Strategic Management Journal

Strat. Mgmt. J., 26: 855–871 (2005)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.481

EXPORT AND DOMESTIC SALES: THEIR

INTERRELATIONSHIP AND DETERMINANTS
ROBERT SALOMON1 and J. MYLES SHAVER2 *
1
Stern School of Business, New York University, New York, New York, U.S.A.
2
Carlson School of Management, University of Minnesota, Minneapolis, Minnesota,
U.S.A.

We examine the interrelationship between export and domestic sales. Our expectation is that
they are simultaneously determined, and as such should not be examined in isolation. We
also investigate how firm factors—such as R&D and advertising investments—and external
factors—such as market growth and exchange rate changes—impact export and domestic sales.
Using a non-recursive system of equations, we test our arguments on a representative sample
of Spanish manufacturing firms between 1990 and 1997. We find significant interrelationships
between export and domestic sales with striking differences between Spanish-owned firms and
foreign-owned firms operating in Spain. For Spanish-owned firms, domestic and export sales are
complements. These firms appear to focus on the domestic market and strength in the domestic
market drives their export sales. In contrast, domestic and export sales are substitutes for foreign-
owned firms. These firms’ export strategies appear subsumed under strategies of managing a
multinational network in which the focus is sales outside of Spain. We discuss the importance of
these findings for understanding and managing export strategies. Copyright  2005 John Wiley
& Sons, Ltd.

INTRODUCTION The perspective that we take is that firms, not

nations, engage in trade. Although goods flow
Exporting is the most widely used firm strat- between nations, it is generally firms that make
egy for international expansion. For example, total the decision to export. Therefore, we consider
world trade in merchandise and commercial ser- exporting a firm strategy that will have important
vices totaled U.S $7.8 trillion in 2002 (WTO press implications for performance.
release April 22, 2003). However, from a strate- Our paper makes three contributions. First, we
gic management perspective, we know relatively examine firms’ export sales and domestic sales
little about this strategy compared to other forms in our hypothesis generation and empirical tests.
of international expansion such as direct invest- We argue that export sales and domestic sales
ment and joint ventures. In this paper, we attempt are simultaneously determined. By acknowledging
a step toward deepening our understanding of firm their interdependence, we reach a more complete
exporting strategies. understanding of the relationship between export
and domestic sales, in addition to a more precise
Keywords: exporting; sales; export performance; interna- understanding of their determinants. Moreover,
tional expansion this approach corrects for potential biases in much
∗
Correspondence to: J. Myles Shaver, Carlson School of Man- of the prevailing literature that implicitly assumes
agement, University of Minnesota, 321-19th Avenue S, 3-365,
Minneapolis, MN 55455, U.S.A. export and domestic sales are independent. Sec-
E-mail: [email protected] ond, we examine how firm and external market

Copyright  2005 John Wiley & Sons, Ltd. Received 25 October 2000
Final revision received 21 March 2005
856 R. Salomon and J. M. Shaver

factors affect export and domestic sales. The firm the methods we employ to test the hypotheses.
factors that we examine include investments in The subsequent section presents the results from
advertising and R&D. The external factors that our analyses and discusses the findings. The final
we examine include exchange rates and economy- section concludes.
wide growth (i.e., changes in gross domestic prod-
uct). Third, we assess whether there are differences
in what affects export sales for domestic firms and PREVIOUS RESEARCH
foreign firms (i.e., in our context Spanish-owned
firms vs. foreign-owned firms operating in Spain). Scholars from disparate fields such as economics,
Because multinational firms operating subsidiaries management, and marketing have explored var-
in Spain manage their operations with a larger ied questions including: why exporting occurs,
global network in mind, there might be very dif- in which direction trade flows, how public pol-
ferent patterns in their export behavior compared icy affects export behavior, what generic export
with Spanish-owned firms. strategies entail, and what determines export per-
To test our arguments, we examine a stratified formance. Moreover, exporting research has been
representative sample of the Spanish manufactur- conducted at both micro (i.e., firm) and macro (i.e.,
ing sector between 1990 and 1997. These data industry or country) levels. At the macro level,
provide a very comprehensive and detailed view an abundant international trade literature addresses
of economic activity within a country. Moreover, issues such as comparative advantage, patterns of
the panel structure of the data allows us to better trade, gains from trade, and government policy (for
isolate how our hypothesized variables influence reviews see Helpman and Krugman, 1985; Gan-
export and domestic sales, versus other sources of dolfo, 1987). Scholars in this tradition see trade
firm heterogeneity. flows and gains as a function of factor endow-
We find striking differences in what drives ments, production technology, government policy,
export and domestic sales for Spanish-owned firms and size.
vs. foreign-owned firms operating in Spain. Specif- In contrast, the micro-level research points out
ically, we find that domestic and export sales are that firms generally make export decisions; there-
complements for Spanish-owned firms. For these fore, firm characteristics deserve greater attention
firms, export sales appear to be driven by preexist- than they have previously received because they
ing strengths in the domestic market. For foreign- stand to substantially influence export behavior. As
owned firms, domestic sales and export sales are a result, the factors that make a particular country’s
substitutes; that is, domestic sales negatively affect exports competitive might not equally influence
export sales. It appears that foreign-owned firms, all firms that make up those exports (Kravis and
when managing their domestic sales in the larger Lipsey, 1992). Rather, firm-specific characteristics
context of the multinational network, make trade- can lead to variance in behavior and performance
offs between sales in Spain and sales outside of across firms from the same country in the same
Spain. We find that both groups of firms’ domestic industry.
sales are positively related to advertising expen- Although researchers have recognized the im-
ditures. However, we find that R&D investments portance of studying export behavior at the firm
influence domestic and export sales in heteroge- level, this stream has not received nearly the atten-
neous ways across the two groups of firms. For tion devoted to macro level issues for two rea-
both sets of firms, growth in foreign economies sons. First, many research questions are predomi-
affects export sales and growth in the Spanish nantly concerned with macro factors—especially
economy affects domestic sales. The exchange rate questions motivated by economic and legisla-
results are mixed, although we generally find that a tive policy-makers. Second, the data necessary to
devaluation of the local currency leads to increased examine micro-level trade phenomena have been
exports, especially for Spanish-owned firms. difficult to obtain.
In the next section, we review the existing Although firm-level research lacks the depth of
research on exporting. We discuss how factors attention devoted to its macro counterpart, there
both internal and external to the firm affect export- is an established body of theoretical and empirical
ing behavior. The third section generates hypothe- work. Much of this research investigates the fol-
ses and the fourth section describes the data and lowing two issues: the export decision and export
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 857

