Patterns of migration, trade and foreign direct investment across ... - degit

12th Dynamics, Economic Growth and International Trade Conference University of Melbourne, Australia, June 29-30, 2007

Patterns of migration, trade and foreign direct investment across OECD countries Ben Dolman∗ Productivity Commission Canberra PRELIMINARY DRAFT: NOT FOR QUOTATION Abstract: Previous studies have shown that countries trade and invest more with partner countries from which they have received more migrants, presumably because migrant networks provide information on financial opportunities abroad. This literature focusing on migrants within individual countries is extended by constructing a dataset covering 28 OECD countries and up to 162 trading partners and the results confirm these bilateral correlations. The effect of migrants on trade flows is also found to be smaller where countries exchange greater direct investment, suggesting that more formal business networks partly displace the effects of migrant networks. The data set also allows analysis of whether migrants increase the aggregate trade and investment of their country-of-residence, a question not previously addressed in the literature. This paper shows that migrants have a much larger effect on the directions of international trade and investment than on aggregate volumes. In the case of trade, this is shown to be consistent with the theory underpinning the gravity equation.

∗ PO Box 80, Belconnen ACT 2616, Australia. Email: [email protected]. This paper is a preliminary exposition of results and comments, criticism and suggestions for extensions are encouraged. A quotable version of this paper may subsequently be released on the Productivity Commission website. The views expressed are those of the author and do not necessarily reflect those of the Productivity Commission.

1

Introduction and summary

The past few decades have seen rapid growth in the movement of goods and factors of production. The volume of international trade grew twice as fast as world output during the 1990s and the volume of foreign direct investment (FDI) grew twice as fast as trade. Economies are rapidly integrating and becoming more closely dependent upon each other. The international movement of people is an important feature of this integrated global economy. The declining cost of travel and communications has lowered information barriers and encouraged the circulation of people across national borders. Australia has been strongly participating in this process. Among a total population of around 20 million, more than 4 million Australians were born abroad. Australia has the largest share of foreign-born residents in the OECD. At the same time, around 350,000 Australian-born people reside overseas in other OECD countries and all up 1 million Australians are abroad on any given day. An assessment of the costs and benefits of migration to the currently resident Australian population is complex and difficult. Migration has played a pervasive role in shaping Australia’s culture, society and economy. The role of migration in meeting the long-term challenges presented by Australia’s ageing population has been the subject of public debate and the Productivity Commission (2006) provides a broad analysis, focusing particularly on the labour market performance of migrants and the macroeconomic effects of capital dilution. Other recent public debate has focused on the role that migrants may play in meeting the short-term challenges presented by tight labour market conditions and, perhaps, skill shortages in parts of the Australian economy. While migration brings many costs and benefits, the current paper has a narrow focus. Among their many influences on the Australian economy, Australia’s migrants and expatriates might facilitate the development of international social and business networks that ease the flow of merchandise across Australian docks and help to identify investment opportunities abroad. Recent discussion of Australia’s ‘diaspora’ has particularly focused on this role: Expatriates can contribute to their home country by influencing trade, investment and philanthropic flows, connecting local organisations to international developments and opportunities, and projecting a contemporary national image. … Some of these benefits are already flowing to Australia. A logical approach for our country, which is small in INTRODUCTION AND SUMMARY

1

population and physically isolated, is to try to capture more of these benefits … by engaging more comprehensively with our diaspora. (Fullilove and Flutter 2004, p. viii)

If migrants were to strengthen these networks and this had the effect of boosting the volume of Australian trade and investment then there are potential benefits to Australian wellbeing. Overcoming the information barriers to Australia’s international trade can improve the global allocation of resources and, for a small economy like Australia’s, intensify competition with domestic manufacturers that, in the long-run, can raise their productivity and improve prices for consumers. Trade and foreign investment also facilitate international technology diffusion which, in almost all countries, is the main source of long-term improvement in total factor productivity. Separate literatures suggest that countries that are more open to trade and foreign investment achieve higher productivity. This suggests the policy-relevant question is whether migrant networks increase volumes of international trade and investment. The Productivity Commission (2006) argued that there was potential for migrants to play a role in facilitating trade, but drew attention to the alternative sources of information that are becoming increasingly readily available, such as from international business consultants. The current paper attempts to quantify the relationship between migrants, trade and investment flows by examining patterns in a cross-section of OECD countries. The paper extends the existing literature by considering the question of whether migrants increase the volume, or simply change the direction, of trade and investment.

1.1

What the paper does and says

There are growing literatures looking at the relationship between trade and migration and between investment and migration. This paper replicates the robust qualitative findings in those literatures that countries trade and invest more with partner countries with which they have received more migrants. However, whereas previous studies have focused on migrants within a single country, this is a study of migration, trade and investment between many countries. The paper brings together a wide range of data on bilateral migrant populations, trade flows, investment positions, GDP, corruption and geography for 28 OECD countries around the year 2000 and investigates the patterns that should be in the data if migrant networks do reduce information barriers between countries. The additional data allows more robust analysis than in some previous papers and allows more interesting questions to be addressed.

2

12TH DYNAMICS, ECONOMIC GROWTH AND INTERNATIONAL TRADE CONFERENCE

The paper proceeds as follows. The next chapter outlines the manner in which migrants might affect trade and investment. The most plausible theoretical explanation is that migrants strengthen international business and social networks thereby improving the flow of information between countries on profitable trading opportunities. This suggests certain patterns should be apparent in the data. For example, migrants should be expected to have larger effects on trade and investment between pairs of countries separated by large information barriers (such as distance or language) or for trade in goods for which information is more important because their quality is in doubt. The existing literature finds limited evidence of these patterns. Chapter 3 sketches the empirical approach to be employed. The gravity model, which links trade between countries to the product of their GDPs and to the distance between them, has been the workhorse of empirical trade analysis and, more recently, has been successful in modelling foreign investment. However, one implication of the model is that if the distance between two countries affects their trade and investment then the distance of each country from other trading partners should also play a role. In estimating the effects of migrant numbers on bilateral trade and investment between migrants’ countries-of-residence and countries-of-birth, allowance is made for these differences in ‘remoteness’ by including dummy variables for each trading partner. A similar approach has previously been used to study patterns of international trade of US states or Canadian provinces, but the current paper provides a more international perspective. Where the paper extends the existing literature is in attempting to also answer the question: does increasing the total number of migrants resident in a country increase that country’s total international trade flows and foreign investment? In the case of trade flows, the theory underpinning the gravity-trade equation (due to Anderson and van Wincoop 2003) is applied to calculate the remoteness measures directly using a Taylor approximation (due to Baier and Bergstrand 2006). These calculations are used in two ways. First, controlling for geographic remoteness in this manner allows estimation of the effect of the total number of migrants from all countries-of-birth on the trade flows of their country-of-residence. Second, the same approximation technique can be used to calculate the theoretically expected effects on trade of increasing the total number of migrants from all countries-of-birth. The treatment of foreign investment is more restricted due to the absence of an estimable theoretical model. Two approaches are taken to estimate the effect of the total number of migrants on foreign investment positions, one minimally extending the gravity-type framework and the second applying the knowledge capital model of foreign investment. INTRODUCTION AND SUMMARY

3

Chapter 4 presents the results of analysis of patterns of trade flows. Simple regression analysis suggests that raising the number of migrants by 1 per cent increases exports from their country-of-residence to their country-of-birth by 0.145 per cent and increases imports from their country-of-birth by 0.127 per cent. These effects are significantly larger where migrants hold tertiary qualifications, but only when these migrants were born in other OECD countries. Closer investigation of the role played by migrants is undertaken by interacting the number of migrants with indicators of the strength of information barriers between countries, with limited success. However, there is some evidence that migrants have a smaller effect on trade where more formal business networks are stronger (proxied by foreign investment). Further results suggest that the effect of migrants is mainly to change the pattern of trading partners. That is, the effect of migrants on the total volume of international trade of their country-of-residence is much smaller than the effect on bilateral trade with their country-of-birth. This is consistent with the implications of the theory underpinning the gravity-trade equation, which suggests that the effects of migrants on the total volume of trade should be at least an order of magnitude smaller than their effect on bilateral trade. Analysis of patterns of aggregate trade reinforces the finding that countries with more migrants, as a share of their populations, do not have much higher openness to trade (defined as the ratio of total merchandise trade to GDP), though decades earlier this may not have been the case. Chapter 5 presents the results of analysis of foreign investment positions. Simple regression analysis suggests that raising the number of migrants by 1 per cent increases investment from their country-of-residence to their country-of-birth by 0.183 per cent and increases investment from their country-of-birth to their country-of-residence by 0.225 per cent. These effects are somewhat smaller than those estimated in the existing literature for investment out of the United States, but are significantly larger than the effects on trade (perhaps reflecting the greater informational requirements for long-term investment compared to trading relationships). The effects are found to be larger where migrants hold tertiary qualifications, regardless of whether they were born in OECD countries. However, there is little evidence that migrants play a larger role between countries separated by larger information barriers. Analysis of the effects of migrants on the total volumes of investment out of their country-of-residence is less conclusive than was the analysis of trade in the previous chapter. Estimates from one set of models — based upon minimally extending the gravity-type framework used to estimate the effects on bilateral investment — suggest that migrants change the pattern of countries with which their country-of-residence invests more than they change the total volume of investment. 4


Estimates from models seeking to implement the knowledge capital theory of foreign investment are more mixed, but should perhaps be treated sceptically since the data do not appear to fit that theory very well. The paper’s tentative conclusion (chapter 6) is that the consequences of immigration and emigration for productivity and living standards through their effects on international business and social networks are likely more nuanced than they may have previously appeared. The results confirm that increasing the number of migrants will tend to increase bilateral trade and investment between their country-of-residence and country-of-birth. However, the effects of evenly increasing the number of migrants from all countries-of-birth on the aggregate volumes of trade and investment of their country-of-residence appear to be significantly smaller. That is, migrants appear to have a much larger effect on the directions of international trade and investment than on aggregate volumes.

INTRODUCTION AND SUMMARY

5

6


2

Possible effects of migration

This chapter outlines the manner in which migration may be expected to affect trade and investment flows based on theory and previous empirical analysis. There is evidence that migrants affect the flows of trade and investment between their country-of-residence and country-of-birth. But the magnitude of this effect is unclear, as is its dependence on social and business networks. The chapter concludes by summarising the recent literature on links between trade and investment flows, and productivity and incomes.

2.1

How may migrants affect trade and investment?

Migrants may facilitate the development of social and business networks that improve the quality of information flowing between countries and encourage trade and investment flows. At a macroeconomic level, two well-known ‘puzzles’ hint at the important role of social networks in international economic relations. The first puzzle is that distance appears to matter too much, both for trade and for investment. An expansive literature finds that a 1 per cent increase in the distance separating countries reduces trade between the two countries by around 0.9 per cent (Disdier and Head 2007). In terms of transport costs alone, this seems an unreasonably large effect. If transport costs were typically 5 per cent of the value of traded goods then Grossman (1998) calculates that a 1 per cent increase in the distance separating countries should lower trade volumes by only 0.03 per cent. For investment patterns, the role of distance is even more profoundly puzzling. Early theories explained investment in terms of differences in factor endowments which lead to vertical investment (Helpman 1984) or transport costs savings of production close to consumer markets which leads to horizontal investment (Markusen 1984). While distance hampers vertical integration of plants across countries, the larger transport cost savings should encourage horizontal investment and in practice horizontal FDI appears to be the predominant type.1. This suggests 1 For example, two-thirds of US affiliate sales are within the host country, while only 10 per cent are sales back to the source country (Blonigen 2005). Sales to countries near to the host (exportplatform investment) are also significant, as are sales between affiliates in different countries (suggesting more complex production chains). POSSIBLE EFFECTS OF MIGRATION AND THEIR CONSEQUENCES

7

that foreign investment might be expected to increase with distance between the source and host countries, yet in practice FDI falls away rapidly.2 The second puzzle is commonly called the ‘border puzzle’. Trade between countries is a small fraction of trade within countries. Even after accounting for distance and associated transport costs, trade between countries is 20 to 50 per cent lower than trade within countries (Anderson and van Wincoop 2003).3 Similarly, foreign investment is a small fraction of domestic investment. The most plausible explanation for both puzzles is that information on foreign trading and investment opportunities, and the associated legal and regulatory environments, is both scarce and expensive to gather, and that search costs increase with distance and national borders (Rauch 1999, 2001). Migrant communities may offer several way of improving information flows by strengthening business and social networks between the migrant’s country-of-birth and country-of-residence. Migrants are well placed to act as middlemen on account of their superior language skills, and their knowledge of consumer preferences, business practices, market structure and laws. Migrant business networks may also help with contract enforcement in countries where foreign business people have difficulty enforcing contracts. These information barriers may be larger for FDI than for international trade. Foreign investors may engage in more complex negotiations with a wider range of people (suppliers, workers, government officials) than traders, and require more detailed knowledge on local labour markets, legal and regulatory environments. At the same time, the risks involved with experimenting to gain information are potentially also larger. While trade can involve low fixed costs (establishing distribution networks) but large variable costs (transport), direct investment often requires a large upfront commitment of resources (plant establishment), partially sunk. In sum, the arguments made for explaining why migrants might affect trade appear to apply with greater strength for investment. 2 The same puzzle is evident for other transactions. Portfolio investment falls rapidly with distance despite distance providing an apparent incentive for investment through reduced correlation between business cycles and, hence, potential for diversification (Portes and Rey 2005). Even on the internet, where ‘transport costs’ are nil, people disproportionately visit websites from nearby countries, particularly for cultural products (such as music) and financial transactions (Blum and Goldfarb 2005). 3 Based on apparent effects on volumes of trade, Anderson and van Wincoop (2004) suggest that total trade costs between developed economies are typically around 170 per cent of production costs, composed of: 12 per cent transport charges; 9 per cent time in transport; 44 per cent border-related trade barriers; and 55 per cent retail and wholesale distribution costs. That is, border-related effects are more than 3 times as important as actual transport charges. 8


The international importance of migrant networks is often highlighted by Indian and Chinese case-studies. Ethnic Indians working in Silicon Valley have, both by improving information networks and establishing the reputation of Indian ICT workers, led to the creation of a large and growing ICT service export industry in India (Saxenian 2002). Expatriate ethnic-Chinese business people have played a large role in directing foreign investment into the Chinese manufacturing industry (Weidenbaum and Hughes 1996). The potential importance of migrant networks for Australia has been highlighted by surveys.4 For example, companies that successfully export to East Asia are three to four times as likely as other businesses to employ staff of East Asian descent (Dawkins et al. 1995). And Australian expatriates may also play a role. More than half of Australian expatriates surveyed by Hugo et al (2003) believed that they have created goodwill towards Australia during their time abroad (though only one in five believe they have established business and trade links). These empirical puzzles, case studies and surveys together suggest that migrants may affect trade and investment through improving information flows. To date no formal theory models this relationship so the analysis in subsequent chapters seeks to identify correlations rather than structural relationships. Nevertheless, if migrants increase trade and investment flows by overcoming information barriers then certain patterns of correlations should emerge in the data. First, migrants should be associated with greater trade between their country-of-birth and their country-of-residence in both directions. If, for example, migrants were associated with only higher imports into their country-of-residence, but not exports, then this might indicate simply that they are importing goods for personal consumption, because their tastes favour goods produced in their country-of-birth. Second, migrants should be more strongly associated with trade between pairs of countries for which alternative business and social networks are weaker. For example, migrant networks may play a larger role between countries that do not share common languages or common colonial ties or well-developed business networks or simply between countries that are far apart. Third, migrants should increase more strongly trade in goods for which information is more valuable because quality varies significantly between suppliers. Fourth, migrant

4 Case studies are corroborative. As an example, Diversity Australia, an Australian Government Initiative, tells the story of how Gateway Pharmaceuticals was established to export pharmaceuticals to Vietnam after a retail pharmacist in Cabramatta, a suburb of Sydney with a large ethnic Vietnamese population, noticed that people from his local community were purchasing pharmaceuticals to send back to their relatives. The business has succeeded in part by making use of migrant employees’ language skills and knowledge of consumer preferences and ways of doing business in Vietnam. POSSIBLE EFFECTS OF MIGRATION AND THEIR CONSEQUENCES

9

networks may play a larger role where one of the trading partners has poorly developed legal systems so that contract enforcement is difficult. Finally, personal characteristics of migrants, such as their educational attainment, may affect the extent of their market knowledge and connections with local business networks. Similarly, if these arguments are correct then migrants should be associated with greater investment flows between their country-of-birth and their country-of-residence, and this effect should be larger than for trade because of the larger informational requirements. Further, the effects of migrants should be larger between countries for which alternative networks are weaker or, perhaps, where migrants are better educated. This paper explores whether these patterns of trade (chapter 4) and investment (chapter 5) are in fact apparent.

2.2

Previous literature

Previous studies of the effects of migration on trade flows and investment positions are few, but have grown in number in recent years. These studies have taken two analytical approaches. The first approach is to compare the number of migrants living within a particular country from each country-of-birth with bilateral trade or investment. The second approach is to compare the number of migrants from a particular country living in a number of other countries with trade or investment flows between those other countries. Rauch and Trindade (1999) and Tong (2005) take the second approach and analyse the effects of Chinese expatriate populations on trade flows and FDI positions respectively. All other studies to date have taken the first approach. The trade literature has generally found that larger numbers of migrants are strongly associated with larger flows of goods, but estimates of the size of these effects vary widely (table 2.1). At one end of the range, Gould’s (1994) results imply that doubling the number of migrants in the United States from a given country would increase trade with that country by only 1 or 2 percentage points. At the other end, Dunlevy’s (2006) results imply that this would increase trade by almost one-half. The literature linking FDI and migration is scarcer and more recent (table 2.2). Two studies of foreign investment out of the United States (Javorcik et al. 2006 and Battacharya and Groznik 2005) suggest an elasticity of FDI to migrant numbers of around 0.3. The similarity between these results suggests some robustness in the relationship, because they were derived separately from cross-sectional and time-series analyses. However, it is not yet clear how these results should be 10


expected to carry over to other countries. The only study of the effects of migrants residing outside the United States, Buch et al. (2003), found widely varying results depending upon how their migrant population data were constructed. Empirical exploration of the manner in which migrants may affect trade and investment has been somewhat limited. Most simply, many studies test whether migrants affect trade by their preference for consuming goods from their country-of-birth by comparing elasticities of imports and exports to migrant numbers. Of the 9 studies comparing import and export flows (table 2.1), four studies find imports more responsive to immigrant numbers, four studies find exports more responsive, while Helliwell (1997) reports mixed results depending upon whether the trade flows are international or intra-national. That is, the literature to date does not provide consistent evidence that migrants affect imports differently than exports. The literature has also explored whether migrant networks play a larger role in trade and investment flows where information barriers may be larger, with mixed success. The literature provides some evidence that migrants play a larger role where trading partners do not share a common language (Dunlevy 2006) or colonial ties (Girma and Yu 2000). There is also some evidence that migrants have a larger effect on trade in ‘differentiated goods’ for which information is more valuable because quality may vary significantly between producers (Rauch and Trindade 1999, Gould 1994, Bryant, Genç and Law 2004, Herander and Saavedra 2004). Finally, migrants may play a larger role in facilitating both trade and investment between countries with weaker institutions (Herander and Saavedra 2004, Dunlevy 2006, Tong 2006). The focus in the literature has been on estimating the relationship with bilateral trade and investment. No studies to date have explored the effects that migrants may have on the total trade flows or total investment positions of their countries-of-residence.

