination of the potential for growth at the extensive margin.

Relevance of the Issue

Even skeptics about export-oriented approaches to development recognize that export growth has been a central feature of successful development in the period since World War II. Rodrik ( ) has noted that no country has successfully grown without a large expansion in its trade. Whether export growth is cause or consequence, it is hugely important for any country undertaking successful growth and development. Whether this takes the form of intensive margin growth characterized by declining prices for any given good, or extensive margin growth is hugely important for successful development.

It is also widely believed that export diversification is important. Part of this belief arises from concerns about potential economic instability from reliance on an excessively narrow base of exports (Brenton and Newfarmer 2007). Another part arises from a view that new products provide new opportunities for learning and productivity growth. Another possible benefit of export diversification is a stimulus to demand arising from purchasers’ preference for variety in the goods that they consume. If, as argued by Hummels and Klenow (2005), extensive margin growth is an important part of the trade growth, then it is possible that these demand-side benefits would substantially increase the welfare gains both to countries expanding their exports and to the countries that trade with them.

METHODOLOGY

We begin by setting out a model with firm heterogeneity along the lines set out by Helpman, Melitz and Rubenstein (2007). This model defines import demand in each country using the Dixit-Stiglitz model of preferences and focuses on the supply side as the determinant of the number of varieties observed in trade. To allow for the possibility that the structure of preferences also plays a role in determining the observed variety of products entering trade, we generalize this structure slightly using the specification from Brown, Deardorff and Stern (1995) adopted by Adeleana (2007). This leads to a demand specification:

(1)

where is its consumption of product l and and are the set of country j’s own products available for its consumption and the set of products it imports from countries of the world, respectively (i.e.). is an empty set if country i has zero exports to country j. The elasticity of substitution across exporters is equal to . It is important to note thatdenotes a measure of the mass of firms of country j in country j while denotes a measure of the mass of country i’s firms that export to country j and is to be determined in equilibrium. Finally we assume so that the consumer faces a trade-off between the quantity per variety and the number of varieties imported.

Given the utility function (1) the demand for product l of exporter i by country j with income Yj is (2)

where is the price of product l in country j and is its ideal price index, defined as

(3)

Equation (2) shows that with the increase in the number of varieties imported from country i is not exactly offset by a decrease in the quantity per varieties.

On the production side firms produce output with a cost-minimizing combination of inputs equal to where a measures the number of bundles of the country’s inputs used by the firm per unit of output and is the cost of this bundle. Thus, a more productive firm is associated with a lower a. Let denote the productivity of the most productive and the least productive firm and the productivity distribution, respectively. We assume that country j has a measure of Fj firms, each one producing a distinct product and that its measure of firms Fj is increasing in its size. Specifically, as a country increases its size some of the newly created firms are more productive than the most productive present firm (i.e. decreases) and some of them are less more productive than the least productive present firm (i.e. increases).

If a firm serves only the domestic market it bears only production costs equal to. If it seeks to export its product to country i there are two additional costs it has to bear: a “melting iceberg” transport cost () and a fixed cost associated with the pair. We assume that while a component of fixed cost is an observed measure of country-pair fixed trade costs another component of fixed cost is dependent on the identity of the exporting and importing countries exclusively. In other words, the exporter-specific fixed cost varies across exporting countries but is the same to every firm in each of them: every firm in country j bears the same fixed cost when it seeks to export its product to foreign destinations. Similarly, every firm has to bear the same importer-specific cost when selling to the same importer. Finally, the country-pair fixed trade costs vary across exporter-importer pairs.

Under monopolistic competition in final products every exporter charges the following markup price:

(4)

This is also the market price for foreign consumers. Thus, a country i’s firm only chooses to serve the foreign market j if its operating profits from selling in j is positive:

(5)

Let be defined by =0, then country i has zero exports to country j if is smaller than country i’s. In other words, there is no exports of country i to country j if its most productive firm can’t make a profit that is large enough to cover the fixed costs.

Let is to be the exports of country i to country j then we have

(6) where .

From the setup laid down above it is now straightforward to see how the size of the exporter influences both the extensive and intensive margin of trade. First, an increase in the exporter i’s income increases the pool of firms. As some of the new firms are more productive than the most productive current firm the extensive margin increases. Second since the consumer values not only having more varieties but also higher quantity per variety the increase in also changes the intensive margin.