performance. The former is defined as whether or firms more so than it induces entry by non-
not a firm is an exporter. The latter is defined as exporters. Taken together these results indicate that
the level of export sales or the level of export an exchange rate hysteresis likely has a greater
intensity (i.e., export sales divided by total sales). effect on the entry and exit decisions of firms than
Although the convention in this literature is to use on the export volume of existing exporters.
the term ‘export performance’ when referring to Exporting behavior is also subject to policy con-
‘export sales,’ we prefer to use the term ‘export straints imposed upon firms by foreign govern-
sales’ rather than ‘export performance’ because it ments. Export restrictions (voluntary or involun-
better describes the phenomenon. tary) encourage firms to forego entry (Yamawaki,
Scholars have identified numerous factors that 1986; Harrigan, 1993; Trefler, 1993; Brainard,
are related to exporting behavior at the firm level. 1993). Likewise, Cooper and Kleinschmidt (1985)
These include external determinants such as policy and Brainard (1997) discover a negative relation-
constraints, market growth, geographical proxim- ship between the level of country-specific export
ity, and exchange rate fluctuations; and firm char- restraints and export intensity. Finally, market
acteristics such as firm product attributes, manage- growth and geographical distance are likely to
rial attributes, and firm capabilities. influence a firm’s export behavior. Although ana-
lyzed at the macro level, Brainard (1997) shows
that the level of U.S. trade increases in foreign
External market characteristics
GDP and decreases in transportation costs.
Scholars have pointed out that some of the same
forces that affect the export behavior and com-
Firm characteristics
petitiveness of nations should also impact indi-
vidual firms (e.g., Dunning, 1993). For instance, In addition to the external factors just presented,
exchange rate shifts affect individual firms. How- the micro-level exporting literature addresses how
ever, a large body of theoretical work suggests firm characteristics influence exporting behavior.
that the decision to export is affected by a sunk- Firm characteristics include differential resources,
cost hysteresis, which stems from the fixed costs technological orientation, marketing skills, firm
of initiating export activity that cannot be recov- strategies, human capital, and managerial attitudes
ered ex post (e.g., Baldwin, 1988; Baldwin and or perceptions that provide firms with competitive
Krugman, 1989). As a result, firms do not imme- advantage.
diately respond to small exchange rate fluctuations. Prior studies have successfully linked techno-
Small fluctuations do not make exports sufficiently logical intensity (i.e., R&D expenditure) with the
profitable to warrant assuming the sunk costs of decision to export (Cavusgil and Nevin, 1984;
entry. Therefore, substantial exchange rate shifts Benvignati, 1990). The implications drawn from
are required in order to induce a firm to enter the aforementioned results are that technologically
an export market. The exit decision is similarly endowed firms are exporters. When relating invest-
affected. Firms will often not cease exporting when ment in R&D to export sales, the results are less
faced with small exchange rate changes because consistent. Many researchers offer evidence that
firms that terminate export activity must reincur R&D positively influences export intensity or vol-
the sunk costs to recommence exporting should an ume (Gruber, Mehta, and Vernon, 1967; Cavusgil,
offsetting exchange rate shift occur in the future. 1984; Benvignati, 1990; Braunerhjelm, 1996; Ito
Empirical evidence to date supports the hystere- and Pucik, 1993). Others find no significant rela-
sis arguments (e.g., Campa, 2004; Roberts and tionship between R&D and export intensity or vol-
Tybout, 1997). ume (Cooper and Kleinschmidt, 1985; Kravis and
By contrast, the effect of exchange rate shifts Lipsey, 1992; Ito and Pucik, 1993).
on the level of export sales, given that a firm is Other studies that examine firm characteristics
already an exporter, is more direct. For instance, and exporting show that advertising investment
Bernard and Jensen (2004) attribute increased does not significantly influence the export decision
export intensity of U.S. manufacturing plants to (e.g., Cavusgil and Naor, 1987). However, Benvig-
favorable exchange rate shifts. Similarly, Campa’s nati (1990) and Kravis and Lipsey (1992) find that
(2004) results indicate that a currency depreci- advertising is negatively related to export sales.
ation increases the export volume of exporting In either case, these findings are consistent with
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
858 R. Salomon and J. M. Shaver

Caves’ (1981) argument that advertising does not literature has focused predominantly on firm char-
carry well across national boundaries. He surmised acteristics as determinants of export behavior (i.e.,
that firms that spend heavily on advertising do so firm export sales/intensity). Independent variables
with the intention of increasing domestic sales. have been introduced without considering effects
Another stream of empirical work addresses found in other research, and few studies test the
managerial influences on exporting. These include effects of micro and macro determinants simulta-
managerial expectations and aspirations (Cavusgil, neously.
1984), attitudes toward risk (Wiedersheim-Paul, Third, firm-level data on exporting are often dif-
Olson, and Welch, 1978), attitudes toward foreign- ficult to assess because they tend not to be publicly
ers, managerial language skills and experience in available. Numerous firm-level studies use small-
foreign countries (Dichtl et al., 1983; Reid, 1983), scale, proprietary survey data in which they have
and other related constructs (for a thorough review sampled on the dependent variable—surveying
see Leonidou, Katsikeas, and Piercy, 1998). The only exporting firms to determine their character-
argument put forth by these scholars maintains istics and actions. Moreover, most of the studies
that managers initiate and develop export strat- (with the exception of Campa, 2004; Bernard and
egy and, as such, deserve to be the focal unit of Jensen, 2004, 1999a, 1999b; and Clerides, Lach,
interest. However, more than 20 years of empirical and Tybout, 1998) have been cross-sectional in
work has failed to produce a reliable relationship nature, making it difficult to do more than doc-
between managerial variables and export behavior ument associations between variables.
(Leonidou et al., 1998). Finally, many existing studies treat exporting as
an endeavor undertaken by small firms internation-
alizing for the first time. The implicit assumption
Summary is that firms are domestic-focused and then enter
Empirical studies of firm-level export behavior are international markets incrementally. Multinational
few compared to studies at the industry or country firms operate with a different logic from their
level. Moreover, there are few consistent research small, domestic-focused counterparts; and little
findings. These observations reflect several under- research examines the difference in export behav-
lying causes. ior across these types of firms.
First, many studies have analyzed exporting in In the subsequent sections we generate hypothe-
isolation. That is, they consider export outcomes ses and design the empirical investigation to ex-
without explicit recognition that export and domes- plicitly account for these concerns.
tic sales are likely simultaneously determined, and
interrelated. To the extent that studies have exam-
ined export and domestic sales, they have done so HYPOTHESES
by examining export intensity (export sales/total
sales) as a dependent variable (e.g., Tookey, 1964; The interrelationship between domestic and
Hirsch, 1970), but little attention has been directed export sales
to the fact that both the numerator and denominator We take the approach that firms maximize the sum
are endogenous. Namely, many firm and exter- of profits from export and domestic markets. As
nal factors affect both export sales and domestic such, export sales and domestic sales cannot be
sales. Forces that affect both domestic and export analyzed in isolation. Moreover, our expectation is
sales concurrently influence both the numerator that exports and domestic sales can directly affect
and denominator, making the net effect ambigu- each other. The precise nature of the relationship
ous. Thus, we estimate a system of equations that between domestic and export sales is unclear ex
allow for simultaneous effects between export and ante, because there are arguments that would sug-
domestic sales. gest both positive and negative effects.
Second, due to the various research traditions There are compelling reasons to expect a pos-
in which these studies are based, many different itive relationship. For example, if firms export
independent variables have been considered. The from a position of domestic strength (e.g., Bernard
macro literature has generally been interested in and Jensen, 1999a), and strength in the domestic
the effects of external characteristics on export out- market can be leveraged in international markets,
comes (i.e., aggregate trade flows), while the micro then we would expect domestic sales to positively
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 859