POSSIBLE EFFECTS OF MIGRATION AND THEIR CONSEQUENCES

11

-0.08 0.50***

UK trade with 48 countries, 1981-93

12

Larger effects for consumer goods. Little difference due to migrants’ education or length of stay.

Comments

0.25*** 0.03

0.08* 0.23***

Trade between 5 Canadian regions and 160 countries, 1992-95 Trade between Spain and 40 countries, 1991-98.

0.05 0.13***

0.22*** 0.22***

Trade between 95 French Department, 1978 and 1993

0.06 0.19

1978 1993

Commonwealth Non-Commonwealth

Between Canadian provinces Between Canadian provinces and US states 0.31*** Larger effects for independent migrants (mostly professionals), c.f. family, refugee, entrepreneurs. 0.29***

0.12 0.06

0.01***

Import elasticity


Combes, Lafourcade & Mayer (2002) Wagner, Head & Ries (2002) Blanes-Cristobal (2004)

0.08**

US trade with 17 countries, 1870-1910

Dunlevy & Hutchinson (1999), Hutchinson & Dunlevy (2001) Girma & Yu (2000)b

0.10***

Canadian trade with 136 countries, 1980-92

Head & Ries (1998)

0.03 0.34***

0.02***

Trade between Canadian provinces and US states, 1990

US trade with 47 countries, 19701986

Export elasticity

Helliwell (1997)

Gould (1994)a

Data

Previous literature on trade and migration

Bilateral migrants

Authors

Table 2.1

Country (but not province) OECD and EU

None

None

None

Regional

None

Country

Fixed effects

(continued on next page)

None

None

Remoteness (distance to world GDP); GDP per capita None

Relative price indices; population

Relative price indices; openness

None

Relative price indices; population

Remoteness variables used

Data

Comments

Smaller effects with ancestry, larger effects with poorer institutions, larger effects for producer goods than consumer goods.

.

0.21 0.47***

0.21 0.47***

na

Bilateral trade in homogeneous goods Bilateral trade in differentiated goods

Larger effect for exports to countries with corrupt political systems, smaller effect with Spanish or English speaking countries.

na Migrant links to trade appear due to a small number of ethnic groups.

na

na

na

0.15*** Larger effects excluding agricultural exports and oil imports.

Import elasticity

POSSIBLE EFFECTS OF MIGRATION AND THEIR CONSEQUENCES

13

Fixed effects

Remoteness (distance from world GDP); GDP per capita

None

Population

Trade openness (exports plus imports/GDP); population Remoteness (distance from world GDP); openness

EEC and EFTA

State and country

Trading-pair

None

None

None, but correlated random country effects Population Asia-Pacific. West coast and east coast estimated separately.

Population


a Elasticities are calculated by Wagner, Head and Reis (2002). b Elasticities are from the regression specification including a lagged dependent variable.

Ethnic Chinese networks Trade between 63 countries, Rauch & Trindade a 1980 and 1990 (1999)

Bandyopadhyay, Coughlin & Wall (2006) Dunlevy (2006)

Exports from 51 US states (and DC) to 36 countries, 1993-1996

Herander & Saavedra (2005)

0.239*** (West coast) 0.059 (East coast) 0.300***

0.09***

Export elasticity

0.162*** (migrants in state) 0.073*** (migrants in other states) Exports from 51 US states 0.132*** (and DC) to 29 countries (1988-92 and 1998-2002) Exports from 51 US states 0.39*** (and DC) to 87 foreign countries at a single point in time (average 1990-92)

Exports from 51 US states (and DC) to 28 countries, 1993

Trade between New Zealand and more than 170 countries, 1981-2001 Exports from East and West Coast of the United States to 51 countries, 1994-96

(continued)

Co, Euzent & Martin (2004)

Bardhan & Guhathakurta (2004)

Bryant, Genç & Law (2004)

Authors

Table 2.1

Higher elasticities for investment from developed countries and to countries with high bureaucratic quality

Dependent variable is growth in foreign investment between 1990 and 2000; elasticities are reported for migrants with different educational attainment as indicated.

Higher elasticity to tertiary educated migrants (0.370**).

Cross-sectional bi-variate regression

Migrant numbers are based on flows since 1974, aggregated into stocks, and the results differ depending on whether gross migration or net migration (including return flows) is used. Panel multivariate regression

Comments

Population; remoteness (distance from world GDP)

Not applicable

Population

None

Openness

Population


b

None

Regional

None

None

Trading-pair

None

Fixed effects

14


a Regressions are undertaken in levels not logarithms and the elasticities reported here are the current author’s estimates based on data for the means of the 37-country sample in 2000. b This study does not employ country fixed-effects in the study of aggregate foreign investment positions, but does in a separate analysis of sectoral investment. c These elasticities are for the effect of 1990 migrant numbers on growth in FDI over the following decade.

0.19***

na

0.21*** (primary) -0.56 (secondary) 0.43* (tertiary) 0.19***

na

na

0.42** 0.284*

na

0.33***

US direct investment into 33 countries 1970, 1980, 1990, 2000 US direct investment into 37 countries, 2000 FDI outward position from the US to 56 countries, 1990 and 2000 US FDI outflows, 1990 and 2000

Ethnic Chinese networks Tong (2005) FDI positions between 70 countries around 1990

Javorcik, Özden, Spatareanu & Neagu (2006) Kugler & Rapoport c (2007)

Bhattacharya & a Groznik (2005)

0.06** (net stocks) -0.10 (gross stocks)

Inward FDI

0.01 (net stocks) 0.48*** (gross stocks)

Outward FDI

FDI outward position of 16 German states, 1990-2000

Data

Previous literature on FDI and migration

Bilateral migrants Buch, Kleinert & Toubal (2003)

Authors

Table 2.2

2.3

Why do trade and investment matter?

Trade and foreign investment are not ends in themselves, but are of policy interest to the extent that they improve local productivity and living standards. This section outlines some of the links between the former and the latter. Trade and foreign investment might immediately raise productivity in a few obvious ways. Trade allows countries to specialise in production of those goods to which they are most suited and this specialisation may also permit economies of scale in production. Foreign investment often brings with it foreign technology, skills and managerial know-how so that foreign-owned firms tend to be more productive and pay higher wages than their domestic counterparts.5 Trade also increases competition between producers. Firms operating in more competitive markets tend to have higher productivity. This might be because competition makes inefficient production both more obvious and costly, resulting in more effective management (Winston 1993), or simply because competition drives those firms that happen to have low productivity out of business (Syverson, 2005). Trade and investment are also channels through which knowledge flows into economies.6 The accumulation of knowledge is central to modern theories of economic growth (Aghion and Howitt 1992, Romer 1990) and a large portion of the differences in productivity levels between countries and productivity growth over time appears to be due to technology (Easterly and Levine 2001, Hall and Jones 1999). The majority of technological progress within all countries — with the possible exception of the United States —occurs though absorption and adaptation of knowledge first developed overseas (Eaton and Kortum 1999). Trade may lead to technological progress in many ways. Importation of capital that embodies the latest vintage technology may raise incomes in the importing country because the benefit of these inventions is typically larger than the price foreign 5 Caves (1974) is among the earliest papers showing that foreign-owned Australian firms have higher productivity than domestically-owned firms. 6 Of course, immigrants themselves may transfer knowledge, for example where they possess scarce skills that they can teach existing residents (Stromback 1994). A significant amount of technology is not able to be written down and hence not able to be transmitted over long distances. Rather much productive knowledge is held tacitly by experts in a field and passed on largely through face-to-face interaction and demonstrations. Certainly at the level of elite scientists and inventors, it is clear that knowledge spill-overs are heavily concentrated within their local communities (Jaffe, Trajtenberg, Henderson 1993; Keller 2000) and these spill-overs tend to follow them when they migrate (Zucker and Darby 2006). But away from these elite few, this direct role of migration in international knowledge flows is poorly understood. POSSIBLE EFFECTS OF MIGRATION AND THEIR CONSEQUENCES

15

inventors can charge, due to competitive pressures and difficulties discriminating between customers. Trading relationships may, in themselves, also transfer knowledge. Exporters may learn to improve their products and production processes through interaction with more advanced foreign competitors and more demanding foreign consumers.7 Foreign investment may also play a role in transferring technology to the local economy. Precisely how knowledge spills over from foreign-owned to locally-owned firms is not clear, but candidate mechanisms include: imitation of technology, managerial and organisational practices; the spread of new skills through the movement of labour between foreign and locally-owned firms; development of export distribution networks and transport infrastructure; and the deliberate transfer of technology from foreign-owned firms to local suppliers of intermediate goods or buyers of their final products. While the theoretical links between trade and investment on the one hand and productivity and income on the other are relatively well understood, only in recent years has the weight of empirical evidence moved to support them. Countries that are more open to international trade tend to have higher incomes and faster economic growth. Of course, countries that are open to international trade tend to have very different institutions than those whose economies are more insular, such as democracy and the rule of law, which might also tend to raise their incomes. So this simple correlation has proved difficult to interpret. Does trade cause incomes to rise? Careful analysis within the past decade have suggested an answer. One strand of analyses has shown that countries that are more open to trade because they are closer to world markets have higher productivity and higher incomes (Frankel and Romer, 1999, Alcala and Ciccone, 2004, Redding and Venables, 2004). More recently, it has been shown that a country’s openness to trade rises after the United States cuts tariffs on trade with them and these countries subsequently experience stronger economic growth (Romalis 2007). That is, the evidence suggests that openness to trade causes incomes to rise.8 Identifying the role of foreign investment on productivity and incomes has proved similarly elusive, until recent years. It is clear that foreign investment increases the productivity of the foreign-owned firm, but to the extent this reflects knowledge that 7 For example, MacGarvie (2006) shows that firms are much more likely to cite patents from countries with which they trade. 8 This conclusion is not unanimously accepted. Rodriguez and Rodrik (2000) and Rigobon and Rodrik (2004) argue that studies linking trade to productivity via geography are flawed because geography affects productivity in many other ways (for example, through resource abundance and exposure to disease). 16


remains within the subsidiary the benefits are largely retained within the multi-national corporation. However, recent analysis has identified productivity increases in indigenous manufacturing firms operating in developed economies after foreign-owned firms entered their narrowly defined sub-industries (Haskel, Pereira and Slaughter 2002, Keller and Yeaple 2003). Since these effects appear strongest within high-tech industries or where the foreign-owned firm is operating close to the technological frontier, they seem to be attributable to some kind of knowledge spill-over, though this may become clearer as this literature develops.9 Trade and foreign investment may be especially important for small countries, such as Australia. One reason is that the size of domestic markets makes the trade-off between economies of scale and competition more acute. Another reason is that small countries are inevitably more reliant upon knowledge developed overseas because domestic knowledge creation makes only a small contribution to advancing the global technological frontier.10

9 Further evidence that knowledge spills over from foreign-owned firms is shown by the fact that domestically-owned US firms are more likely to cite the patents of a Japanese firm that has invested in the United States (Branstetter 2006). 10 Australia, for example, undertakes only a little more than 1 per cent of OECD R&D activity. POSSIBLE EFFECTS OF MIGRATION AND THEIR CONSEQUENCES

17

18


3

Methodology and data sources

This chapter outlines the empirical approach to be pursued. The gravity model is applied to analyse both bilateral international trade and foreign investment. Key limitations of the available data are that many observations are missing or zero, that it is difficult to distinguish between the two possibilities, and that the extent of missing data varies across countries. It appears likely that many of the zero observations are in fact small, but unreported, trade flows and investment positions — that is, the data are censored — and for investment the extent of censoring differs between reporting countries. This suggests a tailored approach to empirical estimation.

3.1

Modelling trade flows

The naïve gravity model of international trade flows simply relates the magnitude of trade between economies to the product of their economic ‘masses’ and to the distance between them, in a similar way to Newton’s model of the force between two bodies due to gravity. That is, the value of imports is given by the following equation, where Mij is the total value of imports into country j from country i, Yi and Yj are their respective gross domestic product and Tij captures the cost of moving goods from country i to country j, where originally Tij was assumed to be simply equal to the distance between the two countries. M ij = A

Y jY i (T ij )σ −1

This type of model does a fairly good job of explaining much of the variation in bilateral trade flows. It provides a useful platform for testing the importance of other determinants of trade costs by allowing the modelled trade costs to depend on more than simply distance between trading partners. The current paper particularly focuses on the role that migrants play in affecting trade costs. To estimate accurately the effects of these sorts of trade costs requires an understanding of the theory underpinning the model. In particular, the naïve version of the model omits two variables: the opportunities available for the exporting nation to trade elsewhere in the world and the price of goods available from other METHODOLOGY AND DATA SOURCES

19

trading partners. Together these terms are often referred to as the ‘multilateral resistance’ to bilateral trade or simply the ‘remoteness’ of the exporter and importer. The pattern of trade between Australia and New Zealand provides an example of the importance of multilateral resistance. Trade flows between these two countries are very large. In 2000 almost USD 6 billion in goods moved between the two countries, despite more than 2300 km separating Wellington from Sydney. For comparison, trade between Ireland and the Netherlands — a broadly similarly-sized pair of economies — totalled less than USD 5 billion in the same year, despite only 760 km separating Dublin from Amsterdam. One difference is that whereas Australia and New Zealand are remote from other potential trading partners, Ireland and the Netherlands are able to trade at low cost with many European neighbours. Bilateral trade flows The existing literature linking trade to migration provides guidance on the appropriate way to estimate the gravity equation. A variety of techniques have been used. Some studies (for example, Helliwell 1997, Blanes-Cristobal 2004) have made no attempt to control for differences in remoteness. The risk with this approach is that geography alone may produce a correlation between migration and trade if, for example, remote countries both trade more and exchange migrants more with nearby countries than they otherwise would. The most common method to address this problem involves adding to the regression variables that are correlated with each trading partner’s remoteness, such as the average distance to world GDP or the size of the local population (for example, Dunlevy and Hutchinson 1999, Hutchinson and Dunlevy 2001; Girma and Yu 2000; Co, Euzent and Martin 2004). However, the validity of this approach depends upon whether these remoteness indicators do a reasonable job of capturing all of the relevant characteristics of the trading partners. A second method to address this problem is to include dummy variables for both trading partners (for example, Wagner et al. 2002, Dunlevy 2006). The advantage of this method is that it avoids estimation bias that can arise because of any mis-specified or omitted factors associated with particular countries, including economic remoteness.11 Since in practice this second method produces significantly 11 There are, however, two often-cited advantages of the first approach. First, including importer and exporter fixed effects precludes estimation of the effects on trade of some other variables (language, contiguity, GDP). Second, the differences between countries may be important in identifying the link between migration and trade. That is, including trading-partner fixed effects may remove too much of the information from the dataset without removing enough of the noise. From a practical perspective, this is less of a problem for the current study because it uses data on 20


different empirical results (for example, Wagner et al. 2002) and is to be statistically preferred, it suggest the results from models using the first method may be biased. In studying the effects of migrants on bilateral trade, the current paper uses this trading-partner fixed effects approach. Though this approach solves problems associated with remoteness of trading partners, the regression equation may still be mis-specified because it is not feasible to include variables that capture the unique historical, cultural, political and business relationships between any pair of countries that may both affect trade and have affected historical migrant flows.12 To alleviate this problem, Pakko and Wall (2001) and Bandyopadhyay et al. (2005) advocate panel estimation including trading-pair fixed effects and Bandyopadhyay et al. (2005) show that doing so reduces the estimated elasticity of trade to migration almost by half compared to cross-sectional analysis.13 While data availability means that the current study cannot take a panel approach, an attempt is made to take account of this suggestion. The current study includes lagged trade flows in the regression equation, which may capture the effects of slow-moving, unmeasured, trading-pair-specific factors14 and also includes an indicator of alternative business networks between countries, represented by the stock of foreign investment. Total trade flows The paper also seeks to investigate the effects of migration on a country’s total trade flows, or equivalently whether increasing the total number of migrants in a country from all countries-of-birth increase that country’s trade. This means focusing on the effects on trade of characteristics of trading partners, specifically aggregates of their immigrant and expatriate populations, so that the trading-partner fixed effects migrants within 28 OECD countries, so there remains substantial variation in trade between trading pairs after allowing for trading-partner fixed effects. Much of the previous literature focussed on migrants within a single country, so that introducing trading-partner fixed effects would account for all cross-sectional variation in trade flows leaving only variation within trading pairs over time (if any) to attribute to migrant numbers. 12 Further, only the trading-pair fixed effects specification avoids estimation bias resulting from mismeasurement or omission of pair-specific variables, which could occur quite simply, for example, when distance is used as a proxy for transport costs or the migrant stock is included as the only indicator of business networks between two countries. 13 The authors interpret this as evidence that estimates derived without including trading-pair fixed effects are biased. However, the panel analysis picks up only the short-run effect of migrants on trade. Since the two cross-sections are separated by only 10 years, the average additional migrant has probably only been resident for five years at the time of the latter observation. It may be that migrants have larger effects after somewhat longer periods of residence. 14 The criticism of Bandyopadhyay et al. (2005) cautions against using the resulting regression coefficients to calculate long-run elasticities. METHODOLOGY AND DATA SOURCES

21

approach is no longer useful. Moving away from this approach while still providing a reasonable basis for unbiased estimation of the effects of migration requires the application of theory. Anderson and van Wincoop (2003) most recently pointed out that in theory only relative trade costs matter for the direction and volume of international trade.15 The total volume of trade, adding up trade both within and between countries, is mechanically determined by the size of the economies involved. If trade costs were to fall evenly for transactions within a country and for transactions between all countries then the total volume of international trade would not be affected.16 The implication for the current paper is that the effects of migrant networks on bilateral international trade flows between the country-of-residence and the country-of-birth are likely to be much larger than the effects on total international trade volumes. Strengthening migrant networks between any two countries will increase trade between those countries, at the expense of both trade within each of the two countries and trade between each of the two countries and every other country in the world. Evenly strengthening migrant networks between all countries will have a smaller effect, arising only at the expense of trade within each country. The paper applies two different Taylor approximations to Anderson and van Wincoop’s (2003) ‘multilateral resistance’ terms, as suggested by Baier and Bergstrand (2006), thereby avoiding the need for a non-linear, iterative method of estimation. Each approximation starts by assuming world trade is close to a simple case, and then making a linear correction to approximate the actual ‘multilateral resistance’ to trade. The first approximation starts from the simple case where international trade is completely free of transport costs and information barriers. This leads to estimation of each country’s multilateral resistance to trade with other world economies based on the GDP-weighted average of the indicator of trade barriers with all countries (such as distance, contiguity and migrant shares). A second approximation starts from the simple case where all economies are of equal size and face equal transport costs and information barriers and this leads to estimation of each country’s multilateral resistance to trade as a simple average of the indicator of trade barriers with all countries. The first approximation appears to fit the current data set more closely and is used except where noted otherwise.

