Competition Between Tax Havens: Does Geographic Distribution Matter?



Written by: Luisa Blanco and Cynthia Rogers

1. Introduction

Over the past decade, economic globalization has led to a significant increase in capital mobility and a corresponding increase in demand for tax haven operations. In fact, total liabilities for a sample of 20 tax havens grew by 154 percent between 1998 and 2006 (Kudrle, 2008). Fearing the harmful impacts of tax haven policies, OECD countries launched the 1998 Harmful Tax Competition (HTC) initiatives. In response to such concerns, research has focused on analyzing the efficacy of tax haven policies as well as the impacts of such tax competition on non-tax havens.

The impact of tax competition on the tax havens, themselves, is an important dimension of the international tax competition dialogue. The extent to which tax havens compete in a race to the bottom provides a basis for tax haven countries to join non-tax haven countries in support of tax policy coordination across the international community. However, if tax havens are involved in a race to the top, then tax havens would have little reason to seek international tax harmonization. The dynamics of tax competition between tax havens, however, have been largely ignored in both policy and research venues.

This paper analyzes tax competition between tax havens paying special attention to physical proximity of countries. On one hand, distance to the nearest tax haven acts as a buffer against competitive pressure. On the other hand, close proximity increases the potential for capital to spillover between tax havens. We investigate these effects using two measures: foreign direct investment (FDI) and the number of American firm subsidiaries present in a tax haven country. We also investigate the likelihood that the top 500 American firms will have operations in different tax havens and the potential spatial relationship of these location choices.

Using a panel analysis of 19 tax haven countries we find FDI inflows and the number of American affiliates are positively related to the distance to the closest tax haven.[ 1] We also find a negative relationship between FDI inflows of a tax haven and that of its nearest tax haven neighbor. Combined these findings imply that tax competition has negative (beggar-thy-neighbor type) effects on tax havens in terms of competing for mobile capital. Interestingly, we find a positive relationship between the number of American affiliates in a tax haven and its nearest tax haven neighbor. This supports the argument that agglomeration benefits are important when firms are deciding where to locate subsidiaries. A tax haven benefits by being closer to another tax haven that already has an American subsidiary.

This paper is organized as follow: Section 2 reviews the literature on tax havens and tax competition, Section 3 presents the data and methodology used to study competition between tax havens, Section 4 discusses the results obtained from the empirical analysis, and Section 5 concludes.

2. Defining Tax Havens and Tax Competition

According to the OECD, a sufficient and necessary condition for a country to be a tax haven is that it has little or no tax on relevant income (Ambrosanio and Caroppo, 2005). Under the HTC initiative, tax haven countries are characterized as countries that are unlikely to share information with other countries fur purpose of taxes, lack transparency, or have firms that do not have a substantial activity in the jurisdiction (Kudrle, 2008). Such negative connotations drive the HTC initiative assertion that the existence or tax havens leads to harmful tax competition. Hines (2007a), on the other hand, uses a more general (and perhaps more generous) characterization of tax havens: tax havens are small, well governed countries with low tax rates.[ 2] Besides tax advantages, tax havens offer multinational corporations (MNCs), access to good infrastructure and offshore financing activities (Hines 2005a). These countries also offer a relaxed regulatory system allowing firms and individuals access to business and bank secrecy (Palan, 2002).

Tax haven regimes are manifested in many different varieties across a heterogeneous group of countries. Kudrle (2008) highlights notable differences in geographic location and income. Desai et al. (2006a) notes the relevance of size differences among tax havens given that American MNCs use large and small tax havens for different purposes. Tax havens are further classified based on the differences in their preferential tax regimes (Ambrosanio and Caroppo, 2005; Palan, 2002): tax havens may have no income or corporate taxes (no-tax havens), low taxes (low-tax havens), no taxes on income of foreign sources (foreign-tax havens), or special tax privileges to certain industries or companies (special-tax havens).

Tax havens are also classified according to the main service provided (Kudrle and Eden, 2003): production, headquarters, sham, and secrecy havens. Production havens have low corporate taxes and grant special privileges so as to encourage foreign firms to produce in the country. Headquarters havens allow foreign firms to lower corporate tax earnings in the home jurisdiction. Shams havens allow corporations and individuals to take advantage of the lack of regulation. Secrecy havens are used by individuals and corporations to evade taxes through secrecy laws.

Tax havens target different tax regimes to specialize in specific sectors and to distinguish themselves from each other. Accordingly, strategic tax regime choice is an important component of tax competition between tax havens. Most research notes the relevance and size of tax haven operations. The volume of literature on the increasing demand of tax haven operations is large (Diamond and Diamond, 2002; Hines, 2005a,b; Hines and Rice, 1994; Kudrle, 2008), but a much smaller body of research investigates the impact of tax haven proximity on non-tax havens.