impact export sales. Likewise, should firms ben- Hypothesis 1: Advertising expenditures increase
efit from their export sales in the form of learn- domestic sales, all else equal.
ing that can be applied to domestic markets (e.g.,
Salomon and Shaver, 2005), then we would expect As Caves (1981) notes, advertising messages
the reciprocal effect: export sales positively impact do not generally carry well across international
domestic sales. A mutually positive relationship borders. Moreover, advertising efforts, especially
would suggest that each complements the other. for domestically focused firms, are often targeted
That is, domestic sales promote exports and vice to the domestic market (Benvignati, 1990; Kravis
versa. and Lipsey, 1992). Under these conditions we
There are also arguments to suggest trade-offs would not expect advertising efforts to be related to
between domestic and foreign sales. For example, export sales. Therefore, we do not expect advertis-
firms might produce goods for sale in domestic ing to affect export sales and do not offer a specific
and export markets, and shift output between mar- hypothesis.
kets when it becomes profitable to do so. Such The second firm characteristic that we examine
a trade-off will manifest as a negative relation- is R&D investment. R&D expenditure has gen-
ship between domestic and export sales, because erally been used to measure firm investment in
sales to one market deny sales to another. The innovative effort that is expected to foster techno-
nature of a trade-off between markets will also logical innovation (either product or process) and
be a function of the extent to which firms pro- hence future sales (e.g., Buckley and Casson, 1976;
duce goods that are targeted to domestic and export Morck and Yeung, 1991). Thus, R&D is intended
consumption. For example, if the goods that a to benefit the firm either through increased revenue
firm exports have little demand in the domestic or decreased cost. We therefore hypothesize:
market, then increased sales in the export market
would not reduce sales in the domestic market. Hypothesis 2: R&D expenditures increase do-
Therefore, export sales would not negatively affect mestic sales, all else equal.
domestic sales. Likewise, if the goods that a firm
that produces for the domestic market have lit- Whether R&D investments affect domestic sales
tle demand in export markets, then increased sales only, or export sales as well, has been debated in
in the domestic market would not reduce sales in the literature. Historically, firms have been argued
export markets. to focus new product innovations in domestic mar-
Although theory elicits expectations of interde- kets before they adapt and apply them to foreign
pendence between export and domestic sales, it markets (Vernon, 1966). If firms direct their R&D
does not provide guidance if a positive or nega- efforts to develop new products for the domestic
tive effect will dominate. This remains an empir- market, we would expect contemporaneous R&D
ical question. As such, we do not offer specific spending to have a greater impact on domestic
hypotheses with regard to the direction of such sales compared to export sales. However, it is
effects. In this respect, our findings can help inform possible that firms make R&D investments that
theory. are specifically targeted for export markets. More
recently, scholars have argued that firms do invest
in R&D in an attempt to innovate for the host mar-
Firm characteristics
ket (Kuemmerle, 1999). As such, we expect R&D
With regard to firm characteristics, we first exam- expenditures to be positively related to export
ine the impact of advertising investments on do- sales. Formally stated:
mestic and export sales. Firms invest in advertising
for a variety of reasons that include information Hypothesis 3: R&D expenditures increase export
dissemination, product differentiation, and brand sales, all else equal.
building. These efforts are intended to increase
firm sales by stimulating demand (i.e., increasing
External characteristics
volume), allowing firms to charge more per unit,
or both. We therefore expect advertising invest- With respect to external characteristics, we first
ments to be positively related to domestic sales. examine the effect of market growth on firms’
We hypothesize: domestic and export sales. Specifically, we expect
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
860 R. Salomon and J. M. Shaver

growth in the domestic market to increase sales service domestic and foreign markets. Although
for firms in the domestic market. All things equal, this describes the majority of firms in our sam-
growth stimulates demand and provides opportu- ple, foreign companies own some of the sample
nities for all firms to amplify their sales volume. firms. Multinational companies operating in Spain
We therefore hypothesize: might make decisions to allocate production and
output very differently from their Spanish-owned
Hypothesis 4: All else equal, growth in domestic counterparts if they manage their subsidiaries with
GDP increases domestic sales. a larger global network in mind (e.g., Rangan,
1998; Kogut, 1983, Feinberg and Keane, 2001). If
The international trade literature has long stress- multinational companies with production facilities
ed the connection between foreign GDP and export in Spain predominantly focus on sales outside of
sales. As foreign markets grow, countries are more Spain, then we might find relationships that differ
likely to engage in trade (Gruber and Vernon; for this set of firms. That is, the nature of the inter-
1970; Leamer and Stern, 1970). Although this relationship between export and domestic sales and
effect has been tested empirically in the macro- effects of the variables that we consider might vary
economic trade literature (see in particular Dear- across foreign-owned and Spanish-owned firms.1
dorff, 1984; Thursby and Thursby, 1987), we antic- Based on this expectation, we split the sample
ipate the very same effect to hold for individual into foreign-owned and Spanish-owned firms to
firms. That is, we expect opportunities for increas- examine whether the effects that we examine are
ing foreign sales with growth in distant markets. contingent on foreign ownership.2
We propose: In the next section we describe the data and
statistical approach that we use to test these effects.
Hypothesis 5: All else equal, growth in foreign
GDP increases export sales.

Similarly, we expect exchange rate fluctuations DATA AND STATISTICAL APPROACH

to influence the export behavior of firms. Specif-
ically, a depreciation of a firm’s home currency, Sample
relative to currencies in export markets, should The majority of the data we employ to test the
increase the volume of exports. This is because hypotheses are from a yearly survey of Spanish
currency depreciation makes goods from the home manufacturing firms conducted by the Fundación
country less expensive in the export country with- Empresa Pública. With the financial support of
out changing the home country currency price. the Spanish Ministry of Industry, the Fundación
Exporting firms, therefore, can increase export Empresa Pública surveys a stratified random sam-
sales when their home country currency depreci- ple of Spanish manufacturing firms to get a rep-
ates. Because our dependent variable of interest is resentative picture of the country’s manufacturing
sales, we do not explicitly focus on the exchange sector. The data cover the population of Spanish
rate hysteresis. As Campa (2004) shows, most of manufacturing firms with 200 or more employees
the influence of exchange rate changes on export and include 4 percent of the population of firms
sales comes through changes in export levels, not with at least 10 employees but fewer than 200
entry or exit, which are more prone to the hys- employees. The data we examine include all obser-
teresis effect. We therefore expect the effect of vations between the years 1990 and 1997.
exchange rate changes to be more direct on export The 1990 data provided information on 2188
volume and we hypothesize: firms. Small firms that dropped out of the origi-
nal sample were replaced every year by firms with
Hypothesis 6: All else equal, home-country cur- similar characteristics from the population. Thus,
rency depreciation increases export sales. the resulting base data set is an unbalanced panel

Foreign-owned vs. Spanish-owned firms 1

We thank an anonymous referee for suggesting this approach.
2
Implicit in our hypotheses, as in most existing In these data there are only 11 Spanish-owned firms with any
international production. Therefore, we do not analyze Spanish
studies on exporting, is that a company pro- multinationals separately and we eliminate them entirely from
duces, and then decides simultaneously how to the analysis.

Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 861

of 3060 firms and 15,226 firm-year observations. The export markets of these firms range from
Although the total available sample could have geographically proximal locations such as Europe
reached 17,504 firm-year observations (2188 firms and North Africa to more distant locations includ-
over 8 years), 2842 observations were missing ing the United States, Mexico, China, Japan, Tai-
data.3 MINER (1993) provides a detailed descrip- wan, and Korea. By contrast, only 11 of the 3058
tion of the data collection efforts (see also http:// firms in our sample (0.36%) reported having man-
www.funep.es). ufacturing activities outside of Spain. We remove
The Fundación Empresa Pública expends sub- these 11 firms from the analysis to avoid any
stantial effort in collecting, cleaning, and main- bias resulting from the existence of Spanish multi-
taining the data. Based on our examination of the nationals versus foreign multinationals operating
relevant data fields and after discussions with rep- within Spain. However, the contrast between num-
resentatives from the Fundación Empresa Pública, bers of exporting and investing firms highlights
we removed two firms from the entire panel and the predominance of exporting strategies versus
eliminated 39 observations due to data errors. This other international expansion strategies in these
reduced the sample to 3058 firms and 14,607 firm- data. Further, our analysis provides general insight
year observations. into small firm behavior from an internationalizing
economy such as Spain.
Dependent variables
Because export intensity does not measure the Independent variables
effects that we wish to explore, we measure, and The survey collected company advertising and
estimate simultaneous equations, using domestic R&D expenditures. Both variables are measured at
and export sales. The survey collects data on the firm level (i.e., at the Spanish subsidiary level
total sales and export sales; we therefore calculate if foreign-owned) and expressed in thousands of
domestic sales by subtracting export sales from pesetas.
total sales. Both variables are measured in thou- In order to test Hypothesis 4, we calculate firm-
sands of Spanish pesetas. specific real exchange rates (e.g., Campa, 2004).
Consistent with our assertion that exporting is Firm-specific exchange rates reflect exchange rate
the most widely used firm strategy for international changes most applicable to a firm’s export port-
expansion, more than half of the firms engage in folio. We favor this approach because it recog-
exporting. Table 1 summarizes the export status of nizes that not all firms have the same export mar-
the sample employed in this study. Approximately kets and are, therefore, not equally affected by
45 percent of the domestic firms export and 88 exchange rate movements. Because the measure is
percent of the foreign-owned firms export.4 expressed as pesetas divided by the foreign curren-
cies, increasing values indicate depreciation of the
3
Missing data were a result of non-response to particular survey
questions.
4
We define a firm as foreign-owned if a foreign parent owns any U.S. government, which is when a firm takes an equity stake of
equity in the focal firm. On average, foreign parents hold about 10 percent or more in a foreign facility (Graham and Krugman,
82 percent of the equity in their target investments, and equity 1995). However, the results presented are not sensitive to this
stakes varied from 1 to 100 percent. Our definition for inclusion definition. Similar results were obtained when we specify our
into the set of foreign-owned firms differs from that used by the cut-off at different levels of foreign ownership.