15 A full discussion of their analysis is given in appendix A and only the scarcest detail is provided here. 16 This, for example, may explain why historical improvements in transport technology have not reduced the effects that distance has on volumes of international trade over the past half-century (Disdier and Head 2007). If costs of trade within countries have fallen at the same pace as for trade between countries then no change would be expected. 22


A further implication of the theory is that multilateral resistance to trade arising from any trade barriers should affect trade flows in precisely the same way as the bilateral barriers themselves. That is, the coefficient in the trade equation on distance, for example, should be precisely the negative of the coefficient on each country’s remoteness. To implement Baier and Bergstrand’s (2006) two approximations, two ‘multilateral’ forms of distance between trading partners are constructed based on GDP weights (wk) or equal weights (1/n) for each trading partner as in the equations below. Multilateral distance form 1: Multilateral (ln dist ) ij = ln dist ij − ∑ wk ln dist ik − ∑ wk ln dist kj + ∑∑ wk wl ln dist lk k

k

k

l

Multilateral distance form 2: 1 1 1 Multilateral (ln dist ) ij = ln dist ij − ∑ ln dist ik − ∑ ln dist kj + ∑∑ 2 ln dist lk k n k n k l n

The interpretation of these new variables can be made more concrete by regrouping terms. For example, the first form of the multilateral distance variables can be reorganised as follows. Multilateral (ln dist ) ij = (ln dist ij − −(

1 1 wk ln dist ik − ∑ wk ln dist jk ) ∑ 2 k 2 k

1 1 wk ln dist ik + ∑ wk ln dist jk − ∑∑ wk wl ln dist lk ) ∑ 2 k 2 k k l

The first term shows that if two countries are far apart compared to each country’s other potential trading partners, the variable will take a larger value. The second term shows that if two countries are far from other potential trading partners compared to the average distance between all countries in the world, then the variable will take a smaller value. Since it is presumed that these variables would take a negative coefficient in the trade equation, this simply means that countries that are far apart trade less with each other and countries that are more geographically remote trade more with each other, and the balance of these two effects can be explicitly calculated. Similar multilateral forms are constructed for other variables (contiguity, commonality of language, colonial ties and migrant shares of the population). In the analysis in chapter 5, the multilateral forms of these variables are used in two ways. The first approach is to use these variables (other than the multilateral form of the migrant shares of the population) as control variables to analyse the effect that the METHODOLOGY AND DATA SOURCES

23

total migrant stock from all countries of birth has on bilateral trade flows. The second approach is to use the multilateral form of the migrant share variable in order to calculate the effect that changes in migrant numbers from all countries should be expected to have on bilateral trade patterns.

3.2

Modelling foreign investment positions

The analysis of bilateral foreign investment positions is a straight-forward replication of the analysis for bilateral trade flows. The same gravity-model used to model trade flows has been widely and successfully applied to FDI and this is the approach that has been used in the majority of studies to date looking at the effects of migrants on foreign investment (Javorcik et al. 2006, Tong 2005, Buch et al. 2003). As for the modelling of bilateral trade flows, dummy variables for each partner country are included, thereby avoiding the need for strong theoretical underpinnings.17 Lagged investment stocks are also included in the regression as a broad indicator of historic ties between countries, informed by problems overcome in estimating the trade equation. However, the paper also seeks to explore the effects of migrants on total investment positions. This requires analysis of the effects of the total number of migrants residing within a particular country so that, again, the trading-partner fixed effects approach is not useful. At this point in the analysis of trade flows, the theory underpinning the gravity model was employed to identify which additional variables should be included in the model. However, there is no similar theory of investment underpinning the gravity-framework. This means that modelling total foreign investment positions requires a choice between empirical ease and performance on the one hand, and theoretical underpinning on the other hand. The current paper explores both approaches. The first approach is to minimally relax the restrictions on the gravity equation by replacing the dummy variables for the migrant’s country-of-residence with a few simple indicators: GDP, GDP per capita and an indicator of economic remoteness (the GDP-weighted distance to other countries, as used for the analysis of trade flows). In this approach, dummy variables for the migrants’ countries-of-birth are retained.

17 While including trading-partner fixed effects is the preferred approach in the trade literature, only one study (Bhattacharya and Groznik 2005) has previously used this approach in studying the effect of migration on investment. 24


The second approach is more theoretically motivated. Whereas the motivations for trade are relatively straightforward, the theoretical motivations for foreign investment are more complex and varied. The classic explanation for foreign investment centres around advantages associated with ownership, location and internalisation (the ‘OLI framework’, Dunning 1977). The ownership advantage occurs where a firm has exclusive access to a product or production process, including trade secrets and a reputation for quality. This explains why foreign firms exist. The location advantage explains where firms seek to produce, to exploit lower factor prices or better customer access. The internalisation advantage explains how firms choose to produce offshore, whether through direct investment or licensing. Licensing may not be desirable where there are risks that the knowledge will spread more widely through the market or that the product’s reputation may be debased. The location advantage has been the particular focus of theory. A recent series of papers (Carr et al. 2001, Markusen and Maskus 1999a, Markusen and Maskus 1999b) developed a model that incorporates incentives for both horizontal and vertical investment in what is called the ‘knowledge capital’ model of FDI. The model assumes that production requires the input of head-office services (for example, R&D, management, finance, accounting and marketing) which are: relatively skilled-labour intensive; can be geographically separated from production; and can be provided jointly as an input into production at all plants. While the knowledge capital model is not analytically tractable, the two-country simulations in those papers provide a few key predictions. For the most part, these predictions centre on characteristics of the source and host country.18 First, the volume of investment between two countries is expected to increase with the size of the economies. Second, horizontal investment occurs most often where two countries are similar in size. This is because larger and more similarly sized markets better justify the fixed costs associated with setting up a local plant in preference to the variable costs associated with otherwise large trade flows between the two countries. Third, since the head-office services are skilled-labour intensive, the model predicts that foreign investment will occur more often where there are large differences between levels of skill. Finally, the model makes predictions about the interaction of these determinants. If the skilled-labour abundant country is large then its firms will tend to produce domestically. Foreign investment is predicted to 18 The empirical analysis in this paper is undertaken in logarithms of the key variables. Whereas Carr et al. (2001) estimate their model of foreign investment in levels, Blonigen and Davies (2004) show that this results in a model that under-predicts investment into developed countries and over-predicts investment into less-developed countries, a problem partly corrected by estimating in logarithms. METHODOLOGY AND DATA SOURCES

25

be particularly prevalent from small, skilled-labour abundant countries (like Sweden, Switzerland and the Netherlands) into larger, unskilled-labour abundant countries. The second approach to explore the effects of migrants on total investment positions is to attempt to implement this knowledge capital model. The quality of institutions is also likely an important determinant of investment decisions. Poor legal protection of assets raises the risk of expropriation, poorly functioning markets increase the cost of doing business and poor infrastructure limits the productivity and export capacity of a country. In particular, previous studies have found that corruption significantly reduces inflows of investment (Wei 2000, Habib and Zurawicki 2002). The analysis includes an indicator of corruption in the host country.

3.3

Main data sources

This section outlines only the main data sources on migrant numbers, trade flows and FDI stocks. Full definitions and sources of other data are provided in appendix B. At each stage data sources have been chosen and data manipulated to optimise the available coverage of countries. A bilateral trade, investment and migration cross-section was constructed for 28 OECD countries and up to 162 partner countries (totalling 4508 possible observations) in or around the year 2000.19 The OECD Database of Immigrants and Expatriates provided data on the number of migrants in each OECD country (except Iceland). Data are generally obtained from the census closest to 2000 undertaken in each country and separately identify migrants by country-of-birth. Working-age migrants (those aged 15 years or over) are also identified by whether they have low (not completed secondary school), medium (completed secondary school) or high (tertiary) educational attainment. The data collection is close to complete, with more than 99 per cent of the counted population in OECD countries reporting a country-of-birth and more than 98 per cent of the working-age population reporting education level. Bilateral merchandise trade flows were sourced from the NBER-UN Trade Database as described in Feenstra et al. (2005). While the original database reports trade to or from 72 countries, which account for around 98 per cent of world exports 19 The OECD contains 30 member countries. The analysis aggregates Belgium and Luxembourg and excludes Iceland. 26


in recent years, the analysis focuses on trade to or from 28 OECD countries, which account for around 73 per cent of world exports. Foreign direct investment stocks were obtained from the OECD International Direct Investment Statistics Yearbook 1992-2003. The 28 OECD countries studied hosted 71 per cent of global inward FDI and were the source of 87 per cent of global outward FDI in 2000 (UNCTAD 2006).

3.4

Estimation

One important feature of the data is that many zero trade flows and investment stocks are reported (table 3.1). The most common approach in the gravity-trade literature is to estimate with ordinary least squares the logarithmic form of the gravity equation, necessarily discarding observations with zero trade. However, by discarding observations in a non-random fashion, this may result in biased estimates of the effects of migrants on trade or investment. An important effect of migrant networks may be to establish trade and investment relations where none would have otherwise existed, so the zero trade and investment flows are of interest. Table 3.1

Bilateral cross-section observations with zero or missing data

Data set

Total Sample observations

Observations with no migrants

Observations with no trade flow or investment stock

Total Working age Inward trade Outward trade Inward trade Inward FDI stock Outward FDI stock Inward FDI stock

Full sample Full sample OECD sample Full sample Full sample OECD sample

4508 4508 756 4228 4266 756

464 464 28 411 423 28

749 749 65 673 698 65

694 551 1 3120 2772 214

Source: See text.

Another important feature of the dataset is that some of these zero observations result from censoring. For example, trade flows were generally excluded from the original dataset where their value was less than US$100,000. In the case of the investment database, concepts, sources and methods vary widely between countries despite OECD attempts to standardise.20 The thresholds for censoring data also appear to vary. For example, Ireland reports positive inward foreign investment

20 Figure C.10 in appendix C illustrates how widely reports vary by reporting country. METHODOLOGY AND DATA SOURCES

27

from only 7 countries in the dataset while Italy and France report positive inward foreign investment from more than 100 countries. Finally, preliminary analysis showed that estimation of the gravity equation in levels suffered from heteroskedasticity. In particular, it appeared that for both trade flows and investment stocks the residual variance was roughly proportional to the square of the expected level of trade flow or investment stock. The simple approach to remove this type of heteroskedasticity is to estimate the logarithmic form of the gravity equation, but this was already precluded by the desire to include the zero observations. Taking into account these features of the dataset, the current paper takes the threshold-tobit approach to estimation introduced by Eaton and Tamura (1994). This method allows the equation to be estimated in logarithms by assuming that zero observations do not strictly represent zero trade flows and investment positions, but rather they simply indicate that these flows and stocks fall below some positive censoring threshold. Specifically, the following equation is estimated by maximum likelihood, where the stochastic error term eij is drawn from a normal distribution and αj is the censoring threshold to be estimated. ln( M ij + α j ) = max (X ij β + eij , ln α j )

While this censoring threshold is normally assumed to be the same across all observations, in the analysis of FDI this assumption is relaxed to allow the threshold to vary by reporting country. The down-side of this approach compared to ordinary least squares regression is that interpretation is more difficult. In the following chapters the coefficients, β, will generally be identified in the text as the semi-elasticity of trade (or investment) with respect to the variable X. This interpretation is not strictly correct, but rather becomes increasingly correct as the trade flows or investment stocks become large.

28


4

Trade

The empirical results, presented in this chapter, reproduce the robust positive correlation between bilateral patterns of trade and migrant numbers found in the previous literature. However, it appears that the increase in bilateral trade arises largely at the expense of trade with third countries. Exploratory analysis suggests that increasing the number of migrants from all countries-of-birth has only a small effect on the total trade flows of their country-of-residence.

4.1

Effects on bilateral trade flows

This section presents estimates of the effects of migrant numbers on imports and exports between their country-of-birth and country-of-residence. The analysis is then extended in an attempt to unpick the role played by migrants in providing information regarding profitable trading opportunities. Some methodological details Before discussing results a few methodological comments are warranted. As a result of using trading-partner fixed effects, only a bare-bones version of the gravity equation can be estimated. The effect on trade flows of variables that are identified with particular countries, such as GDP and population, cannot be estimated. The variables included are the great-circle distance between countries (distance), whether the two countries share a land border (contiguity), speak a common language (language), or share historical colonial ties (colonial ties), the stock of foreign investment held between the two countries (FDI in and FDI out), and the average tariff rate levied by the importer on merchandise from the exporter (tariff). The main explanatory variable of interest is the logarithm of the number of migrants born in the trading-partner as a share of the local population (for example, importer’s share is the number of migrants as a share of the importer’s population). This form of the model was chosen in part because it was statistically preferred,21 21 Statistically preferred in that, after controlling for the logarithm of the migrant share of the total population, other potential migrant indicators were individually not significant (including the log of the number of migrants, the migrant share of the population, and the log of the migrant share of the working-age population). TRADE

29

and in part because it fits more closely with the underlying theory.22 A problem with this approach is that some countries report no migrants from a particular country so that the logarithm of the migrant share is not defined. In this case, the share of migrants in the population is taken to be 1 per 1 million population, and an additional dummy variable called share0 takes the value 1.23 Since the coefficient on share0 can be interpreted as the effect on trade of decreasing the migrant population from 1 person per million to zero it would ordinarily be expected to be small. However, since many of the zero observations occur within three countries (Korea, Germany and Austria) this variable may instead pick up idiosyncrasies related to data collection within these countries. The results are correlative and the cross-sectional nature of the available data precludes analysis of causality. Nevertheless, it seems plausible that causality may run from migrants to trade flows. Migration patterns appear to be affected more by migration policy, wage differentials and the existing size of migrant communities rather than potential trading opportunities.24 Initial results Migrants are found to be strongly associated with bilateral trade flows between countries (table 4.1). Results for the larger sample, that relates to trade flows between 28 OECD members and 162 trading partners, suggest an elasticity of bilateral imports to migrant numbers of 0.127 (column I) and an elasticity of bilateral exports to migrant numbers of 0.145 (column III).25 Since the elasticity of imports to migrant numbers is not larger than the corresponding elasticity for exports, there is no evidence that migrants’ preferences to consume goods from their country-of-birth are driving the effects on bilateral trade.

22 First, the logarithmic form of the relationship assumes a constant elasticity of trade to migrant numbers and hence diminishing returns to additional migrants. This is consistent with the information-based explanation for the relationship between migration and trade, in that each additional migrant from a particular country contribute less additional information on trading opportunities because the local businessmen are better informed when they arrive. Second, including migrants as a share of the population rather than as a total number is consistent with the idea that these information benefits are quite localised. In this form of the model migrants into Australia would have the same effect if considered as a single nation or divided into its States and Territories (so long as the migrants were distributed in proportion to the state populations). 23 A similar trick has previously been employed by Wagner et al. (2002) and Bryant et al. (2004). 24 For example, border effects are larger for migration than trade (Helliwell 1997). 25 Since the dataset is not square, these estimates may differ because exports from OECD countries are determined in different ways than imports into OECD countries, or because migrants affect imports and exports differently. 30


The other variables in the equation are found to take roughly their expected sign and magnitude. Countries that are far apart trade much less than countries that are close together. The estimated elasticity of trade to distance is between -0.7 and -0.9, which is close to the norm found across the gravity-trade literature and, as discussed above, such a large elasticity should be interpreted as evidence of information barriers between countries. Countries that share a common language appear to trade somewhat more than countries that do not. The most broadly specified models estimated on the larger sample of countries (columns II and IV) suggest that sharing a common language raises trade between two countries by between 35 and 57 per cent, though these estimates appear quite fragile to specification and the choice of sample. Former colonial powers trade between 27 and 56 per cent more with their former colonies than would otherwise be expected. Results from the smaller sample, which includes trade only between OECD economies, suggest the possibility that the estimated effects of migrants on trade from the larger sample are overstated. This is because, at least across OECD trading-pairs, the number of immigrants is positively correlated with the number of expatriates. Taking account of migrants resident in both sides of the trading pair results in smaller estimates: the estimated elasticity of imports to migrant numbers is 0.060, while the estimated elasticity of exports to migrants is 0.073 (column V). This suggests caution in applying the estimates from the larger sample. The role of information barriers While these initial results suggest that migrants increase bilateral trade, they provide little understanding of the role played by migrants. In theory, if migrants affect trade flows by providing information about business opportunities then their effect should be largest where the information barriers between countries are largest. Similarly, if migrants play an important role in contract enforcement then their effects would be expected to be smaller for trade within the OECD, since OECD countries typically have well-developed legal systems. To explore these relationships, the share of migrants was interacted with a range of other variables that proxy for information barriers (including distance, foreign investment, colonial ties and commonality of language) as well as OECD membership (columns II and IV of table 4.1). The strongest results are found for foreign investment ties. Foreign investment might be expected to provide business connections between countries that may have the same effect as migrant networks (Rauch 2001). In general, the results support

TRADE

31

this expectation. OECD countries import and export considerably more with countries in which they hold foreign investments.26 If the business networks facilitated by foreign investment are a substitute for migrant networks in providing information on trading opportunities then the effect of migrants should be smaller where country pairs exchange significant foreign investment. The results suggest that this is the case.27 The coefficient on lnFDI*lnshare implies that doubling the amount of foreign investment exchanged between countries reduces the elasticity of imports or exports to the migrant share of the population by around 0.008. As one example, Australia exchanged around USD2.4 billion of FDI with Singapore in 2000 and, interpreted literally, the results suggest that the elasticity of Australian imports to increasing the number of migrants who were born in Singapore would be around 0.1. If Australia exchanged only one-tenth of this amount of investment with Singapore, around the USD220 million it exchanged with South Korea, then this elasticity would be around 0.13. As a second and more extreme example, the United States and France exchanged roughly USD170 billion of FDI in 2000, and as a result the estimated elasticity of US imports to increasing the number of French-born migrants is just 0.01. However, migrants do not appear to play a smaller role where other information barriers between countries are smaller, such as due to proximity, language or colonial ties. Nor is there consistent evidence that migrants have a larger effect on trade with countries outside the OECD.

