Learning from foreign investment by rival firms: Theory and evidence

https://doi.org/10.1016/j.ijindorg.2007.12.002Get rights and content

Abstract

We offer an alternative explanation for follow-the-leader behavior in foreign investment decisions based on Bayesian learning by rival firms. We test the implications of the model through a panel count data sample of MNEs that have invested in Central and Eastern Europe over the period 1990–1997. Interacting the measure of rivals' investment in country-industry pairs with uncertainty, we are able to identify the channel of Bayesian learning about revenue postulated by the model as the only one consistently generating the detected follow-the-leader behavior of foreign investments. The empirical findings are robust with respect to different model specifications.

Introduction

In the literature on foreign direct investment (FDI) it is well established the idea that foreign entry by a firm may trigger a rival reaction, leading to a follow-the-leader (FTL) behavior in foreign investment decisions: firms (the followers) invest abroad as a reaction to the set up of a foreign affiliate by a first-mover competitor (the leader). A possible rationale of such a behavior has been originally discussed in the business literature by Knickerbocker (1973) and Flowers (1976), and it is known as, ‘oligopolistic reaction’1: the intuition is that firms, uncertain of production costs in the country to which they currently export, run the risk of being underpriced by a rival that switches from exporting to establishing a manufacturing subsidiary in the host country. By imitating the behavior of the lead investor, the follower firms can instead match the production cost of the rival firm abroad and thus avoid being underpriced.

This paper tests for the presence of learning to rationalize the follow-the-leader behavior observed in the patterns of foreign direct investment in the market. To the best of our knowledge, this is the first paper proposing a learning mechanism to rationalize the observed FTL behavior of multinational enterprises (MNEs) and testing for it in the data.2 In our model, learning can be either about costs (as in Vettas, 2000) or about revenues. Moreover, in line with a recent literature on firm heterogeneity, the assumption we make on the prior distribution of the variables of concern implies a Pareto distribution for the observations through which learning takes place.

In general, follow-the-leader behavior in FDI decisions is supported by broad empirical evidence. Controlling for variables relevant for the decision to undertake FDI (e.g. the market size of the host country and the distance from the investor's home to its host country), Yu and Ito (1988) consider FTL behavior in two industries, the US tire and textiles. By finding follow-the-leader FDI only in the tire industry, they conclude that firms only react oligopolistically in moderately concentrated industries such as the tire one and not in more competitively structured industries as textiles. More in general, by examining all Japanese investment into the U.S., Hennart and Park (1994) find evidence that FDI by a Japanese enterprise group in the U.S. is more likely if other Japanese rivals have already invested in the U.S.; Ito and Rose (2002) show that, in the same tire industry, firms like Continental and Bridgestone imitate FDI decisions by leading firms like Goodyear and Michelin, with follow-the-leader behavior measured as the impact of the total number of foreign firms (regardless of when they entered) on the probability of investment by another foreign firm in a given year.

Though these results provide compelling evidence for the phenomenon and have a straight-forward economic interpretation, they however fail to identify the theoretical channels through which the reaction of rival multinational enterprises can arise. Moreover, all the previously quoted studies are based on the study of only one or two industries, while a broader analysis encompassing the relation between industry-specific characteristics and rivals' reaction is lacking.

In our paper, we link the finding of our theoretical model with the recent flow of FDI to Central and Eastern European Countries (CEECs). The sample is chosen since it provides an interesting ‘natural experiment’: first, the existence of a learning effect seems plausible after the fall of the iron curtain, as many firms considered investing in Eastern Europe because of the expected lower marginal cost and/or possible new market opportunities in the region. Second, the fall of the Berlin wall in 1989 enables us to monitor over time the number of foreign investments taking place in CEECs and the follow-up behavior by rivals, thus controlling for initial conditions. In particular, it is possible to exclude the effect of learning from domestic firms, since these companies were either non-existing or subject to a heavy restructuring process in the early years of transition. Our sample therefore consists of the yearly number of European Union's foreign investors over the period 1990–1997, over a large set of industries and the most important CEECs.3 By identifying the order of entry from the very first investor to late investors, and using a panel negative binomial regression model relating foreign investment in a given year, industry and country to changes in the total number of investors operating in the same industry and country in the previous year,4 we are able to explicitly test for a foreign firm's reaction to other firms'entry. In addition we test for information spillovers from foreign investors in the previous year in the same industry but in other countries.

An additional advantage of our approach is related to the comparison of alternative channels put forward by the literature for explaining the FTL behavior. Head, Mayer and Ries (2002) formalize the original Knickerbocker (1973) rationale within a Cournot-type model where there is cost uncertainty and a certain minimal degree of risk aversion by firms. Though their model is elegant, the derivation of their main result pays however tribute to strong restrictions on the parameters'space and on the underlying assumptions of the model (e.g. equal slopes of the demand curves in both the home and the host country). In addition, if there is not enough uncertainty and/or firms are not sufficiently risk averse, FDI decisions turn out to be, in their framework, strategic substitutes rather than strategic complements.5

Leahy and Pavelin (2003) provide a simpler theoretical explanation for follow-the-leader behavior in FDI decisions. In their model domestic rivals may be motivated to imitate the leader's FDI when this facilitates collusive behavior in the foreign market. However, since FTL foreign investment only hinges on the possibility to collude, neither uncertainty nor risk aversion play a role in driving their main result; moreover, their framework is also exclusively based on the existence of an oligopolistic market structure.