Table 1. Export status by subsample

1990 1991 1992 1993 1994 1995 1996 1997

Domestic firms
Non-exporters 1025 888 857 811 752 636 642 695
Exporters 667 613 627 597 641 612 660 792
Foreign firms
Non-exporters 69 60 56 47 42 31 34 32
Exporters 304 351 346 329 348 342 329 372

Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
862 R. Salomon and J. M. Shaver

peseta and decreasing values indicate appreciation of the survey was sent out, that firms responded
of the peseta. about their export markets.6
We calculate the firm-specific exchange rate Finally, we collected GDP data published by
index in the following manner. First, the desti- the U.S. Department of Energy. The U.S. DOE
nation markets were grouped into three broadly publishes detailed real GDP data obtained from
defined areas: OECD countries in the European Global Insight for over 200 countries for the period
Union, non-European OECD destinations, and the 1990–99. From this source we calculated two
rest of the world. We then calculated a specific such GDP measures. The first, DOMESTIC GDP,
exchange rate applied to each of the destination simply captures the GDP in Spain (reported in
regions as follows: peseta/euro for the European billions of U.S. dollars). The second measure,
Union, peseta/U.S. dollar for non-European OECD defined as FOREIGN GDP, represents a weighted
members, and the effective nominal exchange rate average of foreign GDP in the three potential
for Spain (from the IMF) for other world export destination markets (in billions of U.S. dollars).
destinations. The bilateral exchange rates were
then converted to a common index using U.S. dol- Statistical method
lar equivalents in order to compare the exchange
rates across currencies. Using the World Develop- In order to test the hypotheses, we employ two-
ment Indicators from the International Bank for stage least squares (2SLS) with firm fixed effects
Reconstruction and Development, we converted to estimate the following non-recursive system of
the nominal exchange rates into real rates by mul- equations.
tiplying the nominal rates by the ratio of infla-
tion between each destination region and Spain Exportsit = β1 Domestic Salesit
(see Kennedy, 2001).5 Finally, we converted the + 1 Xit + δi + εit (1)
three individual destination measures into a uni-
tary, firm-specific measure weighted by the pro- Domestic Salesit = β2 Exportsit
portion of firm export sales to each of the three + 2 Wit + δi + ηit (2)
destination regions.
Owing to data constraints, we made some trade- In these equations, Exportsit refers to export
offs and assumptions in order to generate these sales for firm i at time t, DomesticSalesit to domes-
firm-specific exchange rates. Ideally, we would tic sales for firm i at time t, Xit is a vector of inter-
have liked to identify all potential markets to which nal and external characteristics that affect export
each firm considers exporting. However, these data sales, and Wit is a vector of internal and exter-
are not possible to obtain. Therefore, for exporters, nal characteristics that affect domestic sales. The
we considered the relevant markets as the ones δi term represents a firm fixed effect meant to con-
they currently export to. For non-exporting firms, trol for unobserved heterogeneity; εit is an error
exchange rates were estimated using industry aver- term that is uncorrelated with Xit ; and ηit is an
ages for the largest export destination markets. For error term that is uncorrelated with Wit . We allow
instance, a non-exporter in the chemical products εit and ηit to correlate.
industry was assigned a weighted average of the In order to identify this system of equations,
exchange rates faced by exporters in that indus- we must have explanatory variables that predict
try. The assumption is that the non-exporting focal only exports and that predict only domestic sales
firm is most likely to export to the largest indus- (Kennedy, 2001; Greene, 2000). That is, there must
try destinations if it becomes an exporter. Finally, exist some elements in vector Xit that are differ-
we only have access to data regarding the break- ent from those in Wit and elements in Wit that
down of export markets served at two points in the are different from those in Xit . From our hypothe-
sample. Information about destination markets was ses, R&D appears in both equations. Advertising
only available for exporters in 1990 and 1994. It and domestic GDP predict only domestic sales.
was only in these years, when an extended version Our justification for this specification is that pre-
vious research argues that advertising is primarily
5
The regional inflation rates were calculated using weighted-
6
averages of annual inflation rates across all countries in each Due to the nature of the data, we cannot examine what drives
destination region. export levels to different national markets.

Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 863

directed at the domestic market and corroborat- test the hypotheses, the trade-offs of using this
ing empirical evidence suggests that advertising approach versus alternatives, and the sensitivity of
is not consistently related with international busi- our results to using alternative approaches.
ness activity (e.g., Caves, 1981; Morck and Yeung,
1992; Caves, 1996).7 Moreover, we expect that
changes in the domestic business cycle will more RESULTS
directly affect domestic sales rather than export
sales. By contrast, foreign GDP and real exchange In Tables 2 and 3 we present summary statis-
rates predict export sales but not domestic sales. tics and product moment correlations for each of
We model foreign GDP as affecting only export the subsample splits (Spanish-owned and foreign-
sales for the same reason that domestic GDP owned). Although the summary statistics and cor-
affects only domestic sales. In addition, we expect relations are generally as expected, we find some
exchange rates to more directly affect the sale of differences across subsamples. Notably, foreign-
goods in foreign markets.8 owned firms are larger than Spanish-owned firms.
The Appendix discusses in detail the motiva- Likewise, foreign-owned firms spend more (in
tion for choosing this econometric approach to both absolute terms, and as a percentage of sales)
on advertising and R&D. There are also similar-
7
Confirming this notion, we find that advertising is unrelated
ities across the subsamples. With respect to the
to export sales in specifications in which we include it in both dependent variables of interest, export sales and
equations. domestic sales are significantly positively corre-
lated (r = 0.50, p < 0.001 for the Spanish-owned
8
In the first stage of the estimation process we find that the
variables we use to identify the system have significant effects
on the dependent variables. This provides some indication of the firms and r = 0.36, p < 0.001 for the foreign-
appropriateness of the instruments. owned firms). These significant correlations lend

Table 2. Descriptive statistics and product moment correlations: domestic firms

1. 2. 3. 4. 5. 6. 7.

1. Domestic sales 1
2. Export sales 0.50 1
3. R&D 0.36 0.43 1
4. Advertising 0.34 0.15 0.21 1
5. Domestic GDP 0.03 0.06 0.03 0.03 1
6. Foreign GDP 0.03 0.06 0.03 0.03 0.98 1
7. Real exchange rate 0.07 0.07 0.07 0.05 0.56 0.65 1
Mean 2.10E6 6.34E5 2.84E4 5.52E4 516.16 23592.69 122.47
Standard deviation 7.30E6 3.41E6 1.76E5 2.93E5 20.91 1388.94 17.09
Minimum 0.00 0.00 0.00 0.00 491.96 21929.03 84.21
Maximum 2.82E8 1.02E8 4.68E6 6.24E6 557.25 26142.13 196.09

Table 3. Descriptive statistics and product moment correlations: foreign firms

1. 2. 3. 4. 5. 6. 7.