26 The finding that trade and capital tend to flow together is common in the empirical literature (Collins et al. 1997; Head and Ries 2001; Hejazi and Safarian 2001), against predictions of standard trade models based on differences in resource endowments (Mundell 1957). A possible explanation is that intra-firm trade is increasingly important. For example, around half of US international trade is intra-firm (Blonigen 2005). 27 The variable of interest is the interaction between the migrant share variable and the sum of the inward and outward FDI positions between two countries (ln FDI*ln share). To check that the negative coefficient was not a result of reverse causality (from current trade flows to the current stock of foreign investment), two more robust alternatives were explored: using the stocks of foreign investment from around 1990; or using predicted stocks of foreign investment in 2000 based on stocks in 1990, GDP, population, distance, contiguity, common language the improvement in remoteness between two countries, differences in GDP per capita and each country’s Corruption Perceptions Index. In both cases, the interaction between the measure of foreign investment and the migrant share variable was still found to have a negative and significant coefficient. This result also does not appear to be simply an effect of the size of the economies, as it is not weakened by inclusion of an interaction between the countries’ GDPs and the migrant share. 32


Table 4.1

Bilateral trade and migrants: effects of language, distance, the OECD, foreign investment and colonial tiesa Dependent variable is imports or exports, measured in million US dollars Imports

ln lagged trade ln importer’s share importer’s share0

Basic

With interactions

I

II

III

IV

0.093*** (0.010) 0.127*** (0.014) 0.096 (0.091)

language*ln share ln distance*ln share OECD*ln share ln FDI*ln share colony*ln share

language colonial ties tariff ln FDI in ln FDI out n Log-likelihood Standard error

Basic Immigrants Expatriates

With interactions

0.092*** (0.010) 0.148* (0.086) 0.053 (0.093)

exporter’s share0

contiguity

Imports, OECD sample

Basic

ln exporter’s share

ln distance

Exports

-0.701*** (0.040) 0.279** (0.126) 0.011 (0.075) 0.315*** (0.097) -4.521*** (1.124) 0.023* (0.012) 0.068*** (0.011) 4508 -20677.8 1.1

0.060** (0.026) -0.001 (0.009) 0.031 (0.019) -0.012*** (0.002) -0.018 (0.033) -0.707*** (0.055) 0.347** (0.141) 0.299** (0.126) 0.239* (0.134) -3.972*** (1.128) 0.004 (0.012) 0.034*** (0.012) 4508 -20663.4 1.1

0.081*** (0.008)

0.080*** (0.008)

0.145*** (0.012) 0.102 (0.073)

0.047 (0.119) 0.069 (0.074) 0.069*** (0.023) 0.013 (0.013) -0.030* (0.017) -0.012*** (0.002) 0.035 (0.035) -0.848*** (0.054) 0.089 (0.194) 0.453*** (0.114) 0.445*** (0.133) -3.649*** (1.048) -0.009 (0.010) 0.041*** (0.010) 4508 -20442.0 0.8

-0.902*** (0.045) -0.124 (0.154) 0.140** (0.063) 0.360*** (0.088) -3.039*** (1.058) 0.011 (0.010) 0.070*** (0.009) 4508 -20487.1 0.8

V

VI

VII

0.088*** (0.015) 0.060* (0.031) 0.003 (0.191) 0.073** (0.03) 0.290 (0.203)

0.091*** (0.016) 0.110*** (0.02) 0.097 (0.175)

0.091*** (0.016) 0.117*** (0.018) 0.408** (0.188)

-0.680*** (0.046) 0.171* (0.094) 0.137 (0.090) -0.138 (0.119) -4.412*** (1.467) 0.041*** (0.011) 0.014 (0.013) 756 -5629.9 0.5

-0.684*** (0.046) 0.202** (0.097) 0.132 (0.088) -0.112 (0.119) -4.503*** (1.449) 0.042*** (0.011) 0.016 (0.013) 756 -5634.3 0.5

-0.691*** (0.047) 0.185* (0.095) 0.143 (0.089) -0.120 (0.117) -4.528*** (1.5) 0.044*** (0.011) 0.015 (0.013) 756 -5633.2 0.5

a Results are based on maximum likelihood estimation of a threshold tobit model. All regressions include a full suite of importer and exporter dummy variables. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels.

TRADE

33

Table 4.2

Bilateral trade and migrants: the role of educationa Dependent variable is imports or exports, measured in million US dollars

ln lagged trade ln working-age share Non-OECD, share with tertiary education Non-OECD, share with secondary education OECD*ln working-age share OECD, share with tertiary education OECD, share with secondary education ln distance contiguity language colonial ties tariff ln FDI in ln FDI out n Log likelihood Standard error

Imports

Exports

0.095*** (0.010) 0.149*** (0.017) -0.07 (0.155) 0.015 (0.182) -0.012 (0.019) 0.757*** (0.225) 0.232 0.283) -0.68*** (0.041) 0.357*** (0.121) 0.001 (0.076) 0.246** (0.101) -3.286*** (1.263) 0.032*** (0.012) 0.071*** (0.012)

0.085*** (0.008) 0.198*** (0.015) 0.125 (0.129) 0.209 (0.158) -0.064*** (0.019) 0.813*** (0.238) 0.205 (0.266) -0.91*** (0.040) 0.052 (0.134) 0.109* (0.064) 0.288*** (0.089) -3.508*** (1.131) 0.017 (0.010) 0.07*** (0.009)

3754 -17269.6 1.0

3754 -17035.9 0.8

a Results are based on maximum likelihood estimation of a threshold tobit model. All regressions include a full suite of importer and exporter dummy variables. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels.

The role of education Many characteristics of migrants might be important in determining the effect they have on trade flows. From a policy perspective among the most interesting is the skill level of migrants since this has been a target of migration policy in Australia and some other countries. The potential effect of migrant skills on trade is ambiguous. On the one hand, more educated and skilled migrants might bring with them more market information and may enter more influential positions within their countries-of-residence. On the other hand, more educated migrants might be better positioned to pioneer trade-replacing industries within their host countries and might also be less likely to work directly in importing and exporting businesses. Previous studies have found 34


mixed results, with some evidence that more skilled migrants have a larger effect on trade (Head and Ries 1998, Herander and Saavedra 2005), some evidence that they have a smaller effect on trade (Gould 1994) and some evidence that this is different for exports and imports (Mundra 2005). The current analysis suggests that the education levels of migrants appear to play a role only where migrants come from OECD countries (table 4.2).28 In this case, were 10 percentage points more of the migrant population tertiary-educated rather than lacking secondary education then bilateral imports and exports would be higher by around 8 per cent (over and above the normal effect of migrant numbers). There is no evidence that the educational attainment of migrants from non-OECD countries significantly affects trade. Trade by type of good If migrants affect trade through their knowledge and business networks then the effects should be larger for commodities that are differentiated, in the sense that their quality varies significantly between suppliers. For these goods, the buyer has some inherent uncertainty regarding the quality and characteristics of the product purchased from any particular manufacturer and the price alone does not convey all of the information relevant for international trade.29 Following Rauch (1999), these types of goods are distinguished based on whether it is possible to quote ‘reference prices’ for the goods, that is prices that do not specify the supplier. Commodities that do not have reference prices are called ‘non-homogeneous goods’. The remainder, homogeneous goods, are further split based upon whether reference prices are quoted on organised exchanges (‘Exchange-quoted homogeneous goods’) or quoted only in trade publications (‘Publication-quoted homogeneous goods’). For the former goods it is assumed there are many well-informed, specialised traders so that there is likely little room for migrant networks to improve information flows.

28 In Australia, formal qualification for particular occupations and work experience play a role in selection of migrants. Ideally, analysis of occupational classification of migrants and their linkages to trade would be undertaken but data limitations mean that only formal educational attainment as a share of the working-age population can be studied. The sample size is also reduced by omitting observations with no working-age migrants. 29 An alternative way in which migrant networks may affect trade is through contract enforcement. Rauch and Trindade (2002) suggest that ethnic Chinese networks play a role in this regard, such that if either party to a contract acts opportunistically, then that party’s reputation within the network would suffer. If the effect on trade comes through contract enforcement then the effects would be expected to be quite similar across different types of goods. TRADE

35

The results do not accord well with these expectations. Migrants are associated with greater bilateral trade in all types of goods (table 4.3). The effect of migrants on imports of non-homogenous goods into OECD countries is slightly larger than the effect on imports of homogenous goods, but in general the effects of migrants on trade appears remarkably similar across types of goods. As a cross-check on these results, two alternative ways of disaggregating merchandise trade were considered. Previous studies have found that the effects of migrants is smaller for crude or primary products (Bryant, Genç and Law 2004, Dunlevy and Hutchinson 2001) and larger for consumer goods (Gould 1994, Herander and Saavedra 2005). The current results are less convincing (table 4.3). There is some evidence that the effects of migrants on exports of consumer goods out of OECD countries are larger than for non-consumer goods. But generally the effects of migrants on trade appear very similar across these different types of goods.30

30Primary products are defined to include: food and live animals chiefly for food; crude materials (inedible) except fuel; mineral fuels, lubricants and related materials; and animal and vegetable oils, fats and waxes (SITC rev.2 classes 0, 2, 3 and 4). Consumer goods are defined to include: food and live animals chiefly for food; beverages and tobacco; leather products; furniture; travel goods and handbags; clothing and apparel; and footwear (SITC rev.2 classes 0, 1, 61, 82, 83, 84 and 85).

36


0.037*** (0.009) 0.141*** (0.013) 0.118 (0.088) -0.687*** (0.037) 0.120 (0.110) 0.045 (0.066) 0.121 (0.086) -5.113*** (1.179) 0.018 (0.011) 0.063*** (0.01) 4508 -15111.4 0.93

0.071*** (0.01) 0.111*** (0.014) -0.03 (0.095) -0.776*** (0.043) 0.377*** (0.123) -0.009 (0.076) 0.32*** (0.096) -4.239*** (1.228) 0.041*** (0.012) 0.062*** (0.011) 4508 -13577.1 1.02

0.119

0.132*** (0.013) 0.104*** (0.022) 0.435*** (0.139) -0.884*** (0.059) 0.809*** (0.161) -0.006 (0.107) 0.471*** (0.141) -9.135*** (1.714) 0.067*** (0.019) 0.088*** (0.017) 4508 -13516.7 1.54

0.120

0.146

Nonhomoge neous

Exchang Publicati one-quoted quoted

Imports

0.151

0.107*** (0.012) 0.135*** (0.017) 0.108 (0.109) -0.806*** (0.048) 0.722*** (0.141) -0.057 (0.085) 0.474*** (0.111) -4.954*** (1.343) 0.038*** (0.015) 0.086*** (0.014) 4508 -15974.1 1.29

Primary goods

0.141

0.051*** (0.01) 0.134*** (0.014) 0.145 (0.096) -0.733*** (0.042) 0.039 (0.117) 0.183** (0.077) 0.103 (0.095) -5.569*** (1.246) 0.012 (0.011) 0.067*** (0.010) 4508 -16690.8 1.03 0.141

0.074*** (0.009) 0.131*** (0.013) 0.08 (0.089) -0.529*** (0.039) 0.456*** (0.121) -0.020 (0.068) 0.41*** (0.091) -5.941*** (1.284) 0.024** (0.011) 0.084*** (0.010) 4508 -13459.4 0.93 0.142

0.102*** (0.011) 0.128*** (0.016) 0.143 (0.11) -0.849*** (0.046) 0.176 (0.132) 0.061 (0.086) 0.211* (0.109) -2.764** (1.316) 0.025* (0.013) 0.059*** (0.012) 4508 -18743.9 1.25 0.165

0.111*** (0.012) 0.147*** (0.018) 0.058 (0.133) -1.1*** (0.06) 0.387** (0.177) 0.143 (0.092) 0.218* (0.116) -0.121 (1.741) 0.049*** (0.017) 0.097*** (0.014) 4508 -10012.4 1.25 0.133

0.089*** (0.008) 0.121*** (0.012) 0.139* (0.081) -0.934*** (0.042) 0.168 (0.131) 0.020 (0.062) 0.177** (0.08) -2.805** (1.219) 0.021** (0.011) 0.096*** (0.009) 4508 -14032.6 0.85

Non- Consum Non- Exchang Publicati primary er goods consum e-quoted ongoods er goods quoted

Exports

Dependent variable is imports or exports, measured in million US dollars

Bilateral trade by type of good

0.164

0.080*** (0.008) 0.151*** (0.012) 0.090 (0.077) -0.85*** (0.044) -0.196 (0.15) 0.201*** (0.063) 0.360*** (0.09) -4.137*** (1.073) 0.019* (0.01) 0.062*** (0.009) 4508 -18369.0 0.85

Nonhomoge neous

0.157

0.102*** (0.009) 0.141*** (0.013) -0.031 (0.085) -0.978*** (0.046) 0.378** (0.15) 0.071 (0.067) 0.264*** (0.086) -2.193 (1.414) 0.027** (0.012) 0.085*** (0.01) 4508 -13210.3 0.95

Primary goods

0.158

0.079*** (0.008) 0.145*** (0.012) 0.146* (0.077) -0.907*** (0.046) -0.230 (0.151) 0.185*** (0.063) 0.363*** (0.092) -2.85*** (1.091) 0.021** (0.01) 0.063*** (0.009) 4508 -19371.2 0.85

0.183

0.085*** (0.009) 0.168*** (0.013) 0.019 (0.085) -0.903*** (0.043) 0.174 (0.144) 0.060 (0.066) 0.188** (0.088) -4.200*** (1.28) 0.021* (0.012) 0.056*** (0.01) 4508 -12654.6 0.91

0.150

0.088*** (0.008) 0.136*** (0.012) 0.116 (0.078) -0.911*** (0.046) -0.148 (0.154) 0.166** (0.065) 0.383*** (0.09) -2.363** (1.098) 0.020* (0.011) 0.067*** (0.009) 4508 -19616.1 0.87

Non- Consum Nonprimary er goods consume goods r goods

TRADE

37

a Results are based on maximum likelihood estimation of a threshold tobit model. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels. All regressions include a full suite of importer and exporter fixed effects.

n Log likelihood Standard error Long-run elasticity

ln FDI out

ln FDI in

tariff

colonial ties

language

contiguity

ln distance

share0

ln share

ln lagged trade

Table 4.3

4.2

Effects on aggregate trade flows

The analysis to this point provides some support for the theory that migrants increase bilateral imports and exports between their country-of-residence and country-of-birth. If this is the case then a likely side-effect is that because of these migrant networks between some countries, trade will be lower between countries without migrant networks. A natural question is whether migrants increase total international trade or simply change the pattern of trading partners. The paper takes three approaches to attempt to answer this question. The first approach is to study the effect that the total number of migrants from all countries-of-birth has on the bilateral trade flows of their country-of-residence. The second approach is to implement the theory underpinning the gravity-trade equation in order to calculate the effect that changes in migrant numbers from all countries-of-birth should be expected to have on bilateral trade patterns. Finally, because these first two approaches suggest that migrant networks primarily affect the pattern of trade rather than the total volume of trade, more thorough investigation is undertaken of the effects of aggregate migrant numbers on international openness. Do countries with more migrants trade more? A first attempt to estimate the effects of migration on the total volume of trade is to supplement the analysis within the previous section by estimating the effect that the total migrant share of the population (total share), from all countries-of-birth, has on the international trade of the country-of-residence. If migrants affected only the pattern of trading partners and not the total volume of trade then the elasticity of trade to the total migrant share of the population would precisely offset the direct effects of migrants born in that partner country. The analysis is undertaken without trading-partner fixed effects, but by including the GDP-weighted multilateral form of each bilateral variable (distance, contiguity, colonial ties and common language) discussed in chapter 4, besides the bilateral migrant share variable. Four additional variables are added: the product of the two countries’ GDPs (mass); the product of the two countries’ GDP per capita (masspc); and variables counting the number of countries in the pair that is landlocked (landlocked) or an island (island). The results (table 4.4, columns I and II) suggest that migrants mainly affect the pattern of trading partners and that there is little increase in total international trade associated with evenly strengthening migrant networks. The elasticity of bilateral 38


imports to the bilateral migrant share of the population is estimated to be 0.127, which is the same as that obtained above using trading-partner fixed effects (and so which was believed to be unbiased). However, the elasticity of imports to the total migrant share of the population is -0.110 and is also highly significant. This means that if the number of migrants from all countries-of-birth resident in a particular country increased by 1 per cent then that country’s total imports would increase by only 0.017 per cent and this effect is not statistically different from zero. Similarly, the effect of the total share of migrants in the population on exports appears to offset the effects of bilateral migrant shares, so that the effect on total exports is small (the net effect of increasing the number of migrants from all countries by 1 per cent is estimated to be -0.048 per cent). Should countries with more migrants be expected to trade more? The second attempt to estimate the trade-creating effects of migrants is to implement the approximations of Baier and Bergstrand (2006) to calculate the multilateral forms of the migrant share variables. Since this requires calculation of the strength of both the importer’s and the exporter’s migrant networks with other countries, the analysis is restricted to trade between the 28 OECD countries for which migrant data are available. Nevertheless, this should provide a reasonable guide to the global effects because around three-quarters of the international trade undertaken by these 28 OECD countries is with another country in this group. Before discussing the effects on total trade, it is worth noting that the regression results reiterate the finding in the previous section of this chapter that migrants tend to increase bilateral trade between their country-of-birth and country-of-residence. The results using the GDP-weighted multilateral form (table 4.4, column IV) suggest that migrants have a significant direct effect on imports (with an elasticity of around 0.072) and exports (with an elasticity of 0.132). Using the alternative simpler calculation of multilateral resistance terms results in slightly higher elasticities, around 0.100 for imports and 0.160 for exports (column V). These elasticities are around or a little larger than those obtained in the previous section (table 4.1, column V). The theory allows calculation of how the multilateral resistance terms would be affected by changes in the distribution of migrants. As an illustration, increasing the number of migrants resident in Australia from all countries by 10 per cent would raise the multilateral form of the migrant share variable by between 0.3 per cent (if calculated using simple averages) and 0.7 per cent (if calculated using GDP weights).31 The regression results imply that this would tend to raise Australia’s 31 For other countries the estimates derived using GDP weights would vary somewhat. TRADE

39

imports by between 0.03 and 0.05 per cent and raise Australia’s exports by between 0.05 and 0.09 per cent. By contrast, the same regressions suggest that a 10 per cent increase in the number of migrants from a single country would increase bilateral imports from that country by between 0.7 and 1.0 per cent and increase bilateral exports to that country by between 1.3 and 1.6 per cent.

40


Table 4.4

Bilateral trade flows and the total number of migrantsa Regression results; dependent variable is imports or exports measured in million US dollars Full sample

ln lagged trade ln importer’s share importer’s share0

Imports

Exports

Basic

Multilateral form 1

Multilateral form 2

I

II

III

IV

V

0.060*** (0.007) 0.127*** (0.014) 0.450*** (0.096)

ln exporter’s share

ln exporter’s total share Multilateral (ln importer’s share) Multilateral (importer’s share0) Multilateral (ln exporter’s share) Multilateral (exporter’s share0) ln mass ln masspc island landlocked Multilateral (ln distance) Multilateral (contig) Multilateral (language) Multilateral (colonial ties) n Log-likelihood se

0.050*** (0.005)

0.180*** (0.011) 0.759*** (0.076)

exporter’s share0 ln importer’s total share

OECD sample, imports

-0.110*** (0.032) -0.228*** (0.025)

0.874*** 0.684*** (0.017) (0.012) 0.127*** 0.339*** (0.019) (0.015) -0.276*** -0.174*** (0.056) (0.045) 0.019 -0.104*** (0.048) (0.037) -0.756*** -0.832*** (0.04) (0.034) 0.603*** 0.423*** (0.138) (0.139) 0.281*** 0.408*** (0.086) (0.069) 0.153 0.071 (0.12) (0.105) 4508 4508 -21966.3 -21833.6 1.5 1.2

-0.015** (0.006) 0.058*** (0.02) -0.110 (0.182) 0.181*** (0.021) 0.893*** (0.179) -0.086** (0.041) -0.226*** (0.046)

0.748*** (0.024) 0.317*** (0.04) -0.374*** (0.074) 0.049 (0.062) -0.788*** (0.045) 0.116 (0.116) 0.259*** (0.097) -0.296** (0.117) 756 -5924.8 0.7

0.001 (0.007)

-0.013 (0.008)

0.072*** (0.017) -0.414** (0.210) 0.132*** (0.017) 0.577*** (0.205) 0.804*** (0.023) 0.424*** (0.038) -0.429*** (0.082) 0.021 (0.061) -0.719*** (0.050) 0.188 (0.118) 0.155 (0.104) -0.160 (0.114) 756

0.100*** (0.018) -0.250 (0.195) 0.160*** (0.020) 0.753*** (0.248) 0.565*** (0.023) 0.657*** (0.045) -0.985*** (0.083) 0.180*** (0.064) -0.847*** (0.065) -0.032 (0.15) 0.244* (0.129) -0.246 (0.186) 756 -6035.6 0.9

-5936.4 0.8

a Results are based on maximum likelihood estimation of a threshold tobit model. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels.