Recent studies suggest that proximity to tax havens can be beneficial for non-tax haven countries. Desai, Foley, and Hines (2007a,b) argue that increased activity in tax havens spills over to the benefit of nearby non-tax havens. Rose and Spiegel (2007) present evidence that proximity to tax havens serving as offshore financial centers (OFCs) leads to greater financial depth and competitiveness in the financial sector of a non-OFC country. Furthermore, Blanco and Rogers (2008) show that proximity to tax havens has a positive impact on the level of FDI of less developed countries. To our knowledge, however, the impact of proximity to a tax haven for a tax haven, however, has not been analyzed.

From a theoretical standpoint, the impact of tax competition on tax havens is ambiguous. In the basic models of tax competition, introduced by Oates (1972) and developed by Zodrow and Mieszkowski (1986), tax competition has a negative effect (See Wilson and Wildasin (2004) and Wilson (1990) for recent literature reviews).

Countries lower their tax rates to attract capital, which leads to a decrease in government spending and the underprovision of public goods. In contrast, tax competition in a Tiebout (1956) model can bring positive effects. In the extensions of the Tiebout Model, governments compete for mobile individuals who vote with their feet by locating in jurisdictions with the most favorable combination of taxes and public goods provision. This fiscal competition leads to an efficient provision of public goods and places pressure on governments to keep taxes low.

Related models of tax competition promote the idea that tax competition “tames the leviathan” and leads to a more efficient allocation of resources (Wilson and Wildasin, 2004). For example, Honkapohja and Turunen-Red (2007) argue that tax competition creates strong incentive to expand output. Hong and Smart (2006) posit that tax planning can be socially optimal. Eggert and Sorensen (2008) use a theoretical model to show that the efficiency of the public sector increases with more tax competition.

Agglomeration models, in contrast, suggest that tax competition does not necessarily lead to lower taxes. Under Baldwin and Krugman’s (2000) model some countries do not need to reduce taxes since high taxes will be offset by the benefits of industrial agglomeration. In these models, tax competition leads to a “race to the top” and not to a “race to the bottom”.

Empirical evidence shows that tax revenues have not declined with the fall in corporate taxes over time due to the broadening of tax bases (Hines, 2005b; Devereux et al., 2002). It is argued that tax competition and levels of FDI are related, where countries with lower taxes tend to attract more FDI (Hines, 2007b). In addition, higher taxes are associated with lower levels of affiliate assets and output (Desai et al., 2006a). Mutti (2003) argues that FDI inflows resulting from lowering tax rates have significant positive externalities on the economy, such as technology transfers and greater capital accumulation. These positive spillovers allow the economy to grow at a faster rate.

To summarize, the literature review highlights two main points. First, the impact of tax competition on heterogeneous tax havens is ambiguous. Second, there is an established link between tax haven policies and FDI inflows. For the purposes of our analysis, countries are defined to be tax havens following Dharmapala and Hines (2006).[ 3]

We analyze tax competition impacts on tax havens using two approaches, both of which consider proximity of a tax haven and its closest tax haven neighbor. Specifically, we investigate how distance between a tax haven and its closest tax haven neighbor affects (i) FDI inflows and (ii) the presence of American MNCs.

3. Estimation of FDI inflows in Tax Havens

To investigate tax competition effects on FDI inflows, we use 5 year average FDI observations between 1991 and 2005 for a sample of 19 tax havens.[ 4] Our baseline specification is as follows:

Competition between Tax Havens Does Geographic Distribution Matter_html_25e6b5e5

The dependent variable is the 5 year average of the natural logarithm of FDI inflows.[ 5] FDI inflows were obtained from the United Nations Conference on Trade and Development website. The tax haven proximityvariable (TH_proximity) is the inverse of the natural logarithm of the d i stance to the nearest tax haven. It is constructed using distances calculated by Mayer and Zignago (2006) for 31 of the 35 the tax havens identified inAppendix 1.[ 6] A positive estimated coefficient on TH_proximity indicates that FDI inflows increase as distance to the closest tax haven decreases. Therefore, a positive (negative) coefficient implies that being closerto another tax haven is beneficial (harmful) and that tax havens are good (bad) neighbors for other tax havens.