1. Domestic sales 1
2. Export sales 0.36 1
3. R&D 0.26 0.59 1
4. Advertising 0.58 0.57 0.36 1
5. Domestic GDP 0.05 0.07 0.03 0.03 1
6. Foreign GDP 0.05 0.08 0.03 0.03 0.98 1
7. Real exchange rate 0.05 0.03 0.06 0.01 0.56 0.64 1
Mean 1.34E7 5.73E6 2.94E5 5.15E5 517.02 23652.40 121.24
Standard deviation 4.48E7 2.78E7 1.79E6 1.98E6 20.68 1372.81 17.63
Minimum 0.00 0.00 0.00 0.00 491.96 21929.03 84.21
Maximum 9.71E8 5.40E8 3.33E7 3.29E7 557.25 26142.13 195.40

Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
864 R. Salomon and J. M. Shaver
Table 4. 2SLS fixed-effect estimates (t-values in parentheses)

Domestic firms n = 11, 515 Foreign firms n = 3092

Export sales (1) Domestic sales (2) Export sales (3) Domestic sales (4)

Domestic sales 0.26∗∗∗ (6.79) −0.42∗∗∗ (−8.72)

Export sales 0.23 (0.67) 0.02 (0.07)
Advertising 3.31∗∗∗ (8.44) 6.18∗∗∗ (6.94)
R&D 5.17∗∗∗ (28.07) 0.06 (0.03) 0.95∗∗∗ (4.16) 1.01∗∗∗ (3.41)
Domestic GDP 6298.80∗∗ (2.27) 74793.21∗∗ (2.12)
Foreign GDP 47.38∗∗∗ (2.97) 1953.34∗∗∗ (8.27)
Real exchange rate 4443.74∗∗∗ (3.38) 24997.67 (1.23)
Constant −1.72E6∗∗∗ (−6.49) −1.48E6 (−1.15) −3.81E7∗∗∗ (−9.42) −2.89E7∗∗ (−1.79)

∗ ∗∗ ∗∗∗
p < 0.10; p < 0.05; p < 0.01 (one-tailed tests)

support to our contention that domestic sales and for these firms, increased sales in Spain come at
export sales are not independent and as such cannot the expense of sales outside of Spain.9
be analyzed in isolation. Turning to the hypothesized variables, we find
Table 4 presents the results from the systems that advertising has a positive effect on domes-
of equations that we estimate. Columns 1 and 2 tic sales for both subsets of firms, which supports
present the results for the subsample of Spanish- Hypothesis 1. A 1 peseta increase in advertis-
owned firms. Columns 3 and 4 present the results ing increases domestic sales by 3.31 pesetas for
for the subsample of foreign-owned firms. domestic firms and 6.18 pesetas for the foreign-
We first examine the relationship between the owned firms. Our predictions are that the coeffi-
exports and domestic sales. For the Spanish-owned cient estimates on R&D expenditures would like-
firms, we find that domestic sales significantly wise be positive for domestic sales. For domes-
influence export sales. The effect is positive, which tic firms we find that R&D does not have an
is consistent with the pairwise correlation. We find effect on domestic sales. This was rather surpris-
that a 1 peseta increase in domestic sales increases ing, and for this group of firms Hypothesis 2 is not
export sales by 0.26 pesetas. In column 2, we find supported. By contrast, R&D expenditures have
no relationship between export sales and domestic a positive and significant effect on the domestic
sales. These results suggest that domestic sales sales of foreign-owned firms. The magnitude of
and export sales are complements. Moreover, the the effect is such that a 1 peseta increase in R&D
complementarity is such that only increases in expenditure increases domestic sales by 0.95 pese-
domestic sales lead to increases in export sales, tas.
not vice versa. When examining the results with respect to
Columns 3 and 4 show a different relationship R&D and export sales, we find that R&D is pos-
for the firms with foreign ownership. In column itively related to export sales for both sets of
3 we find that domestic sales negatively affect firms. As such, Hypothesis 3 is supported. For
export sales. A 1 peseta increase in domestic sales Spanish-owned firms, the results suggest that an
reduces exports by 0.41 pesetas. This was some- additional peseta in R&D expenditure increases
what surprising, and inconsistent with the pairwise exports by 5.17 pesetas. Coupled with the non-
correlation. We find no evidence of a reciprocal significant finding of R&D on the domestic sales
relationship—exports do not affect domestic sales of Spanish firms, these results suggest that firms
in this system of equations. Therefore, for firms
with foreign ownership we find that exports and
9
domestic sales are substitutes and that the substi- The difference in effect across the two groups of firms does not
appear influenced by large differences in capacity utilization,
tution is such that increased domestic sales lead which could potentially drive substitution versus complemen-
to decreased exports, not vice versa. This is con- tarity. Approximately 10 percent of the foreign-owned firms
sistent with the interpretation that the goods that have capacity utilization above 90 percent versus approximately
8 percent for the Spanish-owned firms. Moreover, we find no
the foreign-owned firms sell in Spain are similar statistical difference in the mean capacity utilization across the
to the ones that they export out of Spain. Namely, two groups of firms.

Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 865

might increase their R&D expenditures in prepa- Foreign GDP positively impacts both foreign-
ration for entry into export markets. That is, owned and Spanish-owned firms’ export sales. As
they incur additional development expenditures such, Hypothesis 5 also receives support. These
to increase foreign sales. For the foreign-owned results suggest that a $1 billion increase in the
firms, R&D expenditures are likewise positively foreign GDP measure leads to a 47,380 peseta
associated with export sales. For foreign-owned increase in the export sales of domestic firms and
firms, a 1 peseta increase in R&D expenditure is a 1.95 million peseta increase in export sales for
associated with a 1.01 peseta increase in export foreign-owned firms. As with domestic GDP, the
sales. export sales of foreign firms (mean exports of
At this point we raise two considerations when 5.73 billion pesetas) are actually more sensitive
interpreting the results with respect to advertising to changes in foreign GDP than the export sales
and R&D. First, the coefficients for these effects of domestic firms (mean exports of 634 million
are rather large in magnitude—namely a 1 peseta pesetas).
increase in R&D or advertising results in a multiple Our test of Hypothesis 6 appears in columns 1
peseta increase in sales—and it begs the question and 3. We find that exchange rate changes affect
of why firms would not increase such expenditures the export sales of domestic firms, but not for-
if they had effects of this magnitude. One impor- eign firms. As such, Hypothesis 6 receives mixed
tant consideration is that these estimates are from support. The results suggest that although domes-
specifications in which sales, rather than profits, tic firms increase their export sales in response to
is the dependent variable. We would expect coef- depreciation in the peseta, foreign firms do not.
ficients of smaller magnitude with profits as the This implies that the export sales of foreign-owned
dependent variable.10 Second, because these effects firms are less sensitive to exchange rates fluctu-
are measured contemporaneously with sales, there ations—perhaps because they are more likely to
is a concern of reverse causality. Namely, the extra employ currency hedges, or to conduct business
cash flow from exports could facilitate increased in a single currency across all markets; thereby
R&D spending if firms alter products for those for- reducing currency exposure. However, this result
eign markets. Likewise, the extra cash flow from warrants some caution in interpretation due to the
domestic sales might encourage firms to invest in substantial correlation between foreign GDP and
advertising.11 the exchange rate. In fact, when we eliminate for-
With regard to the external variables of interest, eign GDP from both equations, we find that the
we find support for Hypothesis 4. Domestic GDP exchange rate has a positive and significant effect
affects domestic sales for both sets of firms. A on the export sales of both Spanish and foreign-
$1 billion increase in domestic GDP translates owned firms.
into a 6.30 million peseta increase in domestic
sales for Spanish firms and a 74.79 million peseta Sensitivity analyses
increase in domestic sales for foreign firms. Given
that the average domestic sales of foreign and First, to account for the possibility that firms
Spanish firms in the sample are 2.10 billion and respond to exchange rate changes with some lag
13.4 billion respectively, as a percentage of sales, rather than immediately (i.e., the J-curve effect;
the domestic sales of foreign-owned firms are more Bahmani-Oskooee, 1985), we examined specifica-
sensitive to changes in domestic GDP. tions that included lagged values of real exchange
rate. The pattern of results with respect to the
10
hypothesized variables, both in terms of magnitude
In the data, the average gross margin divided by sales is 10.38
percent. Therefore, for every 9.63 pesetas in sales, 1 peseta and statistical significance, does not change sub-
represents margin. As a result, a 1 peseta increase in advertising stantially. Moreover, with respect to both sets of
or R&D expenditure on average increases sales by more than 1 firms we find that the lagged exchange rate variable
peseta yet does not increase profits by more than 1 peseta.
11
has a positive, although weak (p < 0.05), effect in
We considered including lagged values of the R&D and adver-
tising expenditure variables to the specification; however, these the hypothesized direction.
measures are very highly correlated with the contemporaneous Second, because the domestic GDP variable
measures, thereby confounding their interpretation. When we does not vary across firms for any given year, we
included the lagged measures in lieu of contemporaneous values,
we found results that were consistent in statistical significance, could not include year fixed effects in the specifi-
albeit slightly weaker in magnitude than those presented. cation. We examined the sensitivity of our results
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
866 R. Salomon and J. M. Shaver