That is, the theory underpinning the gravity-trade equation implies the primary effect of migrants on trade should be to change a country’s pattern of trading partners rather than increasing the total volume of trade. The elasticity of Australia’s total volume of trade to changes in the total number of migrants from all TRADE

41

countries-of-birth is expected to be at least an order of magnitude smaller than the elasticity of bilateral trade to the number of migrants from a single country. Does migration affect openness? The analysis of bilateral trade patterns in the previous two subsections suggest that, on balance, increasing the total number of migrants in a country from all other countries has only a small effect on the total volume of trade and that this is what the theory behind the gravity-trade model predicts. As a check on these results, this section explores the relationship between migrants and aggregate openness to trade. Analysis of aggregate openness (taken here to be the ratio of merchandise imports plus merchandise exports to GDP) has advantages and disadvantages compared to analysis of bilateral trade flows. The main advantage is that aggregate data are available for many more countries and years. The disadvantage is that there is little theory to guide the empirical modelling of aggregate openness. In the absence of a theory explaining openness, Guttman and Richards (2004) considered some factors that might be important and the analysis here builds on their results. First, the size of a country in terms of land area may matter because large countries tend to have more diverse resource bases allowing greater selfsufficiency. Second, the size of a country in terms of population may matter because, with economies of scale and lower intra-national than international transport costs, larger countries will tend to manufacture more to serve their own consumers. Third, remoteness from world economic activity raises the cost of international trade relative to intra-national trade, and so lowers openness. Finally, trade policy plays a role. The current study includes an index of tariffs (with a higher score indicating lower tariffs). A panel was constructed at five-year intervals between 1960 and 2005 covering between 86 and 126 countries. Unlike the bilateral trade analysis, observations with zero openness are not an issue, so the following simple equation is estimated in each year with ordinary least squares. The variable of most interest is the total migrant share in the population (total share). ln(openness)= β0 + β1ln(total share) + β2ln(populationi)+ β3ln(land areai)+ β4ln(remotenessi)+β5tariff indexi+ εi The results suggest that in recent years countries that have a larger share of migrants in their populations do not have noticeably higher openness to trade. In 2000, countries with a 1 per cent larger share of migrants in their populations tended to be 0.03 per cent more open to trade, but this effect was not statistically significant 42


(decennial results are presented in table 4.5).32 This is consistent with the analysis based on bilateral trade patterns. However, there is some evidence that in decades past this may not have been the case. In 1960, for example, countries with a 1 per cent larger share of migrants in the population were 0.12 per cent more open to trade and this effect was statistically significant (figure 4.1). Interestingly, the magnitude of this historical elasticity of openness to the migrant share of the population is similar to the simple elasticities of bilateral imports and exports to migrants discussed at the start of this chapter (table 4.1, columns I and III). A better understanding of the historical role of migrants in international trade would require historical data on bilateral migrant numbers. Table 4.5

Opennessa Dependent variable is the log of merchandise imports plus exports divided by GDP

ln total share ln population ln land area ln remoteness tariff index n 2 R

1960

1970

1980

1990

0.118** (0.049) -0.118*** (0.044) -0.109** (0.044) 0.213 (0.186) 0.047** (0.020) 86 0.5012

0.070** (0.030) -0.184*** (0.032) -0.098*** (0.027) 0.062 (0.118) 0.078*** (0.015) 97 0.7086

0.045 (0.034) -0.137*** (0.046) -0.103*** (0.032) 0.038 (0.125) 0.052** (0.020) 106 0.5600

0.001 (0.041) -0.079* (0.045) -0.106*** (0.036) 0.13 (0.123) 0.082*** (0.024) 111 0.4261

2000 0.028 (0.039) -0.057 (0.048) -0.075** (0.038) -0.108 (0.150) 0.054** (0.026) 126 0.2887

a Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels.

32 The estimated effects of other variables differ from those of Guttman and Richards (2004) in two ways. First, in the current analysis the remoteness measure generally does not play a significant role in determining openness. Second, the role of population is smaller than in Guttman and Richards (2004), but this is compensated by land area playing a larger role. Both of these differences appear to be mainly due to the use of openness measured by merchandise trade as a share of GDP from the World Bank’s World Development Indicators, rather than openness from the Penn World Tables. TRADE

43

Figure 4.1

Estimated elasticity of openness to the total migrant share of the population, 1960 to 2005

0.25

0.25

0.20

0.20

0.15

0.15

0.10

0.10

0.05

0.05

0.00

0.00

-0.05

-0.05 -0.10

-0.10 1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

Data source: Author’s calculations. Light dashed lines show 95 per cent confidence intervals.

44


TRADE

45

5

Foreign Direct Investment

The empirical results, presented in this chapter, identify a robust association between bilateral patterns of foreign investment and migrant numbers. It proves more difficult to go further and identify the effects of migrants on the volume of aggregate foreign investment. Nevertheless, there is some evidence that the effect of migrants on the total investment positions of their country-of-residence is smaller than the effect on bilateral investment positions.

5.1

Results for bilateral investment

This section begins by analysing the effect of migrants on inward and outward positions and is then extended by considering the roles of other information barriers and the role played by the education levels of migrants. More methodological details The analysis uses the standard gravity-type variables (distance, contiguity, common language, colonial ties, tariff). To these was added an additional variable (improvement in remoteness) calculated by summing the reduction in the minimum distance to world markets from two plants (that is, in the source country and in the host country) compared to a single plant in the source country.33 This is intended to capture a motivation for not just horizontal investment, but also export-platform investment.34 (Intel’s investment in Ireland to supply the large and nearby consumer markets of continental Europe is one example). Since this measure varies by country-pair it can be included in the regression together with country fixed effects, where the normal indicator of motivations for horizontal investment (host country GDP) could not. Finally, since FDI figures were taken for the year closest

33 The variable is constructed by adding up, weighted by the GDP in each potential consumer market, the difference in distance from the source country and the host country to each consumer market wherever the host is closer than the source to a consumer market. 34 Of course, in this kind of spatial framework it would also be relevant to include a variable that measured the opportunities a source country has to invest in other host countries that provide similar proximity advantages. No such variable was able to be constructed. 46


to 2000 for which data were available (up to 3 years either side), dummy variables for each year were also included. As with the modelling of trade flows, the regression analysis is correlative and does not necessarily imply causation. Nevertheless, it seems plausible that causality may run from migrant networks to investment, rather than the reverse. Foreign investment on a wide scale is a relatively recent phenomenon and, while multinational corporations do move personnel across borders, this is not likely to have a large effect on the number of migrants. Empirically, Javorcik et al. (2006) has explored the possibility of reverse causality with an instrumental variables approach and found positive and significant effects of migrants on investment stocks.35 Initial results The results suggest that migrants are associated with significantly higher foreign direct investment in both directions between their country-of-birth and country-of-residence. Results from the larger sample, which includes investment between 28 OECD countries and 151 partner countries, suggest that increasing the number of migrants resident in an OECD country by around 1 per cent increases inward investment from their country-of-birth by 0.225 per cent and increases outward investment by 0.183 per cent (table 5.1, columns I and III). Though these elasticities are slightly lower than those estimated in the previous FDI literature (focussing on investment out of the United States), they are, as expected, significantly higher than those for trade flows (estimated in the previous chapter). Other variables were generally found to affect investment in the manner expected. OECD countries invest around 44 per cent more in countries that share a common language (column I). This is somewhat larger than the apparent effects on trade presented in the previous chapter, in line with the previous literature, perhaps reflecting the greater informational requirements. Investment into OECD countries was not similarly affected. Former colonial powers exchange roughly 85 per cent more investment with former colonies (columns I and III). Investment also appears to be larger where establishing an additional plant reduces the source country’s remoteness from consumer markets, though not for investment between OECD countries.

35 In that study, the numbers of migrants residing in the United States were instrumented using migrant numbers from each country-of-birth resident in the EU, population density in the migrant’s country-of-birth, the cost of obtaining a passport and legal restrictions on emigration. This approach could not be used in the current study because country dummy variables are included. FOREIGN DIRECT INVESTMENT

47

Investment was also found to fall away sharply as the distance between countries increases. An elasticity to foreign investment of around -0.7 among OECD countries (column V) means that the volume of investment falls by about 40 per cent as the distance between two countries doubles. For investment out of OECD countries the estimated elasticity is much larger, around -1.3 (column III), indicating that the volume of investment falls by around 60 per cent as the distance between two countries doubles. This difference is consistent with the idea that distance is a proxy for information barriers, since the alternative information sources are probably scarcer regarding investment opportunities in, and characteristics of, less developed countries. The role of other information barriers The remaining analysis sought to understand better the role played by migrants in facilitating investment (table 5.1, columns II and IV). If migrant networks increase investment by reducing information barriers then the effects of migrants would be expected to be larger between pairs of countries for which these information barriers were larger. To explore this, the migrant share variable was interacted with indicators of the size of these information barriers including distance, colonial ties and commonality of language. In general, there was little evidence that the role of migrants was larger between pairs of countries with apparently larger information barriers.36 The one exception was that migrants appear to play a larger role in facilitating investment into OECD countries from more distant partner countries (column II). The coefficient on lndistance*lnshare implies that doubling the distance between two countries increases the elasticity of investment to the migrant share of the population by 0.03. One implication is that the elasticity of Australian inward investment to migrant numbers is generally higher than for most other OECD countries because of Australia’s remoteness. For example, if the United Kingdom were located where New Zealand is then the elasticity of Australian inward investment to the share of the population born in the United Kingdom would be around 0.26 instead of 0.34. The elasticity of outward investment to migrants was not found to vary as significantly with distance.

36 Other interactions were investigated, with results available from the author on request. Tong (2005) found that the effect of co-ethnic networks was larger for investment into developed countries and into countries with better bureaucratic quality. For the current paper, this was explored for migrant networks by interacting partner GDP per capita or indicators of corruption with the migrant share variable, but without finding a significantly different effect. 48


Table 5.1

Bilateral investment and migrants: effects of distance, language, the OECD and colonial ties Dependent variable is inward or outward direct investment position, measured in million US dollars Inward position

ln lagged investment ln host’s share host share0

Outward position

Basic

With interactions

Basic

With interactions

I

II

III

IV

0.240*** (0.022) 0.225*** (0.026) 0.455** (0.200)

0.239*** (0.022) -0.154 (0.151) 0.461** (0.200)

ln source’s share source share0 ln distance * ln share common language * ln share OECD * ln share colony * ln share ln distance

-0.837*** (0.158) ln improvement in 0.905 remoteness (0.594) contiguity -0.164 (0.174) common language 0.366*** (0.131) colonial ties 0.608*** (0.158) host’s tariff -2.590 (2.369) n 4228 Log likelihood -7539.6 Standard error 1.14

0.043** (0.017) 0.055 (0.040) 0.024 (0.033) -0.006 (0.059) -0.732*** (0.163) 1.228** (0.608) -0.028 (0.185) 0.525*** (0.183) 0.566** (0.222) -2.759 (2.391) 4228 -7534.9 1.14

0.259*** (0.017)

0.283*** (0.018)

0.183*** (0.023) 0.474*** (0.150)

0.029 (0.126) 0.452*** (0.153) 0.016 (0.014) -0.006 (0.035) 0.029 (0.028) 0.051 (0.044) -1.301*** (0.132) 2.678*** (0.503) -0.095 (0.182) 0.023 (0.152) 0.705*** (0.191) -1.440 (1.921) 4266 -10577.2 1.02

-1.323*** (0.130) 2.549*** (0.486) -0.170 (0.164) 0.046 (0.103) 0.619*** (0.127) -1.820 (1.962) 4266 -10621.5 1.02

Inward position, OECD sample Basic Immigrants only V

Expatriates only

VI

0.181*** (0.027) 0.192*** (0.061) 0.766** (0.357) 0.033 (0.061) 0.763** (0.337)

0.185*** (0.027) 0.208*** (0.044) 0.682** (0.336)

-0.684*** (0.189) 0.004 (0.013) 0.212 (0.205) 0.800*** (0.189) 0.028 (0.229) 2.465 (2.672) 756 -4429.8 0.94

-0.720*** (0.192) 0.008 (0.013) 0.227 (0.210) 0.778*** (0.191) 0.029 (0.235) 2.721 (2.675) 756 -4432.4 0.94

VII 0.176*** (0.027)

0.177*** (0.045) 1.032*** (0.311)

-0.733*** (0.192) 0.006 (0.013) 0.263 (0.207) 0.802*** (0.191) 0.083 (0.229) 2.461 (2.690) 756 -4435.6 0.94

a Results are based on maximum likelihood estimation of a threshold-tobit model. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels. All regressions include full sets of importer, exporter and year dummy variables and the threshold is allowed to vary by importer.

The role of education More educated migrants may be expected to have a larger effect on foreign investment because they are better positioned financially and socially to help entrepreneurs invest abroad. To date, only Javorcik et al. (2006) has investigated this relationship, finding some evidence that that the effect of migrants is larger FOREIGN DIRECT INVESTMENT

49

where migrants held tertiary qualifications. The current results support this finding (table 5.2). Roughly, a 10 percentage point increase in the share of migrants with tertiary qualifications, at the expense of migrants without complete secondary education, appears to raise investment by around 10 to 20 per cent. The results are qualitatively similar whether or not migrants were born in OECD countries. Table 5.2

Bilateral investment and migrants: effects of education Dependent variable is inward or outward direct investment position, measured in million US dollars

ln lagged investment ln working age share Non-OECD, share with tertiary education Non-OECD, share with secondary education OECD*ln working-age share OECD, share with tertiary education OECD, share with secondary education ln distance ln improvement in remoteness contiguity common language colonial ties host’s tariff n Log likelihood se

Inward position

Outward position

0.234*** (0.025) 0.323*** (0.036) 1.024*** (0.366) -0.122 (0.454) 0.012 (0.042) 1.835*** (0.517) 0.802 (0.552) -0.812*** (0.176) 1.110 (0.677) -0.117 (0.174) 0.297** (0.137) 0.496*** (0.153) -0.638 (2.671) 3551 -6606.5 1.12

0.251*** (0.018) 0.258*** (0.030) 1.467*** (0.326) 1.18*** (0.388) 0.024 (0.035) 1.244*** (0.410) 0.572 (0.528) -1.287*** (0.141) 2.401*** (0.552) -0.134 (0.169) -0.063 (0.110) 0.612*** (0.130) 0.159 (2.166) 3564 -8984.9 1.01

a Results are based on maximum likelihood estimation of a threshold tobit model. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels. All regressions include full sets of importer, exporter and year dummy variables and the threshold is allowed to vary by reporting countries.

5.2

Effects on aggregate investment

The analysis to this point has presented robust evidence of a strong association between migrants and investment between their country-of-residence and country-of-birth. Two approaches are taken to extend the analysis to consider the effects that migrants have on the total foreign investment of their country-of-residence. 50


The first approach is atheoretical, minimally expanding the gravity-type framework of the previous section to now include country fixed effects on only one side of the investment relationship. Three other variables are also added: each country’s GDP; each country’s GDP per capita; and a measure of remoteness (host remoteness or source remoteness), which is the GDP-weighted distance to the rest of the world as constructed in the previous chapter. The second approach is an attempt to use the implications of the knowledge capital framework. To stay close to that model, four variables are used: the product of the two country’s GDP (mass); the difference in skill abundance, measured by average years of education in the source country relative to the host country (relative skill); the absolute value of the logarithm of the ratio of the source country GDP to the host country GDP (relative GDP); and the interaction between the relative skill and relative GDP variables. The analysis also includes a corruption index based on data published by Transparency International, with a higher score indicating lower perceptions of corruption. Results The results for investment are less clear than the results for trade. Whereas the results in the previous chapter suggested that increasing the number of migrants would have only a small effect on the world volume of trade, the broadest implications of the analysis of investment patterns (table 5.3) is that it is difficult to determine the likely effect on the world volume of FDI. Results from the gravity-type model and the knowledge capital model differ somewhat. Whether this is because of limitations in the available data or whether the knowledge capital theory simply does not adequately explain the pattern of investment is not clear. The gravity-type model provides some evidence that migrants have a smaller effect on the total investment positions of their country-of-residence than on the pattern of countries with which it exchanges investment. For investment into OECD countries (column I), the results suggest that increasing the number of migrants from a particular country by 1 per cent increases investment from that country by 0.234 per cent, but that increasing the total number of migrants by 1 per cent reduces investment by 0.136 per cent. Similarly, in the case of investment out of OECD countries (column III), increasing the number of migrants by 1 per cent increases investment into their country-of-birth by 0.211 per cent, but increasing the total number of migrants by 1 per cent reduces outward investment from their country-of-residence by 0.077 per cent (though this latter effect is not quite statistically significant, with a p-value of 0.11). Roughly, these results suggest that FOREIGN DIRECT INVESTMENT

51

the effects of migrants on the total foreign investment positions of their country-of-residence are around half as large as their effects on bilateral investment positions with their country-of-birth. The aggregate effect on total investment out of OECD countries is statistically different from zero, but the aggregate effect on total investment into OECD countries is not. Results from the knowledge capital form of the model are mixed. Increasing the total number of migrants resident in a country appears to reduce investment into OECD countries (column II), but increase investment out of OECD countries (column IV). However, there is reason to be cautious in interpreting the results from the knowledge capital form of the model because the estimated effects of other variables do not match theoretical expectations. •

Investment between countries is found to increase with the size of the economies involved, as theory predicted.

•

The relationship between investment and skill is not found to be simple. A positive difference in skill between the source and host countries increases investment into OECD countries (as predicted), but decreases investment out of OECD countries. Since OECD countries are typically skill abundant, this may simply reflect the fact that foreign investment predominantly occurs between skill abundant countries.

•

Economies that differ more in size do not appear to exchange less investment, against the theory’s prediction.37

•

The effect of the interaction between skill differences and size differences is also unclear.

That the knowledge capital model does not fit the data well suggests placing greater weight on the results from the gravity-type model, which include a wider range of other control variables.