The vector X represents a set of control variables identified in previous empirical analyses as important determinants of FDI.[ 7] In the baseline model, the control variables include the 5 year average of the naturallogarithm of population, exchange rate, and openness to trade. The natural logarithm of the initial level of real GDP per capita in the 5 year period is also included as control variable.[ 8] A description of the variablesand their sources is presented in Table 1 and summary statistics are given in Table 2. The model is estimated using the ordinary least squares (OLS) estimator with White’s robust standard errors and time fixed effects.[ 9]

Given the potential for structural instability in the form of non-constant error variances, we estimate a spatial error model with robust standard errors to check for robustness. The spatial error model corrects for thepotential bias resulting from the possibility that FDI inflows in one country may be dependent on the FDI inflows of nearby countries .[ 1 0] The error term in the spatial error m odel is specif i ed as follows:

Competition between Tax Havens Does Geographic Distribution Matter_html_8f86186

matrix. W is constructed following Blonigen et al. (2007), where W is made by TxT matrices of dimension IxI (T represents the number of periods and I the number of countries).[ 1 1] The diagonal matrices are symmetricmatrices of the ratio of the shortest bi- lateral distance in the sample and the bi-lateral distance from country j to country k, where the w eight for the countries with the shortest distance is equal to 1 .[ 1 2] The other matrices thatcompose W are matrices of zeros of dimensions IxI.

The spatial error model is estimated with the maximum likelihood estimator (MLE) and W is normalized so that each row sums to unity.[ 1 3] Wald test will be used to determine whether errors are spatially correlated.[ 14]

Table 4 shows the estimates from the baseline model using OLS. Estimates in column 1 show that the coefficient for tax haven proximity is negative and statistically significant at the 1 percent level. This indicates thatthe closer a tax haven is to the nearest tax haven neighbor, the lower its FDI inflows. This result is robust to including a variable that controls for institutions (Colu m n 2) .[ 1 5] Based on esti m ates from colu m n 1, if the distance tothe closest tax haven increases by one standard deviation (713 kilometers), then the natural log of FDI inflows increases by 0.2, which is approximately 1.23 million US dollars.

Another way to look at tax competition between tax havens is to look at the level of FDI in the closest tax haven which helps to identify potential spillovers. In Table 4, columns 3 and 4 include the natural logarithm ofFDI inflows to the closest tax haven as an explanatory variable instead of proximity to the nearest tax haven. The specification in column 4 includes a variable to account for institutions. The estimated negative coefficients on the level of FDI are statistically significant at the 1 and 5 percent level, for column 3 and 4, respectively. Based on the estimates in column 3, a decrease of 1 standard deviation of the natural log of FDI inflows in the closest tax haven would lead to a predicted increase in FDI inflows by 1.24 million US dollars.

Table 5 shows the spatial error model estimates corresponding to the models in Table 4. Again columns 2 and 4 add the institutions variable to the models in columns 1 and 3, respectively. In all these estimations, we fail to reject the hypothesis of no spatially error correlated term at the 5 percent, which implies that OLS estimates are not biased.

From the estimates shown in Table 4, it can be concluded that being close to another tax haven is a bad thing for tax havens. This evidence shows that tax competition between tax havens negatively affects FDI inflows. This is an interesting result since it shows the importance of geographic location when tax havens are competing with each other for FDI.

4. Estimation of American MNCs Activity in Tax Havens

We also analyze competition between tax havens by looking at the impact of tax haven proximity on the activity of American MNCs in a specific tax haven. For this part of the analysis, we use our baseline modelspecified in equation 1, where the dependent variable is the number of American MNCs affiliates in a tax haven and observations are annual for the years 1999 to 2005.[ 1 6] The number of affiliates of American MNCs was obtained from the US Bureau of Economic Analysis. We estimate the baseline model using the OLS estimator with time fixed effects and the MLE estimator with time fixed effects for the spatial error model.[ 1 7] Summary statistics are shown in Table 3.

Table 6 presents the estimates using observations on American MNCs in tax havens. As shown in columns 1 and 2, the estimated coefficients on tax haven proximity are negative and significant at the 1 percent level.Based on column 1 estimates, increasing the distance to the closest tax haven by one standard deviation (726 kilometers) results in an estimated increase in the natural log of number of affiliates by 0.15; representinga predicted increase of one affiliate (approximately 1.16).

Columns 3 and 4 show the estimates when the natural log of affiliates in closest tax haven is included as independent variable instead of the tax haven proximity variable. The coefficient on the affiliates variable ispositive and significant at the 1 percent level (Table 6, columns 3 and 4). This is an interesting finding since it seemingly contradicts the previous findings where being closer to the nearest neighboring tax haven was associated with less economic activity in a Tax Haven. According to agglomeration models of tax competition, however, firms tend to concentrate in specific areas. Estimates from column 3 suggest that as if there is a increase in the natural log of the number of affiliates in the closest tax haven by one standard deviation (2.5), then the number of affiliates in a tax haven increases by 0.25; which also represents an increase of one affiliate (approximately 1.28).