when we removed the GDP measures and included rate changes rather than small changes, the true
year effects. There were no material changes in the effect of exchange rates would actually be stronger
results. than that measured. Given that we find a positive
Third, we split the sample by time in order and significant effect of exchange rates on export
to explore the stability of the parameters across volume, this implies a conservative estimate. Nev-
time periods. For instance, during the early 1990s ertheless, we acknowledge that such effects might
Spain experienced a severe economic downturn be present in our data and we would encourage
followed by a recovery in the mid to late 1990s. It future research to investigate these issues further.
could be that the effects of the independent vari- The aforementioned limitations notwithstanding,
ables differ depending upon the economic cycle. our results have implications for scholars and man-
Therefore, we split each of the subsamples and agers. Below we discuss the implications of such
ran separate regressions for two distinct periods: findings for both research and practice.
recessionary (1992–94) and recovery (1995–97).
Although the results were generally consistent
with those presented (especially with regard to the DISCUSSION
relationship between domestic and export sales),
there were some informative contrasts. Specifi- We find at least partial support for all of our
cally, both dependent variables (for both Spanish- predictions. Moreover, we find striking differences
and foreign-owned firms alike) were more sen- in many of these effects across Spanish-owned
sitive to external market fluctuations (GDP and and foreign-owned firms operating in Spain. The
exchange rate changes) in the recessionary period. support of our predictions combined with the often
Because greater variation in external market char- differing effect across the two groups of firms
acteristics exists during the recessionary period has important implications for guiding managerial
versus the recovery period and because each period actions, reconciling equivocal findings in previous
is only 3 years in length, the different levels of research, and guiding future exporting research.
variability in the data might drive these differ- We find that for Spanish-owned firms exports
ences over time. Therefore, we do not want to draw and domestic sales are complements because do-
strong conclusions based on these results.12 mestic sales positively affect export sales. Their
domestic sales are sensitive to changes in the
Limitations growth of the Spanish economy and to changes in
advertising spending. The export sales of Spanish-
At this point we draw two caveats that we discuss owned firms are influenced by changes in R&D
in the Appendix. First, we ignored the bounded expenditure and changes in external factors such
nature of the dependent variables in favor of con- as foreign GDP and exchange rates. The interpre-
trolling for unobserved firm heterogeneity. Al- tation of these results in concert is that the Spanish-
though we did check the robustness of the results owned firms tend to focus on the domestic market
across various specifications, we recognize that and that strength in the domestic market drives
accounting for this factor might lead to more their expansion into export markets—potentially
precise estimates and conclusions. As such, we with some investment in R&D to adapt their prod-
encourage future work to reconsider such issues ucts to foreign markets. Therefore, export strength
when more advanced methodological techniques appears to be a benefit that many firms realize from
become available. domestic strength.
Second, the influence of hysteresis, to the extent In contrast, domestic GDP, R&D investments,
it is present in these data, might induce a sample and advertising expenditures all influence local
bias that leads to inaccurate coefficient estimates. sales of foreign-owned firms. Moreover, export
To the extent that hysteresis affects the results we and domestic sales are substitutes for the foreign-
present, the bias suggests that our estimates with owned firms operating in Spain because domestic
respect to exchange rates would likely be underes- sales negatively affect export sales. And although
timated. Because a hysteresis effect suggests that these firms’ export sales are sensitive to growth in
firms are more likely to respond to large exchange foreign markets (and, to a certain extent, exchange
rate changes), it appears that these foreign firms
12
We thank an anonymous reviewer for suggesting this analysis. operating in Spain are managed as part of a
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 867

network of manufacturing operations and that the CONCLUSIONS

sales focus of these operations tends to be outside
of Spain. Therefore, the export strategies of these We examine how firm and external forces affect
firms appear to be subsumed under broader strate- domestic and export sales while explicitly con-
gies that include direct investments. sidering the interdependence between export and
The implications of these findings for managers domestic sales. Specifically, we examined the
are consistent with findings in the literature that influence of firm advertising investments, firm
firms should undertake export strategies from posi- R&D investments, market growth, and exchange
tions of strength in the domestic market (e.g., rate changes on domestic sales and export sales.
Bernard and Jensen, 1999a). It appears that domes- In doing so, we add to a relatively small num-
tic strength encourages export sales. Moreover, ber of studies that examine firms’ strategies with
should firms expand from strength and realize sales respect to exporting. Although exporting is the
in the foreign market, this can facilitate other bene- most prevalent form of international expansion, the
fits such as potentially aiding innovation (Salomon firm-level study of this phenomenon—especially
and Shaver, 2005; Salomon, 2002) and helping from a strategic management perspective—is rare.
mitigate liquidity constraints (Campa and Shaver, Our findings have important implications for
2002). However, such implications would apply managers. First, the interdependence between do-
only for firms that do not have foreign produc- mestic and export sales that we document implies
tion facilities. That is, exporting is the way that that strategic decisions that focus on domestic or
these firms participate internationally. For firms export sales in isolation can lead to suboptimal firm
with direct investments, exporting is often the tool choices. A complicating factor in jointly consider-
that allows for the realization of location advan- ing these decisions is that for the domestic firms
tages. For example, exporting allows firms to move in our sample domestic sales positively influenced
production capacity around the world while selling export sales, but not vice versa. This suggests
in various domestic markets. that the path to firm growth, both for exports and
Our findings also provide insight into why many domestic sales, involves exporting from a position
results in the firm-level exporting literature have of domestic strength. The implication is that effec-
been equivocal. First, many existing studies ignore tively managing domestic operations aids exports.
the potential effects of domestic sales on export For the foreign-owned firms operating in Spain, the
sales. Furthermore, many fail to control for unob- finding that domestic sales negatively affect export
served firm heterogeneity when analyzing export sales suggests that there exists a trade-off within
sales.13 Failure to control for forces such as these foreign-owned firms’ network of sales when sales
that affect both domestic and export sales concur- increase in the market where goods are produced.
rently might have led to inconsistent, and poten- Second, we show that firm investments such
as R&D affect export sales. This suggests that
tially biased, results across various studies. Sec-
export sales are not solely driven by environmen-
ond, we find that firm and external characteristics
tal factors, such as exchange rate changes and
are important determinants of export sales. Studies
market growth, which are outside of the control
that do not account for both effects might suffer
of managers. Therefore, a firm’s export strategy
biases due to omitted variables. Third, we find that
must involve more than simply scanning for and
exports behave differently for firms with and with-
responding to changes in the macro business envi-
out direct investments. Therefore, to the extent that
ronment. Moreover, the interdependence between
previous samples had unequal weighting of these
export and domestic sales suggests that managers
types of firms and did not explicitly control for must make investments in R&D and advertising,
these effects, this alone could account for the lack understanding that they can have both direct and
of consistent results. In summary, attention to all of indirect effects on domestic and export sales.
these points can guide firm-level export research. Finally, we emphasize the importance of consid-
ering domestic-owned and foreign-owned firms as
entities with different internal logics. The Spanish-
13
Not surprisingly, when we run our results without controlling owned companies in our sample seem to resem-
for firm effects we find a positive and significant effect of
domestic sales on exporting sales for both groups of firms ble small, domestic-focused firms that are the
(consistent with the pairwise correlations). implicitly assumed actor in much of the exporting
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
868 R. Salomon and J. M. Shaver