For an initial, exploratory analysis, we obtained the dataset from Hummels and Klenow (2005), for which we thank David Hummels. The component of this dataset that we used involves data at the six-digit level for 120 exporters and 76 importers in 1995. We also used data on the costs of doing business in a wide range of countries from

For an initial, exploratory analysis, we estimated a gravity equation based on the formulation above. This allows us to estimate the determinants of bilateral trade flows, as well as the changes in overall trade. We estimated it in difference form, comparing each exporter, j, with the rest of the world as an exporter to each market. We made the Tobit-type adjustment advocated by Helpman, Melitz and Rubenstein (2007) to deal with the frequency of zero entries in the bilateral trade matrices at this level of disaggregation. Like Hummels and Klenow (2005, p710), we specify the extensive margin as the ratio of exports of the products exported by country i from the rest of the world to the market of interest to exports of all products exported by the rest of the world to that market. The intensive margin is defined as the ratio j’s exports to the rest of the world’s exports of the same products. The product of the two margins is therefore the share of country j to the rest of the world’s exports to that market.

Table 1
Product-Based Extensive and Intensive Margins
(OLS Estimation)
A / B
Dependent Variables / Extensive / Intensive / Extensive / Intensive
Independent Variable
Ln(Yi/Yk) / 1.056 ** / 0.819 ** / 0.237 ** / 1.143 ** / 0.810 ** / 0.333 **
(24.84) / (21.68) / (9.59) / (25.13) / (19.22) / (11.05)
Ln(Distji/Distjk) / -1.213 ** / -1.063 ** / -0.150 ** / -1.263 ** / -1.022 ** / -0.241 **
(-17.58) / (-20.08) / (-3.27) / (-18.22) / (-17.99) / (-5.25)
Diff(Indexj –Mean_Indexk) / -1.046 ** / -0.755 ** / -0.291 ** / -1.157 ** / -0.827 ** / -0.330 **
(-4.63) / (-3.58) / (-3.81) / (-5.59) / (-4.19) / (-4.09)
Contiguity / 0.642 ** / 0.143 / 0.499 ** / 0.648 ** / 0.167 / 0.481 **
(3.16) / (0.82) / (3.86) / (3.11) / (0.97) / (3.88)
Language / 0.605 ** / 0.468 ** / 0.137 * / 0.705 ** / 0.461 ** / 0.244 **
(4.13) / (3.48) / (1.84) / (4.56) / (3.19) / (3.29)
Colony / 0.719 ** / 0.362 * / 0.357 ** / 0.737 ** / 0.345 * / 0.392 **
(3.53) / (2.05) / (3.58) / (3.79) / (2.01) / (3.94)
Importer Fixed Effects / Yes / Yes / Yes / Yes / Yes / Yes
Control for Sample / No / No / No / Yes / Yes / Yes
Selection Bias and Firm
Heterogeneity
Number of Obs. / 5490 / 5490 / 5490 / 5490 / 5490 / 5490
R2 / 0.66 / 0.55 / 0.35 / 0.67 / 0.57 / 0.37

We focus on the results in the right hand panel of Figure 1. This shows the effect of changes in key variables such as GDP, Distance and the cost of doing business on total export growth and the portion of export growth due to extensive margin growth and intensive margin growth. The striking feature of these results is the very large importance of extensive margin growth. As income grows, 70 percent of the growth in exports for our full sample is explained by extensive margin growth. As the distance between two trading partners increases, trade declines overall, and 80 percent of that decline occurs at the extensive margin, with the variety of products exported declining sharply. More importantly from a policy point of view, increases in the cost of doing business are found to have strong, significant, negative impacts on the extensive margin.

This result is interesting, but stands in strong contrast to the time-series results of Brenton and Newfarmer (2007) and Amiti and Freund (2007), both of whom find low rates of extensive margin growth over time. Our result is more in line with estimates of growt in exports of new varieties in Kehoe and Ruhl (200?) and in Martin and Manole (200?), both of whom found high rates of growth in new exports in dynamic exporters.

Part of the explanation relative to the Brenton and Newfarmer result may arise from greater disaggregation in the Hummels-Klenow database. While Brenton and Newfarmer work with around 3000 SITC categories, the Hummels-Klenow database contains 5017 Harmonized System categories. The much lower rate of extensive margin growth between the US and China in the Amiti and Freund study may reflect the large size of each partner in this bilateral trading relationship, and the consequently small number of cells initially unfilled relative to our global database.

One other possible explanation may be a sharp difference between the results based n cross-section and those based on time series. Perhaps, as in the celebrated case of the consumption function, the relationship needs to be more carefully specified before estimates from time series and cross-sectional analysis can be compared? Clearly, this is an issue that needs careful examination as we go forward.

References

Brenton, P. and Newfarmer, R. (2007), Watching more than the discovery channel: export cycles and diversification in development, Paper presented to the Conference on Export Diversification, World Bank, March 15.

Dimaranan, B., Ianchovichina, E. and Martin, W. (2007), ‘Competing with giants, who wins, who loses?’ in Winters, L. A. and Yusuf, S. eds. Dancing with Giants: China, India and the Global Economy. World Bank and the Institute for Policy Studies, WashingtonDC and Singapore.

Dixit, A.and Stiglitz, J. (1977), ‘Monopolistic competition and optimum product diversity’ American Economic Review 72(3):

1