The estimates obtained from the spatial error model using American MNC data are shown in Table

7. We reject the hypothesis of no spatially correlated error terms at the 1 percent level. Controlling for spatial correlation of the error term, the estimates are consistent with those found using the OLS estimator: thetax haven proximity variable has a significant negative effect on the number of affiliates, and the number of affiliates in the closest tax haven has a positive and significant effect on the number of affiliates. Themagnitude of the effect of these variables is similar in magnitude to those based on the OLS estimator. Estimates from column 1 (column 3) show that when the distance to the closest tax haven increases (number of affiliates in the closest tax haven) by one standard deviation, there is an increase of the number of affiliates by 1.16 (1.42).

Geography plays an important role in reconciling the different impacts of American affiliates. The combined results suggest that tax havens compete with each other to attract MNCs (thus the negative impact of proximity of nearest tax haven on affiliate levels). However there are positive agglomeration effects: tax havens with higher levels of affiliates in the closest tax haven neighbor have higher levels themselves. Thus, if a firm is already in a tax haven and it wants to expand, then it is likely that the firm will be interested on having an affiliate in the closest low tax jurisdiction (i.e. tax haven).

To study these contradictory results in more depth, we investigate the presence of the top 500 American companies in tax havens as specified by the Forbes 500 list of 2008. We look at the number of affiliates thateach of these companies has in a tax haven according to the directory of American firms operating in foreign countries.[ 18]

Of the top 500 American firms, 46 percent has at least one affiliate in a tax haven, and 36 percent has an affiliate in more than one tax haven. In fact, the probability of having more than one tax affiliate, given that thefirm has one tax haven affiliate, is equal to 77 percent. The tax havens with the highest number of American subsidiaries are Hong Kong, Singapore, Switzerland, and Ireland (in descending order).[ 1 9] There is anapparent synergy between tax haven subsidiaries in these locations. Given that the firm has a subsidiary in Singapore, there is a 79 percent of probability of having a subsidiary in Hong Kong; and given that the firm has a subsidiary in Hong Kong, there is a 77 percent of probability that the firm has a subsidiary in Singapore. In the case of Ireland and Switzerland, we find similar results. Given that the firm has a subsidiary in Ireland,there is a 66 percent of probability of having a subsidiary in Switzerland; and given that the firm has a subsidiary in Switzerland, there is a probability of 58 percent that the firm has a subsidiary in Ireland.

4. Conclusion

To summarize, there are two main empirical findings from this analysis. First, proximity to the closest neighboring tax haven is negatively related to FDI inflows and the number of American subsidiaries in a tax haven.In addition, an increase in the amount of FDI inflow in the closest nearby tax haven has a negative effect on the amount of FDI inflows in tax havens. These results suggest that there is race to the bottom between taxhavens, where they all compete for mobile capital and geographic location matters. The policy implication is that as a group, tax haven countries experience potentially harmful impacts from tax competition and havereason to support initiatives aimed at international tax harmonization.

Our second main finding is that agglomeration benefits are evident for firms when they are deciding where to locate their subsidiaries. As the number of affiliates in the closest nearby tax haven increases, there is anincrease in the number of affiliates in tax havens. This result suggests that there are positive spillovers among neighboring tax havens. Given that a firm is already located in a tax haven, there is a higher probability that it will expand operations to a nearby tax haven. So tax havens may benefit by being nearer to other tax havens with high presence of American firm affiliates.

There are many fruitful avenues for further research. In particular, extensions should explore the implications of tax competition between tax haven regimes with different characteristics. As discussed in theliterature review, the group of tax havens is a heterogeneous group and determining the relevance of location for competition between these countries is important. In addition, a future research might evaluate notonly the location of American MNCs but also their level sales, employment or assets in tax havens. Such factors are undoubtedly important for gaining a better understanding of the role of tax competition and geography for tax havens.

References

Anselin, Luc. (1999). Spatial econometrics. Center for Spatially Integrated Social

Science, http://csiss.ncgia.ucsb.edu/learning_resources/content/papers/

Anselin, Luc, Le Gallo, Julie, and Jayet, Hubert. (2008). Spatial Panel Econometrics. In Matyas, Laszlo and Sevestre, Patrick (Eds.), The econometrics of panel data, Springer 3 rd Edition, Berlin.