37 Most of the literature to date that has studied this effect has focussed on investment from unusually large source countries. For example, Carr et al. (2001) consider only the United States as a source of foreign investment, while di Mauro (2000) considers investment out of France, Germany, Italy, the United Kingdom, the United States, Japan, Korea and Canada. When restricted to these smaller samples the current analysis can also identify a negative relationship between size differences and investment, but this finding may simply mean that investment normally flows into large countries. 52


Table 5.3

Bilateral investment and the total number of migrants Dependent variable is inward or outward direct investment position, measured in million US dollars Full sample, inward FDI Knowledge capital

Gravity ln lagged investment ln host’s share host’s share0

Full sample, outward FDI

IV

V

VI

0.231*** (0.019) 0.234*** (0.026) 0.39** (0.177)

0.430*** (0.023) 0.164*** (0.034) 0.192 (0.266)

0.284*** (0.017)

0.405*** (0.018)

0.211*** (0.024) 0.636*** (0.132)

0.168*** (0.026) 0.819*** (0.168)

-0.077 (0.05)

0.242*** (0.057)

0.242*** (0.026) 0.228*** (0.06) 0.134 (0.472) 0.118** (0.055) 0.952*** (0.327) -0.195* (0.106) -0.012 (0.084) 0.538*** (0.099) 0.544*** (0.064) -0.479* (0.257) 1.334*** (0.142) -0.557*** (0.09) -0.022 (0.224) -0.491** (0.222) 0.005 (0.006) -0.409* (0.213) 0.899*** (0.207) 0.095 (0.268) 0.066 (0.083)

0.294*** (0.03) 0.177*** (0.054) -0.067 (0.621) 0.053 (0.05) 0.977** (0.377) -0.144 (0.117) 0.459*** (0.095)

-0.136* (0.075)

-0.238** (0.102)

0.704*** (0.064)

ln source GDP

0.514*** (0.038) -0.706*** (0.177)

ln source GDP per capita ln distance ln host remoteness

-0.745*** (0.149) 0.022 (0.427)

-0.792*** (0.064)

ln source remoteness ln (improvement in remoteness) contiguity common language colonial ties host’s corruption index

Knowledge capital

III

ln source’s total migrant share

ln host GDP per capita

Gravity

II

source’s share0

ln host GDP

Knowledge capital

I

ln source’s share

ln host’s total migrant share

Gravity

OECD sample, inward FDI

0.533 (0.544) -0.145 (0.178) 0.436*** (0.128) 0.567*** (0.154) 0.096* (0.053)

0.898*** (0.098) -1.282*** (0.134)

-0.373*** (0.046)

-1.324*** (0.203) 2.598*** (0.484) -0.298* (0.169) 0.116 (0.100) 0.432*** (0.121) 0.076 (0.053)

0.042** (0.021)

-0.584*** (0.071)

-0.048 (0.059)

(continued on next page)

FOREIGN DIRECT INVESTMENT

53

Table 5.3

(continued) Full sample, inward FDI Knowledge capital

Gravity I

II

Gravity III

0.586*** (0.042) 0.016 (0.044) 2.154*** (0.171) 0.168*** (0.050)

ln mass | ln(relative GDP) | relative skill relative skill * relative GDP n Log likelihood se

Full sample, outward FDI

4228 -7608.0

2444 -6516.3

OECD sample, inward FDI

Knowledge capital IV

Gravity V

0.421*** (0.030) -0.013 (0.028) -0.306** (0.141) -0.023 (0.026) 4266 -10690.0 1.08

2400 -8192.7

Knowledge capital VI 0.620*** (0.060) 0.007 (0.065) 0.432* (0.246) 0.098 (0.116)

756 -4465.4

650 -4158.7

1.19 1.64 1.16 1.29 1.40 a Results are based on maximum likelihood estimation of a threshold tobit model. Standard errors calculated using the White-robust estimator are reported in parentheses. ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent levels. Gravity-type models include source dummy variables for the inward (full sample) regression, and host dummy variables for the outward regression. In all regressions the threshold is allowed to vary by reporting country, and additional dummy variables are added indicating the year in which the investment stocks were recorded.

54


FOREIGN DIRECT INVESTMENT

55

6

Conclusions

The existing literature has shown that countries trade and invest more with partner countries from which they have received more migrants and the current results confirm this bilateral correlation. Other studies have demonstrated linkages between aggregate international trade and investment and a country’s productivity and living standards. However, the effects of migrants on aggregate trade and investment have not been previously explored. This paper extended the existing literature in that direction. The tentative conclusion is that the consequences of immigration and emigration for productivity and living standards through their effects on international business and social networks are likely more nuanced than they may have previously appeared. The results suggest that increasing the total number of migrants from all countries-of-birth has at most a very small effect on the aggregate trade of their country-of-residence. The results for FDI are less conclusive, but again there is some evidence that migrants have a smaller effect on total investment than on bilateral investment with their country-of-birth. The theoretical explanation for why migrants affect bilateral trade and investment — that migrants strengthen business and social networks thereby reducing information barriers — was also explored, but the results provided only limited support. Evidence was found that migrants have a smaller effect on trade where there are strong business networks already in place (indicated by the stock of FDI) and that migrants have a larger effect on investment between countries that are further apart. But in general it was not clear that the effects of migrants were larger where information barriers between countries were higher. Nor was there strong evidence that the effects of migrants were much larger for trade in differentiated goods, for which information would be expected to be more valuable.

56


CONCLUSIONS

57

A

The Gravity Equation

This appendix outlines why trade flows between countries should follow a gravity equation pattern and how that equation can most effectively be estimated. While the discussion is technical, the effort is justified because the theoretical model provides strong guidance on how the gravity equation can be successfully implemented to provide consistent estimates of the effects of migration on trade flows. Finally, the threshold tobit method of estimation, used for both trade flows and foreign investment positions, is expounded.

A.1

Why a gravity equation?

The standard explanation for the gravity equation based on properties of expenditure systems follows the work of Anderson (1979), and is further developed in Anderson and van Wincoop (2003), Feenstra (2004) and Baldwin (2006). The following discussion most closely follows Baldwin (2006). The goal is to develop a theory of how the trade flows between economies are affected by the size of the economies and the transport costs involved with trading between them. The successful theory should help to explain why there is so much more trade between states within a country than there is between countries, and why there is so much more trade between Australia and New Zealand than between the Netherlands and Ireland (as discussed in chapter 4). To focus attention on the role of transport costs, it is assumed that people in different countries all have the same tastes and that these tastes are homothetic. A consequence of homothetic preferences is that if all prices are the same in two countries then consumption of each good will be in proportion to these countries’ incomes. More specifically, consumers are presumed to have preferences expressed by a constant elasticity of substitution (CES) utility function. A consequence of CES preferences, in particular, is that consumers like variety so much that they want to consume at least a little of all goods available regardless of the price. While consumers in all countries are the same, it is assumed that different countries produce entirely different products so that consumers elsewhere in the world have to import these products. This specialisation of production together with consumers’ desire for variety provides the motivation for international trade. However, the price 58


that producers receive for their goods abroad will in part be required to cover transport costs and this curbs international trade. This sort of model is best suited as an explanation for trade in non-homogenous goods and this raises questions as to what the likely errors are in applying the model for homogenous goods or where there are non-traded goods. Now for some notation. Suppose that a country i produces Ni different varieties of goods. Assuming consumers find each of these goods equally desirable and the transport costs between the two countries and production costs in country i are the same for all goods, then all of the goods from country i should arrive in country j with the same price, pij, and in the same quantity, cij. The total value of imports, Mij, into country j from country i is just the sum of expenditure on all of these goods M ij = N i p ij c ij . Since each good is only produced in one country the imports of a

good in country j must equal the exports of a good from country i. The remainder of the exercise is to find the prices pi that result in the export supply equalling the import demand. Import demand Import demand depends only upon the relative prices of goods from different countries and the total amount consumers spend, denoted Ej. This observation that relative prices matters seems simple enough but in fact ends up being the trickiest part of the estimation of the gravity equation and, as was discussed in the text, has caused trouble with empirical trade modelling in the past. Given the assumption that consumer preferences take the constant elasticity of substitution form, and that the aggregate value of a country’s consumption is equal to the value of a country’s production, it can be shown that demand in country j for each good produced in country i takes the following form. 1 / 1−σ

⎛ C k kj 1−σ ⎞ ( p ij ) −σ j ∆ = ⎜∑N (p ) ⎟ c =E (∆ j )1−σ where ⎠ ⎝ k =1 ij

j

(1)

Here the variable ∆j is an index of the landed prices of all goods that are available for consumption in country j. Adding up the value of all of the goods imported by country j from country i yields total import demand. ( p ij )1−σ M =E N (∆ j )1−σ ij

j

i

(2)

The number σ is the assumed elasticity of substitution between goods in consumption, which is an indication of how willing consumers are to switch between goods when prices change. For example, if this takes the value 0 then THE GRAVITY EQUATION

59

consumers choose the same volume of each good regardless of their relative prices, while if it takes the value 1 then consumers choose to spend the same share of their income on all goods. In the context of global trade it seems reasonable to assume that generally consumers can find close substitutes for most products, so that σ>1. Transport costs The simplest way to include transport costs is to assume that that they can be represented as a mark-up over producer prices that is borne by the exporter and fully passed through to the consumer (as it would be for example with a monopolistically competitive manufacturing industry).38 We assume the relationship between the prices, pi, the producer receives and the prices the consumer in country j pays is given by p = T p , where Tij captures the costs of transporting goods from country i to country j. Here Tij takes the value 1 for trade within countries and is otherwise greater than 1. With transport costs the total import demand equation is given in (3). ij

M ij = E j N i ( p i )1−σ

ij

i

(T ij )1−σ ( ∆ j )1−σ

(3)

Balanced trade The equation above contains all of the structure of the model. The problem is that the number of varieties produced in each country and their producer prices are not observable. A way around this is to apply the equilibrium condition that the total value of goods bought from country i, both locally and from abroad, is equal to that country’s GDP. The whole system is closed by the assumption that each country’s trade is balanced, that is that each country’s total imports and exports are equal and total expenditure, Ej, is equal to total production, Yj. C C Yi (T ik )1−σ Y k = ∑ M ik = N i ( p i )1−σ ∑ j 1−σ Y , where Y is the total of world k =1 ( ∆ ) This means that Y k =1 i i 1−ο GDP, and substituting the solution for N ( p ) into the import equation yields the

gravity trade equation (4). Y jY i ij 1−σ 1 M = (T ) j 1−σ Y (∆ ) (Ωi )1−σ ij

(4)

38 This subsumes the transport industry within the manufacturing industry and avoids the need to separately model its structure. 60


1 /(1−σ )

⎛ C (T ik )1−σ Y k ⎞ ⎟ Ω = ⎜⎜ ∑ k 1−σ Y ⎟⎠ k =1 ( ∆ ) ⎝ where i

(4a)

The derivation of the gravity equation is complete. However, one final result will i i 1−ο prove useful below. Substituting the N ( p ) back into the definition of ∆j reveals the similarity between these terms. 1 /(1−σ )

⎛ C (T kj )1−σ Y k ⎞ ⎟⎟ ∆ = ⎜⎜ ∑ k 1−σ Ω ( ) Y k = 1 ⎝ ⎠ j

(5)

What does it all mean? The gravity equation states that imports between any two countries will tend to increase in proportion to the size of each of the partner economies and should decline as a function of transport costs for those imports. But this was already known. What extra information has the theory bought? The theory shows that trade between two countries is affected by trade costs with other countries. The additional variables ∆j and Ωi are composites of world trade costs. They go by many names. Anderson and van Wincoop (2003) call them the ‘multilateral resistance’ to bilateral trade. The equation shows that if country j faces larger transport costs in importing from all countries, then ∆j will be small and that country will tend to import more from any partner country than would be predicted based only on bilateral trade costs. In this sense, ∆j can be considered to be a measure of the openness of country j to imports. Similarly, if country i faces large transport costs in exporting to all countries then Ωi will be small and the country will tend to export more to any partner country than would otherwise be predicted based only on bilateral trade costs. Ωi can be considered a measure of the access of local firms to foreign markets. These two terms go some way to solve the puzzle of why Australia trades so much with New Zealand. Both countries are far from the rest of the world so while transport costs between the two countries are high, there are not other markets available for them to trade with more cheaply. Moving down from the national to the regional level, these same two terms explain at least part of the reason that a country like the United States trades so little with the rest of the world. American consumers have such a wide variety of goods available within easy reach that the imperative to look outwards is lessened.

THE GRAVITY EQUATION

61

A.2

Calculating multilateral resistance

The theory shows that the relative prices of goods available from other countries (multilateral resistance) affects the level of imports. Ignoring these prices may bias the estimates of all other terms in the gravity equation because it is likely the omitted multilateral resistance terms are correlated with the transport cost variables. Four solutions have been used to address this problem. One solution comes from observing that the two multilateral resistance terms separately describe characteristics of the exporter and importer. For this reason, the simplest way to avoid the omitted variable bias is to include two full sets of dummy variables for importer and exporter. A second solution is to include in the regression explanatory variables that are likely to be correlated with the omitted multilateral resistance terms. These are commonly termed ‘remoteness’ indicators and generally involve some form of averaging of distances to trading partners using weights based on the size of each trading partner’s economy. A third solution is to include estimates of these multilateral price indices directly. The fourth solution is to notice, as did Anderson and van Wincoop (2003), that the gravity model actually provides enough structure to construct these multilateral resistance terms. When trade costs are symmetric a simplifying solution to (4a) and (5) is ∆i=Ωi. The whole system can be estimated by first estimating the coefficients on the trade cost terms in the absence of multilateral resistance, solving for the implied multilateral resistances, then making better guesses at the trade cost terms that are consistent with these resistances, and so forth. Doing so requires a customised optimisation routine. In practice it is possible to instead use a simple Taylor approximation by assuming world trade is close to a simple case, and then making a linear correction to approximate the actual ‘multilateral resistance’ to trade. As discussed in chapter 3, Baier and Bergstrand (2006) suggest two alternative approximations. This section illustrates their approximation technique starting from the simple case where transport costs and information barriers are all zero. In the absence of transport costs the multilateral resistance must be the same in all countries and a solution to the system of equations is Tij=∆j =Ωi=1. The proposed solution is to estimate the j 1−σ i 1−σ product of the multilateral resistance terms (∆ ) (Ω ) by taking a first-order Taylor approximation with respect to ln ∆j, ln Ωi and all ln Tkl about this zero transport cost solution.

(T ik )1−σ k θ k 1−σ k =1 ( ∆ ) C

(Ω i )1−σ = ∑

62


(4a)

(T kj )1−σ k θ k 1−σ k =1 (Ω ) C

(∆ j )1−σ = ∑

(5)

C

C

k =1

k =1

C

C

k =1

k =1

1 + (1 − σ ) ln Ω i = 1 + ∑θ k (1 − σ ) ln T ik − ∑θ k (1 − σ ) ln ∆k

(6a)

1 + (1 − σ ) ln ∆ j = 1 + ∑θ k (1 − σ ) ln T kj − ∑ θ k (1 − σ ) ln Ω k

(6b)

Subtracting 1 from both sides and dividing by (1-σ) yields the slightly simpler expressions. C

C

k =1

k =1

C

C

k =1

k =1

ln Ω i = ∑ θ k ln T ik − ∑ θ k ln ∆k

(7a)

ln ∆ j = ∑ θ k ln T kj − ∑ θ k ln Ω k

(7b)

Multiplying the first by θi and summing over all i, multiplying the second by θj and summing over all j, and adding the two equations yields. C

C

C

C

∑ θ k ln Ω k = ∑∑θ lθ k ln T lk − ∑θ k ln ∆k k =1 C

l =1 k =1

k =1

C

C

C

(8a)

∑ θ k ln ∆k = ∑∑θ lθ k ln T lk − ∑θ k ln Ω k k =1

l =1 k =1

(8b)

k =1

Adding these together and collecting common terms yields C

C

C

∑ θ k (ln Ω k + ln ∆k ) = ∑∑ θ lθ k ln T lk k =1

(9)

l =1 k =1

Substituting each of these equations into the other yields C

C

C

C

ln Ω i = ∑ θ k ln T ik − ∑∑ θ lθ k ln T lk + ∑ θ k ln Ω k k =1

l =1 k =1

k =1

C

C

C

C

(10a)

ln ∆ j = ∑ θ k ln T kj − ∑∑ θ lθ k ln T lk + ∑ θ k ln ∆k k =1

l =1 k =1

(10b)

k =1

Summing (10a) and (10b) and substituting in (9) yields (11).

THE GRAVITY EQUATION

63

C

C

k =1

k =1

C

C

ln Ω i + ln ∆ j = ∑ θ k ln T ik + ∑ θ k ln T kj − ∑∑ θ lθ k ln T lk

(11)

l =1 k =1

Equation (11) can now be calculated directly using knowledge of distance, migrants stocks, and a variety of other variables that affect trade costs, and substituted into the gravity equation. C

C

k =1

k =1

C

C

ln M ij = ln Y j + ln Y i + (1 − σ )(ln T ij − ∑θ k ln T ik − ∑ θ k ln T kj + ∑∑ θ lθ k ln T lk ) l =1 k =1

(12)

This is simpler to implement than the customised optimisation routine proposed by Anderson and van Wincoop (2003). Baier and Bergstrand (2006) show that this approximation produces very similar coefficient estimates and similar comparative statics for at least some policy questions, such as the effects of borders on trade. The reason this approximation works well is that the higher-order terms in the Taylor series are largely uncorrelated with the remaining terms in the gravity equation, so that the estimation bias resulting from omitting these terms is small.

A.3

Estimating the gravity equation

The earlier discussion resulted in a gravity trade model of the form given in (4). The problem is that the deterministic theory from which this is derived does not clearly guide the stochastic interpretation of the model. If theory is generally believed to provide an indication of the expected or average values seen in practice then this suggests equation (13a), written more simply as (13b). Alternatively, the theory could be interpreted as indicating an expected value in the log scale (13c). E ( M ij | Y i , Y j , T ij , ∆ j , Ω i ) = exp( β1 + β 2 ln Y i Y j + β 3 (ln T ij − ln ∆ j − ln Ω i ))

(13a)

E ( M | x ) = exp( xβ ) = µ ( x; β )

(13b)

E (ln M | x ) = xβ

(13c)

Thus the theory provides limited guidance as to an appropriate estimation method. The logarithmic form can be simply estimated with ordinary least squares, while the exponential form of the conditional mean equation can be used for estimation within the Generalised Linear Model framework. The latter has the advantage of allowing estimation retaining zero trade flows but preliminary analysis using non-linear least squares, Poisson regression and gamma-class GLM regression

64


showed that the residuals exhibited severe heteroskedasticity.39 In particular, the Park test, suggested by Manning and Mullahy (2001), demonstrated that the point variance went roughly like the square of the predicted volume of trade.40 Manning and Mullahy (2001) provide advice on model selection in this circumstance. While in principle gamma-class regression could take account of this, it can be more simply addressed through logarithmic regression, and since the errors in the log scale also show kurtosis this is preferred. However, a customised approach to estimation in the log scale was pursued that allowed retention of the zero trade flow and investment position observations. For investment positions and trade flows Vij from country i to country j separately we estimate the equations. Vij = max (− α j + exp( X ij β + µ i + ν j + eij ),0 )

Here Xij are a range of variables expected to affect flows between the two countries (eg distance), µi and νj are country-specific fixed effects and eij is assumed to be a normally distributed error term with variance σ2. Of particular note, the αj are the data reporting thresholds. Thus a positive value is reported if the volume of trade exceeds some minimum threshold, αj. These are assumed to be the same across all countries for trade flows but to differ by reporting country for investment positions.41 These equations are estimated by maximum likelihood. To derive the log likelihood function, start by defining a latent variable V*ij with the data reporting rule that Vij=V*ij if V*ij>0 and Vij=0 otherwise. Vij* = −α j + exp( X ij β + µ i +ν j + eij ) ln(Vij* + α j ) = X ij β + µi + ν j + eij Reorganising terms and taking logarithms yields . Hence the cumulative density function for the latent variable has the simple form given in the following expression.