An alternative explanation for follow-the-leader FDI can be inferred within the theories of economic geography. FTL foreign investment is in fact consistent with a pattern of FDI that is spatially agglomerated: if the trade-off between competition effects and agglomeration forces is solved in favor of the latter, it becomes profitable to follow abroad the leader investor.6 Thus, the agglomeration channel for FTL foreign investment predicts that the latter is more likely the higher the number of early investors, in contrast with the findings of the previously quoted papers, where an oligopolistic market structure is crucial for the generation of a follow-the-leader behavior. In particular, Crozet et al. (2004) perform an analysis of FDI in French regions, showing that the location of new entrants is positively and significantly infuenced by the proximity of other MNEs of the same nationality.7 Analogously, Buch et al. (2005) in their study of German FDI determinants find positive agglomeration effects working through the number of other German firms that are active in a given host country.

Finally, a follow-the-leader pattern in the undertaking of foreign direct investment might be exogenously generated by firm or industry-specific characteristics: firms (thus with no leader-follower relation) would commonly observe a signal, unobserved by the econometrician, that reduces their uncertainty, but some of them might be more efficient in reacting to this signal and exploit first-mover advantages, with the ‘losers’ bunching behind them in terms of investment timing. In a similar way, time-to build heterogeneity might generate an exogenous sequential pattern in FDI inflows.

Our framework is able to encompass the main implications of all these alternative FTL models. Therefore, the exercise allows us to precisely identify the channel which, among the possible alternative explanations, is more consistent with the detected pattern of sequential investment by MNEs. To this extent, our results show that, alongside more traditional determinants of FDI, follow-the-leader behavior driven by our proposed channel of Bayesian learning by rival firms plays a significant role in driving MNEs' decisions to invest abroad. More specifically, the results indicate that firms learn about revenue rather than cost. This result is robust with respect to different model specifications which control for both industry and country heterogeneity.

The paper is structured as follows. Section 2 presents a simple model of rival MNEs' reaction through Bayesian learning, whose implications are tested through the econometric approach presented in Section 3 against possible alternative channels driving FTL foreign investment. The results are discussed in Section 4, while Section 5 extends the empirical approach by considering learning from investments in other CEECs than the host country. Finally, Section 6 concludes.

Section snippets

FDI and Bayesian learning

Following related studies on herd behavior and informational cascades (e.g. Banerjee, 1992, Bikhchandani et al., 1998), we assume that the expected payoff from adopting or rejecting an action, i.e. investing abroad in this case, depends on observable signals, e.g. the cost of production that a rival pays in the host country. The undertaking of a FDI by a rival firm resolves some uncertainty surrounding the profitability of the investing firm in the host country. This information then becomes

Econometric approach

In order to avoid a simultaneity bias, we have lagged all the covariates (not only the investment dummies) one year.19 Note that by using categorical dummies for

Results

Table 1 shows the results for the pooled specifications of the estimation (i.e. Models 1 to 4 in the previous scheme). The first set of control variables have the expected sign and are overall significant. More specifically, when considering the unconditional regression excluding the investor dummies and their interactions (Model 1), traditional FDI determinants measuring population size and GDP per capita are positive. These results show that horizontal (market-seeking) investment explains a

An extension: country-specific heterogeneity

In Table 2 we have reported the results for model specifications in which the panel group structure of the data was considered only across industries i, ignoring possible biases arising from country heterogeneity. A possible additional set of dummies convenient for our analysis considers the fact that information about production costs may also be revealed by firms that invest in the same industry i, but in a country s  j. Investment dummies CLijt, CMijt, CHijt, CV Hijt therefore indicate,

Conclusion

Paying tribute to the original intuition by Knickerbocker (1973), we have been able to derive a general theoretical model of rival MNEs' reaction based on Bayesian learning from first mover investors, encompassing the main implications of alternative models developed by the literature, and testing the resulting propositions on the actual behavior of rival MNEs. We find evidence for our theoretical propositions, showing that, alongside more traditional determinants of FDI, follow-the-leader

Acknowledgements

We wish to thank the Editor (Ali Hortacsu), an anonymous referee and seminar participants at the 5th annual meeting of the European Trade Study Group in Madrid, Erasmus University in Rotterdam and the 31st EARIE meeting in Berlin for helpful comments and suggestions. The usual disclaimer applies. This paper has been written as part of a CEPR research network on ‘The Economic Geography of Europe: Measurement, Testing and Policy Simulations’, funded by the EC under the Research Training Network

References (29)

  • S.J. Chang

    International expansion strategy of japanese firms: capability building through sequential entry

    Academy of Management Journal

    (1995)
  • A. Cukierman

    The effects of uncertainty on investment under risk neutrality with endogenous information

    Journal of Political Economy

    (1980)
  • S. Davies et al.

    Industrial Organization in the European Union

    (1996)
  • E.B. Flowers

    Oligopolistic reactions in eastern and Canadian direct investment in the United States

    Journal of International Business Studies

    (1976)
  • View full text