literature. Our results for this set of firms are Bernard A, Jensen JB. 1999b. Exporting and productiv-
consistent with the prevailing literature that sug- ity. NBER working paper #7135.
Bernard A, Jensen JB. 2004. Entry, expansion and
gests that firms should expand internationally from
intensity in the U.S. export boom. Review of
a position of domestic strength. By contrast, sub- International Economics 12(4): 662–675.
sidiaries of foreign parents that are a part of multi- Blonigen B. 2001. In search of substitution between for-
national networks are managed with consideration eign production and exports. Journal of International
to the overall network. This highlights the impor- Economics 53(1): 81–104.
tance of a burgeoning literature examining whether Blundell RW, Smith RJ. 1989. Estimation in a class
of simultaneous equation limited dependent variable
foreign production and exporting are substitutes or models. Review of Economic Studies 56(1): 37–57.
complements (see Blonigen, 2001). Brainard SL. 1993. A simple theory of multinational
To conclude, we reiterate that exporting is corporations and trade with a trade-off between
the most widely used foreign expansion strategy. proximity and concentration. NBER working paper
In this paper we document the interrelationship #4269.
Brainard SL. 1997. An empirical assessment of the prox-
between export and domestic sales and highlight imity–concentration trade-off between multinational
how both firm and external market characteristics sales and trade. American Economic Review 87(4):
are important determinants of export and domestic 520–544.
sales. Moreover, we find important differences Braunerhjelm P. 1996. The relation between firm-
in the exporting behavior of Spanish-owned and specific intangibles and exports. Economic Letters 53:
foreign-owned companies. These findings have 213–219.
Buckley PJ, Casson MC. 1976. The Economic Theory of
important implications for effectively managing the Multinational Enterprise. Macmillan: London.
this important international expansion strategy. Campa J. 2004. Exchange rates and trade: how important
is hysteresis in trade? European Economic Review
48(3): 527–548.
ACKNOWLEDGEMENTS Campa J, Shaver JM. 2002. Exporting and capital
investment: on the strategic behavior of exporters.
IESE and University of Minnesota working paper.
We are grateful to the Fundación Empresa Pública Caves RE. 1981. Intra-industry trade and market structure
for providing us with access to these data and in the industrial countries. Oxford Economic Papers
to José Campa for providing us with the firm- 33: 203–223.
specific exchange rate data. We appreciate help- Caves RE. 1996. Multinational Enterprise and Economic
ful, detailed, and constructive comments from Analysis. Cambridge University Press: Cambridge,
MA.
the anonymous reviewers, Associate Editor Will Cavusgil ST. 1984. Organizational characteristics asso-
Mitchell, José Campa, Wilbur Chung, Bill Greene, ciated with export activity. Journal of Management
Sevgin Eroglu, and Michael Lenox. Studies 21: 3–22.
Cavusgil ST, Nevin J. 1984. Internal determinants of
export marketing behavior: an empirical investigation.
REFERENCES Journal of Marketing Research 18: 114–119.
Cavusgil ST, Naor J. 1987. Firm and management
Arellano M, Bond S. 1991. Some tests of specification for characteristics as discriminators of export marketing
panel data: Monte Carlo evidence and an application activity. Journal of Business Research 15: 221–235.
to employment equations. Review of Economic Studies Clerides SK, Lach S, Tybout JR. 1998. Is learning by
58: 277–297. exporting important? Micro-dynamic evidence from
Bahmani-Oskooee M. 1985. Devaluation and the J-curve: Colombia, Mexico, and Morocco. Quarterly Journal
some evidence from LDCs. Review of Economics and of Economics 113(3): 903–947.
Statistics 67(3): 500–504. Cooper R, Kleinschmidt E. 1985. The impact of export
Baldwin R. 1988. Hysteresis in import prices: the strategy on export sales performance. Journal of
beachhead effect. American Economic Review 78: International Business Studies 16(1): 37–55.
773–785. Deardorff A. 1984. Testing trade theories and predicting
Baldwin R, Krugman P. 1989. Persistent trade effects trade flows. In Handbook of International Economics,
of large exchange rate shocks. Quarterly Journal of Vol. 1, Jones RW, Kenen PR (eds). North-Holland:
Economics 104: 635–654. Amsterdam; 467–517.
Benvignati A. 1990. Industry determinants and ‘differ- Dichtl E, Leibold M, Koglmayr HG, Muller S. 1983.
ences’ in U.S. intrafirm and arms-length exports. The foreign orientation of management as a
Review of Economics and Statistics 72: 481–488. central construct in export-centered decision making
Bernard A, Jensen JB. 1999a. Exceptional exporter processes. Research for Marketing 10: 7–14.
performance: cause, effect, or both? Journal of Dunning J. 1993. Multinational Enterprises and the
International Economics 47(1): 1–25. Global Economy. Addison-Wesley: Reading, MA.
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 869
Feinberg SE, Keane MP. 2001. U.S.–Canada trade decision to export. American Economic Review 87(4):
liberalization and MNC production location. Review 545–564.
of Economics and Statistics 83(1): 118–132. Salomon R. 2002. Spillovers to foreign market partici-
Gandolfo G. 1987. International Economics I: The Pure pants: assessing the impact of exporting and firm het-
Theory of Trade. Springer: New York. erogeneity on innovative outcomes. PhD dissertation,
Graham EM, Krugman P. 1995. Foreign Direct Invest- New York University.
ment in the United States. Institute for International Salomon R, Shaver JM. 2005. Learning by exporting:
Economics: Washington, DC. new insights from examining firm innovation. Journal
Greene WH. 2000. Econometric Analysis. Macmillan: of Economics and Management Strategy 14(2):
New York. 431–460.
Gruber W, Mehta S, Vernon R. 1967. The R&D factor Thursby JG, Thursby MC. 1987. Bilateral trade flows,
in international trade and international investment of the Linder hypothesis, and exchange risk. Review of
United States industries. Journal of Political Economy Economics and Statistics 69(3): 488–495.
75: 20–37. Tookey DA. 1964. Factors associated with success
Gruber W, Vernon R. 1970. The technology factor in in exporting. Journal of Management Studies 1:
a world trade matrix. In The Technology Factor 48–66.
in International Trade, Vernon R (ed). Columbia Trefler D. 1993. Trade liberalization and the theory of
University Press: New York; 232–272. endogenous protection: an econometric study of U.S.
Harrigan J. 1993. OECD imports and trade barriers in import policy. Journal of Political Economy 101(1):
1983. Journal of International Economics 35(1–2): 138–160.
451–471. Vernon R. 1966. International investment and interna-
Helpman E, Krugman P. 1985. Market Structure and tional trade in the product cycle. Quarterly Journal
International Trade. MIT Press: Boston, MA. of Economics 80: 190–207.
Hirsch S. 1970. The Export Performance of Six Wiedersheim-Paul F, Olson HC, Welch LS. 1978. Pre-
Manufacturing Industries. Praeger: New York. export activity: the first step in internationalization.
Ito K, Pucik V. 1993. R&D spending, domestic com- Journal of International Business Studies 9(1):
petition, and export performance of Japanese manu- 47–58.
facturing firms. Strategic Management Journal 14(1): Yamawaki H. 1986. Exports, foreign market struc-
61–75. ture and profitability in Japanese and U.S. manu-
Kennedy P. 2001. A Guide to Econometrics. MIT Press: facturing. Review of Economics and Statistics 68:
Cambridge, MA. 618–627.
Kogut B. 1983. Foreign direct investment as a sequential
process. In The Multinational Corporation in the
1980s, Kindleberger CP, Audtresch D (eds). MIT
Press: Cambridge, MA; 38–56. APPENDIX: ECONOMETRIC
Kravis I, Lipsey R. 1992. Sources of competitiveness of APPROACH
the United States and of its multinational firms. Review
of Economics and Statistics 74: 193–201.
Kuemmerle W. 1999. The drivers of foreign direct In determining the appropriate econometric ap-
investment into research and development: an proach to test our hypotheses, we were faced
empirical investigation. Journal of International
Business Studies 30(1): 1–24.
with a number of trade-offs given the underlying
Leamer E, Stern R. 1970. Quantitative International question that we wish to examine and the nature
Economics. Aldine: Chicago, IL. of the data. We wish to simultaneously estimate
Leonidou LC, Katsikeas CS, Piercy NF. 1998. Identify- domestic and export sales; nevertheless, our data
ing managerial influences on exporting: past research have the following features: (a) we observe firms
and future directions. Journal of International Market-
ing 6: 74–102.
over multiple years—hence the concern of lack
MINER (Ministerio de Industria y Energı́a). 1993. Un of independence in the error terms across observa-
Panorama de la Industria Española. Madrid, Spain. tions for each firms (i.e., there exists unobserved
Morck R, Yeung B. 1991. Why investors value multina- firm heterogeneity); (b) the dependent variables
tionality. Journal of Business 64: 165–187. are limited because we do not observe negative
Morck R, Yeung B. 1992. Internalization: an event study.
Journal of International Economics 33: 41–56. values for export or domestic sales; and (c) there
Rangan S. 1998. Do multinationals operate flexibly? might be persistence in export sales if exchange
Theory and evidence. Journal of International rate hysteresis is an important element of export
Business Studies 29(3): 217–237. sales.
Reid SD. 1983. Firm internationalization, transaction Because the focal contribution of our paper is
costs, and strategic choice. Journal of International
Business Studies 2: 44–56. the consideration of the simultaneity of export
Roberts M, Tybout J. 1997. The decision to export in and domestic sales, our primary approach is to
Columbia: an empirical model of sunk costs and the estimate a non-recursive system of equations of
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
870 R. Salomon and J. M. Shaver