39 The poisson approach in particular has recently been used for estimating gravity-trade type equations. Silva and Tenreyo (2003) find that estimation of gravity-trade models using the Poisson-like structure is preferable to non-linear least squares in the presence of a range of forms of heteroskedasticity, to the point the authors suggest the latter estimator is rendered useless in practice. 40 Estimates suggested the point variance was proportional to the predicted volume of trade raised to a power of around 1.7 to 1.9. 41 The exception is analysis of investment patterns between OECD countries where the data have been harmonised, as discussed in appendix B. THE GRAVITY EQUATION

65

⎛ ln(Z + α ) − X ij β − µ i −ν j Pr(Vij* < Z ) = Φ⎜⎜ σ ⎝

⎞ ⎟⎟ ⎠

The likelihood has a continuous part for trade above the threshold and a discrete part for trade below the threshold. The discrete part is given by a simple expression. ⎛ ln α − X ij β − µi −ν j Pr(Vij = 0) = Pr(Vij* < 0) = Φ⎜⎜ σ ⎝

⎞ ⎟⎟ ⎠

The continuous part is derived by differentiating the cumulative density. f (Vij = Z ) =

⎛ ln(Z + α ) − X ij β − µi −ν j ⎞ ∂ ⎛ ln(Z + α ) − X ij β − µi −ν j ⎞ 1 ⎟⎟ = ⎟⎟ Φ⎜⎜ φ ⎜⎜ ( ) ∂Z ⎝ Z + σ α σ σ ⎝ ⎠ ⎠

The estimation proceeds by obtaining the parameters that maximize the log likelihood given by the following expression. ⎛ ln α j − X ij β − µ i −ν j ⎞ ⎟⎟ ln L(V , X ; β ,α , µi ,ν j ) = ∑ ln Φ⎜⎜ σ V =0 ⎝ ⎠ ⎛ ln(Vij + α j ) − X ij β − µ i − ν j + ∑ ln φ ⎜⎜ σ V >0 ⎝

⎞ 1 ⎟ − ln(Vij + α j ) − ln σ 2 ⎟ 2 ⎠

Due to the censoring the coefficients β converge only asymptotically to the semi-elasticities.

66


B

Data definition and sources

Bilateral trade, investment and migration cross-section Corruption: Data was taken from Transparency International’s Corruption Perceptions Index, 2006. A higher score indicates lower corruption perceptions. Distance: Great-circle distances in kilometres between the largest cities in each country were obtained from the Centre d'Etudes Prospectives et d'Informations Internationales (http://www.cepii.fr/anglaisgraph/bdd/distances.htm). Internal distances within countries were estimated by 0.67 landarea / π (Head and Mayer 2002). The same dataset provided indicators of contiguity (two countries sharing a common land border), common language (two countries sharing a language spoken by at least 9 per cent of the population), historical colonial ties, and whether a country is landlocked or an island. Belgium was used to represent Belgium-Luxembourg. Foreign direct investment positions: Foreign direct investment positions (inward and outward) were sourced from the OECD’s International Direct Investment Statistics Yearbook 1992-2003. In many cases data were missing. Three steps were taken to produce the largest possible data sample. First, inward and outward foreign direct investment position data are used for the nearest available year to 2000, searching up to 3 years either side. Data are expressed in US dollars in the prevailing exchange rate in that year. Second, where data are missing they have been coded to zero. Finally, for the subset of investment positions between OECD countries both inward and outward investment should be separately reported for each country pair. For this subset, instead of taking data from the nearest available year, data were averaged across reports of inward and outward investment in 2000 where both were available, and missing data were replaced with reports of the corresponding position where only one observation was available. This harmonised dataset was used where only OECD countries were considered. Gross domestic product (GDP) and population: Estimates of GDP in current US dollars and population for the cross-section analysis were sourced from the IMF World Economic Outlook, September 2006. Where this data was unavailable for particular countries and years, data were sourced from the World Bank World DATA DEFINITION AND SOURCES

67

Development Indicators. Failing these two sources, population data were sourced from Maddison (2003), while GDP was estimated based on the GDP per capita of the nearest available year. Migration: Populations of migrants within OECD countries were obtained from the OECD Immigrants and Expatriates Database, as detailed in Dumont and Lemaître (2005). These are drawn from each member country’s census closest to the year 2000 and migrants are identified by country-of-birth (exceptions are Germany where data are based on a household survey, and Korea and Japan where data are based upon nationality rather than country-of-birth). Switzerland was aggregated with Liechtenstein and Monaco with France. Where the former Czechoslovakia, former Yugoslavia and former USSR were indicated as countries-of-birth these were disaggregated in proportion to the average shares across countries for which data were available. The education variables used are the number of people with each level of educational attainment (tertiary, upper secondary, less than upper secondary) as a share of the working-age population reporting their educational attainment. Skill: Skill was measured by average years of schooling of the population aged 25+ years in 1999, from Barro and Lee (2001). Tariffs: Data on tariffs were obtained from the World Bank’s Bilateral Tariff Data, http://econ.worldbank.org, as detailed in Bouët et al (2004). These data are provided at the 6-digit Harmonised System (HS6) level based on simple averages of line-item tariffs and take account of both ad valorem tariffs and the ad valorem equivalent of specific tariffs. The current paper uses a simple average across HS6 classes. Missing data are filled using the average for that importer, and the indicator is used only in models that include importer dummy variables. Trade flows: International trade flows by commodity were obtained from the NBER-UN World Trade Data, as outlined in Feenstra et al (2004). Missing US export and Indian import data for 2000 was replaced with the corresponding data for 1999. China was aggregated with China Free Trade Zone and China Macau SAR. Trade flows by differentiation of goods: Trade in goods was aggregated from three and four-digit SITC level into three categories: ‘non-homogeneous’; ‘exchange-quoted homogeneous’; and ‘publication-quoted homogeneous’. The classification was obtained from Jon Haveman’s International Trade Data website, but originates with Rauch (1999). Because ambiguity resulted from aggregation of commodities, both ‘conservative’ and ‘liberal’ classifications were provided, with the latter using a more liberal definition of non-homogeneous goods. The results in the current paper use the conservative definition; results using the liberal definition are similar and are available from the author on request. 68


Openness analysis Remoteness: Remoteness was calculated as the GDP-weighted geometric average great-circle distance between each country’s main city. Nominal market exchange rate GDP weights were calculated based on the Penn World Table 6.2. Tariff index: A tariff index was obtained from the Institute for Economic Freedom’s Economic Freedom of the World Annual Report 2006 (indicator 4A). Where historical data are missing these are back-cast from the following periods observation using ordinary least squares regression. Data from 2004 were used to represent 2005. A higher score indicates less restrictive tariffs. Trade openness: Openness data, measured as merchandise imports plus exports as a share of GDP, were sourced from the World Bank World Development Indicators, 2006. The same source was used for land area (in square kilometres), population and the total number of migrants as a share of the population.

DATA DEFINITION AND SOURCES

69

C

Australia's patterns of migration, trade and foreign investment

This appendix outlines changing volumes of Australia’s trade, migration and foreign investment flows over time, how these compare with other OECD countries and the patterns of bilateral relationships. Two observations motivate exploring the link between migration and trade. First, on the one hand, migrants are a larger share of the Australian population and are better educated than in any other OECD country. On the other hand, previous studies have suggested that Australia’s bilateral trade is higher than would be predicted given its remoteness. Second, over the past half-century both the proportion of migrants in the Australian population and the proportion of trade in the Australian economy have been rising, and the directions of both migration and trade have been shifting away from the United Kingdom and Ireland and towards Asia.

C.1

Migration

Migrants have always played a large role in the Australian economy. At Federation in 1901 around 23 per cent of the population was foreign born (figure C.1). This proportion declined steadily through to the end of the Second World War, at which point less than 10 per cent of the population was foreign born. Since that time, successive waves of migration from Europe and, later, Asia have returned the proportion of the population that was foreign born to around 23 per cent in 2001 (or more than 4 million people), while increasing the cultural diversity of the population. Elsewhere in the OECD, migrants generally play a smaller role than in Australian society (figure C.2). Migrant populations remain large in historical ‘settler’ economies — New Zealand, Canada and the United States — and in some small, open European economies such as Switzerland. In most European countries migrants represent a much smaller proportion of the population (10 per cent or less), though this is still large compared to Japan and Korea (where 1 per cent and 0.3 per cent respectively of the populations are migrants). Nevertheless, in the past decade or two these patterns of migrant flows have been changing (Coppel et al. 2001). Migration into EU countries has been growing, peaking in the early 1990s 70


with the increased migration from Eastern European countries and a larger number of asylum seekers. Historically strict migration restrictions in Japan have been eased somewhat as that country confronts economic challenges associated with its changing demographics. Figure C.1

Australians born overseas by region of birth Foreign-born per cent of population identifying country of birth

25 20

United Kingdom and Ireland

Other European countries

Asian countries

New Zealand

25

Other countries

20

15

15

10

10

5

5

0 1901

1911

1921

1931

1941

1951

1961

1971

1981

1991

0 2001

Data source: ABS, Australian Historical Population Statistics, cat. no. 3105.0.65.001.

Though more difficult to measure, the Australian population overseas may also play an important role in Australia’s society. Reliable data are hard to pin down. One reason is that the notion of an expatriate community is not precise. The oft-quoted figure is that on any given day as many as one million Australian citizens are overseas, of which around three-quarters are long-term residents abroad and the balance short-term visitors.42 More accurate data are available for expatriates defined as those residing outside of their country-of-birth. Around 346,000 Australian-born people were residing in other OECD countries around the year 2000 (Dumont and Lemaitre 2005). For consistency with the definition of migrants used above, this latter definition of expatriates is used throughout the paper.43

42 These data reflect Department of Foreign Affairs and Trade consular estimates of the number of Australian citizens abroad (DFAT 2002, 2004). 43 Nevertheless, these two sets of estimates appear to be broadly consistent. Department of Foreign Affairs and Trade estimates suggest around 550,000 Australians are resident in Western Europe, the United Kingdom and Ireland, North America or New Zealand. While census data show 346,000 Australian born people resident in OECD countries, these data are broadly reconciled by noting that one-quarter of Australians were not born in Australia and among foreign-resident Australian citizens the number born overseas is likely to be larger. For example, just over half of emigrants permanently departing Australia in recent years were born overseas (Hugo et al 2001). AUSTRALIA'S PATTERNS OF MIGRATION, TRADE AND FOREIGN INVESTMENT

71

Figure C.2

Immigrants in OECD countries by country of residence Per cent of country-of-residence population identifying country-of-birth 25

20

20

15

15

10

10

5

5

0

0

A Sw us N itz tra ew e lia Ze rlan a d C lan an d U ni A ad te us a d tr S ia Swtate e s Be de l- n Ire Lux N G lan et re d he e U rl ce ni te F and d ra s Ki n ng ce N dom D orw en a y Po ma r k r G tu er ga C m l ze ch S any R pa ep in ub lic H Ita u n ly Fi gar y Sl nlan ov d a Po kia la Tu nd rk Ja ey M pan ex Ko ico re a

25

Data source: OECD, Database on Immigrants and Expatriates.

Figure C.3

Expatriates living in other OECD countries by country of birth Per cent of country-of-birth population 25

20

20

15

15

10

10

5

5

0

0

Ir N P elan ew or d Ze tug al al M and ex SwSlo ico itz v a k U e ia ni te G rlan d re d Ki e ng ce d Fi om na Au lnd st ri G It a N er al et m y he a rla ny C nd an s B ad D el-L a en u m x a Ko rk r Po ea H lan un d ga N C o r ry ze ch T way R u rk ep ey Sw ubl e ic Fr den an ce S Au pa st in r U ni J alia te a d pa St n at es

25

Data source: Author’s calculations based on OECD, Database on Immigrants and Expatriates.

Whereas Australia’s migrant population is unusually large, Australia’s expatriate population is unusually small. Expressed as a fraction of the Australian population, the number of Australian-born people residing in other OECD countries amounts to less than 2 per cent of the Australian population (figure C.3). By contrast, many European countries have expatriate communities that are two or three times as large and in some cases far larger. Only Japan and the United States have smaller expatriate communities than Australia, relative to their population. 72


Figure C.4

Where do Australians come from and where do they go? Thousand people

Immigrants by country of birth

Expatriates by OECD country of residence

1200

1200

120

120

1000

1000

100

100

800

800

80

80

600

600

60

60

400

400

40

40

200

200

20

20

0

0

0

U ni te d U Kin ni g t N ed dom e w St Ze ate al s G and re C ece an ad N et I a he ta rla ly n Ja d s pa Ire n la Fr nd an ce

U ni te d N Kin ew g Ze dom al an d Vi Ita et ly na C m h G ina G re e Ph erm ce ilip an pi y ne N et In s he d rla ia nd s

0


While Australia’s migrant and expatriate communities differ markedly in size, they share some characteristics. One feature is that migrants’ countries of origin are similar to expatriates’ countries of destination. The United Kingdom is both the most common source and most common destination, and other European countries such as Italy and Greece are also popular (figure C.4). This two-way migration is not unique; immigrant and expatriate populations are positively correlated across the OECD. Another characteristic that Australia’s migrant and expatriate communities share is that they are highly educated (table C.1). In most OECD countries, migrants have lower educational attainment than the locally-born population. Notable exceptions are Canada, New Zealand and Australia, where skill plays a larger role in the migration programme.44 In most other OECD countries the majority of new arrivals are linked to family reunion. For example, in the United States and France this is the motivation for around three-quarters of new arrivals (Coppel et al. 2001). In Australia, 70 per cent of the annual migrant intake currently comes through the skilled migration programme, and a larger share of the migrant population is tertiary qualified than in any other OECD country.

44 Greece and Italy are other examples where immigrants have better educational attainment, but this reflects the lower qualifications of the locally-born labour force. AUSTRALIA'S PATTERNS OF MIGRATION, TRADE AND FOREIGN INVESTMENT

73

Table C.1

Educational attainment of Australian migrants and expatriates Per cent of working-age populations Not completed secondary education

Completed secondary education

Australian born and resident

46

16

39

Foreign-born Australian residents

39

19

42

Australian-born residents of other OECD countries

18

38

44

Populations

Bachelor’s degree or higher


C.2

Trade

Australia’s trade has risen rapidly in recent decades accompanying declines in tariffs and shipping costs, and the development of trading partners in Asia. This has reversed the trend towards smaller trade flows evident in the first half of the 20th century (figure C.5). With the exception of the wool boom in the early 1950s (associated with the Korean war), total trade as a proportion of Australia’s GDP has not been as high as it is today since the end of the First World War. Australia’s patterns of trade have also shifted over the past few decades. The United Kingdom and Ireland accounted for 18 per cent of Australian merchandise trade in 1970 but just 5 per cent in 2000. Over the same period, the proportion of Australia’s trade undertaken with Asia increased from 37 per cent to 56 per cent (excluding countries formerly part of the USSR). Despite this trend towards greater openness, even today trade plays a smaller role in the Australian economy than in almost all other OECD economies (figure C.6). Indeed, while Japan and the United States trade less than Australia as a share of GDP, this reflects the self-sufficiency that comes with size and diverse industrial structures. Geographic remoteness likely explains why Australia trades so little. Unlike most other developed economies, Australia is a long way from all other developed economies. It is situated half a world away from the large markets in Europe and the United States. The trip from Sydney to Wellington, the capital of Australia’s nearest developed neighbour, passes over 2300 km of open water. A similar length trip from Paris to Moscow passes over four other countries and 140 million people. Australia’s isolation raises the cost of transporting goods between Australia and the rest of the world. The puzzle, perhaps, is why despite these high trade costs Australia trades so much. Previous studies suggest that Australia’s bilateral trade is

74


greater than would be expected after taking account of its size and remoteness from trading partners (Battersby and Ewing 2005, Guttman and Richards 2005). Figure C.5

Australia’s international trade, 1900/01 to 2005/06 Per cent of GDP

60

60

50

50

40

40

30

30

20

20

10

10

0 1900/01

0 1920/21

1940/41

1960/61

1980/81

2000/01

Data source: Australian Bureau of Statistics, Australian System of National Accounts, 2005-06, cat. no. 5204.0; Butlin (1977).

Figure C.6

Merchandise trade in OECD countries, 2000 a

Imports plus exports as a per cent of GDP

150

100

100

50

50

0

0

Ir H ela Sl un nd o C vak Be gar ze R l- y ch e Lu p N Re ub x et p lic he ub rla lic Sw Ca nd itz na s er da la K nd Fi ore n Sw lana e d Au de n M stri a e D x en ic o N mar o G rw k e N P rmaay ew o n r Ze tug y a al Ic lan e d Poland U Fr land ni a te d S nce Ki p ng ain do m TuItaly Au rk st ey G rali re a ec U e Ja SA pa n

150

Data source: NBER-UN World Trade Data; IMF, World Economic Outlook, September 2006.

Australia’s isolation also affects the direction of its trade. Remoteness from Europe and the United States means that Australia trades more with (relatively) near neighbours. Australia’s largest trading partners are Japan, the US and China, but AUSTRALIA'S PATTERNS OF MIGRATION, TRADE AND FOREIGN INVESTMENT

75

this is because they are such large economies. The effect of remoteness is more clearly seen if these trade flows are expressed as a percentage of trading-partner GDP. By this metric, Australia’s leading trading partners are closer to home: Kiribati, Papua New Guinea, Fiji, Samoa and New Zealand (figure C.7). Figure C.7

Whom does Australia trade with? Total merchandise trade (imports plus exports), 2000 By value USD billion

By share of trading partner GDP per cent 120

120

25

25

100

100

20

20

80

80

15

15

60

60

10

10

40

40

5

5

20

20

0

0

0

0

W es

Ki rib at i PN G

te rn F N Sa iji ew m Ze oa al Vi and et Si na ng m ap M or al e a Th ysia ai In lan do d ne si a

30

U ni J a te pa d St n at es C h Un in ite K a or d K e N ing a ew d Ze o m al a Ta nd iw G er an m Si an ng y ap M ore al ay si a

30

Data source: NBER-UN World Trade Data; IMF, World Economic Outlook, September 2006.