the following form: in our system are limited, we explored the pos-
sibility of employing the simultaneous Tobit esti-
Exports = f (DomesticSales, R&D, Foreign mator of Blundell and Smith (1989). There were
two complications in using this approach. First,
GDP, Real exchange rate) (A1) the Blundell and Smith estimator is for a recursive
DomesticSales = g(Exports, Advertising, system of equations and we have a non-recursive
system of equations. Second, our data are panel
R&D, Domestic GDP) (A2) in nature, which raises the concern about lack of
independence across observations as we discussed
This system of equations is identified because we previously in this Appendix. The Blundell and
model Foreign GDP and Real exchange rates as Smith estimator does not address non-independent
direct determinants of export sales, but not domes- observations.
tic sales. Likewise, we model Advertising and The results that we present in the main body
Domestic GDP as direct determinants of Domes- of the text show that export sales are a function
tic sales and not export sales. The first step in our of domestic sales and not vice versa. Therefore,
efforts to estimate this system of equations was to although data-driven, we decided that we could
employ 2SLS on each equation. We then estimated ‘overlook’ the first issue and model a recursive sys-
2SLS with firm random effects to assess whether tem of equations. However, the second issue posed
unobserved heterogeneity was a concern. We found more of a problem given that the tests we describe
that the random effects were statistically signifi- above point to the existence of unobserved firm
cant, supporting our concern that there exists unob- heterogeneity in the data.
served heterogeneity and that panel data techniques Therefore, we found ourselves having to make a
were appropriate. The assumption underlying the trade-off. Given that our results show the existence
random effects estimator is that the unobserved of unobserved heterogeneity we decided it was
heterogeneity is not correlated with the regres- more appropriate to employ the 2SLS approach.
sors. To assess this, we estimated a 2SLS fixed- We did, however, run some sensitivity analyses
effects specification, which does not assume that across recursive simultaneous equation models that
the unobserved heterogeneity is uncorrelated with did not account for firm effects or limited depen-
the regressors. Using a Hausman test, we evaluated dent variables using the Blundell and Smith esti-
whether the random effects assumption was valid. mator. We found that a recursive model without
The Hausman test rejected the null hypothesis that controlling for unobserved firm effects provided
the random effect was uncorrelated with the regres- extremely consistent results (in terms of level of
sors. Therefore, we favored the 2SLS fixed-effect statistical significance and sign of coefficient esti-
approach. mates for all explanatory variables). Therefore,
We examined the sensitivity of our estimates to the extent it exists, it appears that the lim-
to using a within-firm 3SLS estimator. The 3SLS ited nature of the exporting variable does not
results were entirely consistent with the 2SLS fixed lead to qualitatively different interpretations of our
effect results. The benefit of 3SLS is that the esti- results.
mates are consistent and asymptotically efficient (a Another approach we considered was to ignore
property that the 2SLS estimator does not have). the simultaneity of the dependent variables and
However, a trade-off with 3SLS vs. 2SLS is that estimate a panel data (i.e., random effect) Tobit
if one equation is misspecified then the entire sys- model for export sales. We did not choose this
tem is misspecified. Given this trade-off we chose approach because (a) the simultaneity of domestic
to present the 2SLS results. Results from any of and export sales is a key element of our paper
the specifications that we describe in this Appendix and (b) the Hausman test in the 2SLS specification
are available upon request. indicates that the unobserved effects correlate with
the regressors.
Limited dependent variables
Persistence in exporting
As previously mentioned, a focal element of this
paper is to account for the simultaneity of domestic In order to assess the potential bias induced by
and export sales. Because the dependent variables an exchange rate hysteresis effect, we estimated
Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)
 Interrelationship between Export and Domestic Sales 871

a dynamic panel model of export sales using the the 2SLS estimates that we present in the paper.
Arellano and Bond (1991) estimator. The concern With regard to export persistence, we found that
is that our results might be misleading should there lagged export sales were positively related to cur-
exist persistence in export sales. Unfortunately, the rent export sales for the foreign firms but nega-
Arellano and Bond (1991) estimator does not allow tively related to contemporaneous export sales for
us to directly assess the simultaneity of export the Spanish firms. Our interpretation is that hys-
and domestic sales; however, we include domestic teresis does not play a prominent role because
sales as an explanatory variable in the export sales persistence in export volume was not prevalent
equation. for domestic firms. We interpret these results to
With respect to the sign of coefficient estimates suggest that our findings would not change sub-
and level of statistical significance, the results from stantially if we were to more explicitly control for
this estimator were consistent with those from the possibility of hysteresis in the data.

Copyright  2005 John Wiley & Sons, Ltd. Strat. Mgmt. J., 26: 855–871 (2005)