C.3

Foreign investment

The stock of foreign direct investment into Australia (that is Australia’s inward position) has risen steadily over recent decades, doubling from 15 per cent of GDP in 1980 to 30 per cent in 2006 (figure C.8). Over the same period, the stock of Australian investment abroad (Australia’s outward position) grew from 3 per cent to 28 per cent of GDP. The growth in investment over the past 25 years has closely mirrored foreign investment trends across the world in aggregate, which increased from 6 per cent of world GDP in 1980 to 24 per cent in 2005.

76


Figure C.8

Australia’s foreign direct investment, 1980 to 2005 Per cent of GDP

40

Australian direct investment abroad (outward position)

40

Direct investment in Australia (inward position) 30

30

20

20

10

10

0 June-80

0 June-85

June-90

June-95

June-00

June-05

Data source: ABS, Balance of Payments and International Investment Position, cat. no. 5302.0, December 2006; ABS, National Accounts: National Income Expenditure and Product, cat. no. 5206.0, December 2006.

Figure C.9

Foreign direct investment into and out of OECD countries, 2000 per cent of GDP Inward position

Outward position Norway Switzerland Netherlands Bel-Lux UK Sweden Denmark Finland France Canada Ireland Germany Spain Australia Portugal Italy NZ USA Austria Iceland Japan Greece Korea Hungary Turkey Slovakia Mexico Czech R. Poland

Ireland Bel-Lux Netherlands Hungary NZ Denmark Sweden Czech R. Switzerland UK Portugal Canada Australia Spain Poland Finland France Slovakia Norway Mexico Austria Germany USA Greece Italy Turkey Korea Iceland Japan

0

40

80

120 160 200 240

0

40

80 120 160 200 240

Data source: UNCTAD, World Investment Report, 2006.

AUSTRALIA'S PATTERNS OF MIGRATION, TRADE AND FOREIGN INVESTMENT

77

Figure C.10 Where do Australians invest, and whom invests in Australia?a Billion US dollars Direct investment in Australia

Australian direct investment abroad

40

40

30

30

30

30

20

20

20

20

10

10

10

10

0

0

0

0

Switz erla

Germ an

Bel-L u

PNG Sing apor e Arge ntina Cana da

40

Unite d S ta tes Unite d Kin gdom New Zeala nd Berm uda Hong Kong Neth erlan ds

40

nd Sing apor e

50

y Fran ce

50

x

50

Unite d S ta tes Unite d Kin gdom J apa n Neth erlan ds New Zeala nd

50

a The figure shows investment positions in the year closest to 2000 for which data are available, originally derived from data reported to the OECD by the Australian Bureau of Statistics. Error bars shows another estimate of the same investment position, derived from data reported to the OECD by the partner country’s statistics bureau where these data are also available. Data source: OECD, International Direct Investment Statistics Yearbook, 1992-2003.

Whereas Australia’s volumes of trade are small by comparison with other OECD countries, its foreign investment positions are closer to the OECD average (figure C.9). Many of Australia’s largest trading partners are also key investment partners (figure C.10). The United States, United Kingdom, New Zealand and Japan account for 72 per cent of direct investment into Australia and 81 per cent of direct investment out of Australia. There are two significant problems with these investment data. First, where foreign investment is directed to companies registered in financially convenient locations, such as Bermuda, it is not possible to identify the countries in which the physical investment occurs. Second, and more fundamentally, different countries record foreign investment using different criteria and data sources. For transactions between OECD countries this can be illustrated because each investment position is reported by both the host and source country. The error bars in figure C.10 show that in these cases there are some very large differences between these reports of Australian investment positions.

78


References

Aghion, P. and Howitt, P. 1992, ‘A model of growth through creative destruction’, Econometrica, vol. 60, no. 2, pp. 323-51. Alcalá, F. and Ciccone, A. 2004, ‘Trade and productivity’, Quarterly Journal of Economics, vol. 119, no. 2, pp. 613-46. Anderson, J. 1979, ‘A theoretical foundation for the gravity equation’, American Economic Review, vol. 69, no. 1, pp. 106-16. —— and van Wincoop, E. 2003, ‘Gravity with gravitas: A solution to the border puzzle’, American Economic Review, vol. 93, no. 1, pp. 170-92. —— and —— 2004, ‘Trade costs’, Journal of Economic Literature, vol. 42, no. 3, pp. 691-751. Baier, S. and Bergstrand, J. 2006, ‘Bonus vetus OLS: A simple approach for addressing the “border puzzle” and other gravity-equation issues’, working paper, University of Notre Dame, March. Baldwin, R. 2006, ‘The Euro’s trade effects’, European Central Bank Working Paper Series, no. 594, March. Bandyopadhyay, S., Coughlin, C. and Wall, H. 2006, ‘Ethnic networks and US exports’, Federal Reserve Bank of St. Louis Working Paper Series, 2005-069D (September 2006 revision). Bardhan, A. and Guhathakurta, S. 2004, ‘Global linkages of subnational regions: Coastal exports and international networks’, Contemporary Economic Policy, vol. 22, no.2, pp. 225-36. Barro, R. and Lee, J. 2001, ‘Schooling quality in a cross-section of countries’, Economica, vol. 68, November, pp. 465-88. Battersby, B. and Ewing, R. 2005, ‘International trade performance: The gravity of Australia’s remoteness’, Australian Government Treasury Working Paper Series, no. 2005-03. Bergstrand, J. 1989, ‘The gravity equation in international trade: Some microeconomic foundations and empirical evidence’, Review of Economics and Statistics, vol. 67, pp. 474-81.

REFERENCES

79

Bhattacharya, U. and Groznik, P. 2005, ‘Melting pot or salad bowl: Some evidence from US investment abroad’, paper presented to the 2003 European Finance Association annual meeting, March. Blanes-Cristobal, J. 2004. ‘Does immigration help to explain intra-industry trade? Evidence for Spain’, Economic Working Papers at Centro de Estudios Andaluces, no. E2004/29. Blonigen, B. 2005, ‘A review of the empirical literature on FDI determinants’, Atlantic Economic Journal, vol. 33, no. 4, pp. 383-403. —— and Davies, R. 2004, ‘The effects of bilateral tax treaties on US FDI activity’, International Tax and Public Finance, vol. 11, no. 5, pp. 601-22. Blum, B. and Goldfarb, A. 2006, ‘Does the Internet defy the law of gravity?’, Journal of International Economics, vol. 70, no. 2, pp. 384-405. Bouët, A., Decreux, Y., Fontagné, L., Jean, S. and Laborde, D. 2004, ‘Computing an exhaustive and consistent, ad-valorem equivalent measure of applied protection: a detailed description of MAcMap-HS6 methodology’, version 2, June. Branstetter, L. 2006, ‘Is foreign direct investment a channel of knowledge spillovers? Evidence from Japan’s FDI in the United States’, Journal of International Economics, vol. 68, pp. 325-34. Bryant, J., Genç, M. and Law, D. 2004, ‘Trade and migration to New Zealand’, New Zealand Treasury Working Paper Series, 04/18. Buch, C., Kleinert, J. and Toubal, F. 2003, ‘Where enterprises lead, people follow? Links between migration and German FDI’, Kiel Institute for World Economics Working Papers, no. 1190. Butlin, M. 1977, ‘A preliminary annual database, 1900/01 to 1973/74’, Reserve Bank of Australia Research Discussion Paper Series, no. 7701. Carr, D., Markusen, J. and Maskus, K. 2002, ‘Estimating the knowledge-capital model of multinational enterprise’, American Economic Review, vol. 91, no. 3, pp. 693-708. Caves, R. 1974, ‘Multinational firms, competition and productivity in host-country markets’, Economica, vol. 41, no. 162, pp. 176-93. Co, C, Euzent, P and Martin, T 2004, ‘The export effect of immigration into the USA’, Applied Economics, vol. 36, pp. 573-83. Collins, W., O’Rourke, K. and Williamson, J. 1997, ‘Were trade and factor mobility substitutes in history?, NBER Working Paper Series, no. 6059.

80


Combes, P, Lafourcade, M and Mayer, T 2003, ‘Can business and social networks explain the border effect puzzle?’, CEPII Working Paper Series, 2003-02, January. Dawkins, P., Kemp, S. and Cabalu, H. 1995, Trade and Investment with East Asia in Selected Service Industries: The Role of Immigrants, Bureau of Immigration, Multicultural and Population Research, Canberra. Deardorff, A. 1998, ‘Determinants of bilateral trade: Does gravity work in a neoclassical world,’ in JA Frankel (ed.), The Regionalization of the World Economy, University of Chicago Press: Chicago and London. Department of Foreign Affairs and Trade (DFAT) 2002, ‘Estimated number of Australians overseas in 2001’, mimeo, http://www.southern-crossgroup.org/archives/Statistics/ —— 2004, Submission no. 646 to the Senate Inquiry into Australian Expatriates. Disdier, A. and Head, K. 2007, ‘The puzzling persistence of the distance effect on bilateral trade’, forthcoming in Review of Economics and Statistics. Dumont, J. and Lemaître, G. 2005, ‘Counting immigrants and expatriates in OECD countries: a new perspective’, OECD Social, Employment and Migration Working Papers, no. 25, June. Dunlevy, J. 2006, ‘The influence of corruption and language on the pro-trade effect of immigrants: Evidence from the American States’, Review of Economics and Statistics, vol. 88, no. 1, pp. 182-6. —— and Hutchinson, W. 1999, ‘The impact of immigration on American import trade in the late nineteenth and early twentieth centuries’, Journal of Economic History, vol. 59, pp. 1043–62. Eaton, J. and Kortum, S. 1999, ‘International technology diffusion: Theory and measurement’, International Economic Review, August 1999, vol. 40, no. 3, pp. 537-70. Eaton, J. and Tamura, A. 1994, ‘Bilateralism and regionalism in Japanese and U.S. trade and direct foreign investment patterns’, Journal of the Japanese and International Economies, vol. 8, no. 4, pp. 478-510. Easterly, W. and Levine, R. 2001, ‘It's not factor accumulation: stylized facts and growth models’, World Bank Economic Review, vol. 15, no. 2, pp. 177-219. Feenstra, R. 2004, Advanced International Trade: Theory and Evidence, Princeton University Press, Princeton. ——, Lipsey, R., Deng, H., Ma, A. and Mo, H. 2004, ‘World trade flows: 19622000’, NBER Working Papers, no. 11040. REFERENCES

81

Frankel, J. and Romer. D. 1999, ‘Does trade cause growth?’, American Economic Review, vol. 89, no. 3, pp. 379-99. Fullilove, M. and Flutter, C. 2004, ‘Diaspora: The world-wide web of Australians’, Lowy Institute Papers, no. 4. Girma, S. and Yu, Z. 2002, ‘The link between immigration and trade: Evidence from the United Kingdom’, Weltwirtschaftliches Archiv/Review of World Economies, vol. 138, no. 1, pp. 115-30. Gould, D. 1994, ‘Immigrant links to the home country: empirical implications for US bilateral trade flows’, The Review of Economics and Statistics, vol. 76, no.2, May, pp. 302-16. Grossman, M. 1998, Comment in Frankel, J. (ed.), The Regionalization of the World Economy, University of Chicago Press, Chicago and London. Guttman, S. and Richards, A. 2004, ‘Trade openness: An Australian perspective’, Reserve Bank of Australia Research Discussion Papers, no. 2004-11. Habib, M. and Zurawicki, L. 2002, ‘Corruption and foreign direct investment’, Journal of International Business Studies, vol. 33, no. 2, pp. 291-307. Hall, R. and Jones, C. 1999, ‘Why do some countries produce so much more output per worker than others?’, Quarterly Journal of Economics, vol. 114, no. 1, pp. 83-116. Haskel, J., Pereira, S. and Slaughter, M. 2002, ‘Does inward foreign direct investment boost the productivity of domestic firms?’, NBER Working Papers, no. 8724. Head, K. and Ries, J. 1998, ‘Immigration and trade creation: econometric evidence from Canada’, Canadian Journal of Economics, vol. 31, no. 1, pp .47-62. —— and —— 2001, ‘Overseas investment and firm exports’, Review of International Economics, vol. 9, no. 1, pp. 108-22. Head, K. and Mayer, T. 2002, ‘Illusory border effects: Distance measurement inflates estimates of home bias in trade’, CEPII Working Papers, no. 2002-01. Hejazi, W. and Safarian, A. 2001, ‘The complementarity between U.S. foreign direct investment stock and trade’, Atlantic Economic Journal, vol. 29, no. 4, pp. 420-37. Helliwell, J. 1997, ‘National borders, trade and migration’, Pacific Economic Review, vol. 2, no. 3, pp. 165-85. Helpman, E. 1984, ‘A simple theory of international trade with multinational corporations’, Journal of Political Economy, vol. 92, no. 3, pp. 451-71.

82


Herander, M. and Saavedra, L. 2005, ‘Exports and the structure of immigrant-based networks: The role of geographic proximity’, Review of Economics and Statistics, vol. 87, no.2, pp. 323-35. Hugo, G., Rudd, D. and Harris, K. 2001, ‘Emigration from Australia: Economic implications’, Committee for Economic Development of Australia (CEDA) Information Papers, no. 77. ——, —— and —— 2003, ‘Australia’s diaspora: Its size, nature and policy implications’, CEDA Information Papers, no. 80. Hutchinson, W. and Dunlevy, J. 2001, ‘The pro-trade effect of immigration on American exports during the period 1870 to 1910’, Vanderbilt University Department of Economics Working Papers, no. 0125. Javorcik, B., Özden, Ç., Spatareanu, M. and Neagu, C. 2006, ‘Migrant networks and foreign direct investment’, World Bank Working Paper Series, no. 4046 Jaffe, A., Trajtenberg, M. and Henderson, R. 1993, ‘Geographic localization of knowledge spillovers as evidenced by patent citations’, Quarterly Journal of Economics, vol. 108, no. 3, pp. 577-98. Keller, W. 2000, ‘Geographic localization of international technology diffusion’, NBER Working Paper Series, no. 7509. —— and Yeaple, S. 2003, ‘Multinational enterprises, international trade, and productivity growth: Firm-level evidence from the United States’, NBER Working Papers, no. 9504. Kulendran, N. and Wilson, K. 2000, ‘Is there a relationship between international trade and international travel?, Applied Economics, vol. 32, pp. 1001-9. Kugler, M. and Rapoport, H. 2007, ‘International labour and capital flows: complements or substitutes?, Economics Letters, vol. 94, pp. 155-62. Maddison, A. 2003, The World Economy: Historical Statistics, OECD Development Centre Studies, OECD, Paris. Manning, W. and Mullahy, J. 2001, ‘Estimating log models: To transform or not to transform’, Journal of Health Economics, vol. 20, pp.461-94. Markusen, J. 1984, ‘Multinationals, multi-plant economies and the gains from trade’, Journal of International Economics, vol. 16, no. 3-4, pp. 205-26. —— et al 1996, —— and Maskus, K. 1999a, ‘Multinational firms: Reconciling theory and evidence’, NBER Working Paper Series, no. 7163. —— and —— 1999b, ‘Discriminating among alternative theories of multinational enterprise’, Review of International Economics, vol. 10, no. 4, pp. 694-707. REFERENCES

83

MacGarvie, M. 2006 ‘Do firms learn from international trade?’, Review of Economics and Statistics, February 2006, vol. 88, no. 1, pp. 46-60. Mundell, R. 1957, ‘International trade and factor mobility’, American Economic Review, vol. 47, no. 3, pp. 321-35. Mundra, K. 2005, ‘Immigration and international trade: A semi-parametric empirical investigation’, Journal of International Trade and Economic Development, vol. 14, no. 1, pp. 65-91. OECD 2004, Internationalisation of Higher Education, OECD, Paris. Pakko, M. and Wall, H. 2001, ‘Reconsidering the trade-creating effects of a currency union’, Federal Reserve Bank of St Louis Review, September/October, pp.37-45. Park, R. 1966, ‘Estimation with heteroscedastic error terms’, Econometrica, vol. 34, pp. 888. Portes, R. and Rey, H. 2005, ‘The determinants of cross-border equity flows’, Journal of International Economics, vol. 65, no. 2, pp. 269-96. Productivity Commission 2006, Economic Impacts of Migration and Population Growth, Final Report, April. Rauch, J. 1999, ‘Network versus markets in international trade’, Journal of International Economics, vol. 48, no. 1, pp .7-35. —— 2001, ‘Business and social networks in international trade’ Journal of Economic Literature, vol. 39, pp. 1177-1203. —— and Trindade, V. 1999, ‘Ethnic Chinese networks in international trade’, NBER Working Papers, no. 7189. Redding, S. and Venables, A. 2004, ‘Economic geography and income inequality’, Journal of International Economics, vol. 62, pp. 53-82. Rigobon, R. and Rodrik, D. 2004, ‘Rule of law, democracy, openness, and income: Estimating the interrelationships’, NBER Working Papers, no. 10750. Rodrik, D. 2000, ‘How far will economic integration go?’, Journal of Economic Perspectives, 14, Winter, pp. 177-86. Rodriguez, F. and Rodrik, D. 2000, ‘Trade policy and economic growth: A skeptic’s guide to the cross-national evidence’, NBER Working Papers, no. 7081. Romalis, J. 2007, ‘Market access, openness and growth’, NBER Working Papers, no. 13048. Romer, P. 1990, ‘Endogenous technological change’, Journal of Political Economy, vol. 98, no. 5, pp. S71-102. 84


Saxenian, A. 2002, Local and Global Networks of Immigrant Professionals in Silicon Valley, Public Policy Institute of California, San Francisco. Silva, J. and Tenreyo, S. 2003, ‘Gravity-defying trade’, Federal Reserve Bank of Boston Working Paper Series, no. 03-01. Stromback, T. 1994, ‘Immigration and skill transfer’, Curtin University School of Economics and Finance Working Paper Series, no. 9408. Syverson, C. 2004, ‘Market structure and productivity: A concrete example’, Journal of Political Economy, vol. 112, no. 6, pp. 1181-222. Tong, S. 2005, ‘Ethnic networks in FDI and the impact of institutional development’, Review of Development Economics, vol. 9, no. 4, pp. 563-80. United Nations Committee on Trade and Development (UNCTAD) 2006, World Investment Report, UNCTAD, Geneva. Wagner, D., Head, K. and Ries, J. 2002, ‘Immigration and the trade of provinces’, Scottish Journal of Political Economy, vol. 49, no. 5, pp. 507-25. Wei, S. 2000, ‘Local corruption and global capital flows’, Brookings Papers on Economic Activity, no. 2, pp. 303-54. Weidenbaum, M. and Hughes, S. 1996, ‘Asia's bamboo network’, American Enterprise, vol. 7, no. 5, pp. 68-9. Winston, C. 1993, ‘Economic deregulation: Days of reckoning for microeconomists’, Journal of Economic Literature, vol. 31, no. 3, pp. 1263-89. Zucker, L. and Darby, M. 2006, ‘Movement of star scientists and engineers and high-tech firm entry’, NBER working papers, no. 12172, April.

REFERENCES

85