Ambrosanio, Maria, and Caroppo, Maria. (2005). Eliminating harmful tax practices in tax havens: defensive measures by major EU countries in tax haven reforms. Canadian Tax Journal, 53(3), 685-719.

Baldwin, Richard, and Krugman, Paul. (2000). Agglomeration, integration and tax harmonization. European Economic Review, 48, 1-23.

Blanco, Luisa and Rogers, Cynthia. (2008). Are tax havens good neighbors? An LDC perspective. Working paper.

Blonigen, Bruce. (2005) A review of the empirical literature on FDI determinants. Atlantic Economic Journal, 33(4), 383-303.

Blonigen, Bruce, Davies, Ronald, Waddell, Glen, and Naughton, Helen. (2007). FDI in space: Spatial autoregressive relationships in foreign direct investment. European Economic Review, 51(5), 1303-1325.

Bureau of Economic Analysis. (2008). U.S. Direct Investment Abroad http://www.bea.gov/international/xls/usdiacap.xls

Desai, Mihir, Foley, Fritz, and Hines, James. (2006a). The Demand for Tax Haven Operations. Journal of Public Economics, 90(3), 513-531.

Desai, Mihir, Foley, Fritz, and Hines, James. (2006b). Do Tax Havens Divert Economic Activity? Economic Letters, 90(2), 219-224.

Devereux, Michael, Griffith, Rachel and Klemm, Alexander. (2002). Corporate income tax reforms and international tax competition. Economic Policy, Vol.17(35), 451-

495.

Dharmapala, Dhammika and James Hines. (2006). Which countries become tax havens? National Bureau of Economic Research, Working paper No. 12802.

Diamond, Walter and Diamond, Dorothy. (2002). Tax Havens of the World. Matthew Bender Books, New Jersey.

Directory of A m erican fir m s operating in foreign countries. (2007). Uniworld Business Publications, 19 th Edition, Massachusetts.

Eggert, Wolfgang and Sørensen, Peter. (2008). The effects of tax competition when politicians create rents to buy political support. Journal of Public Economics, 92, 1142–1163

Elhorst, Paul. (2003). Specification and estimation of spatial panel data models. International Regional Science Review, 26(3), 244-268.

Hines, Ja m e s. (2007a). Tax Havens in Lawrence Blu m e and Steven Durlauf (eds.) The New Palgrave Dictionary of Economics Online, 2 n d edition.

Hines, Ja m e s. (2007b). Taxation of fore i gn inco m e in Lawr e nce Blu m e a nd Steven Durlauf (eds.) The New Palgrave Dictionary of Economics Online, 2 n d edition.

Hines, James. (2005a). Do tax havens flourish? in James Poterba (ed.) Tax Policy and the Economy 19. Cambridge, Massachusetts: NBER and MIT Press.

Hines, James. (2005b). Corporate taxation and international competition. Ross School of Business working paper, No. 1026.

Hines, James and Rice, Eric. (1994). Fiscal paradise: foreign tax havens and American business. Quarterly Journal of Economics, 109(1), 149-182.

Hong, Qing and Smart, Michael. (2006). In praise of tax havens: International tax planning and foreign direct investment. Working paper.

Honkapohja, Seppo and Turunen-Red, Arja. (2007). Increasing returns, learning, and beneficial tax competition. Journal of Public Economic Theory, 9(6), 927–958.

Kaufmann, Daniel, Kraay, Aart, and Mastruzzi, Massimo. (2008). Worldwide

Governance Indicators Project (online).http://info.worldbank.org/governance/wgi/

Kudrle, Robert (2008). The OECD’s harmful tax competition initiative and the tax havens: from bombshell to damp squid. Global Economy Journal, 8(1), article 1.

Kudrle, Robert and Eden, Lorraine. (2003). The campaign against tax havens: will it last?

Will it work? Stanford Journal of Law, Business, and Finance, 9, 37-68.

LeSage, James. (1999). Spatial econometrics, Web Books of Regional Science, Regional

Research Institute. http://www.rri.wvu.edu/regscweb.htm

Mayer, Thierry and Zignago, Soledad (2006) CEPII Distance data. Centre D’Etudes Prospectives et D’Informations Internationales (online). http://www.cepii.fr/anglaisgraph/bdd/distances.htm

Mutti, John. (2003). Foreign direct investment and tax competition. Institute for International Economics, Washington, DC Oates, W.E. (1972). Fiscal Federalism. Harcourt Brace Jovanovich, New York. Palan, Ronen.(2002). Tax havens and the commercialization of state sovereignty.

International Organization, 56(1), 151-171.

Rose, Andrew and Spiegel, Mark. (2006). Offshore financial centers: parasites and symbionts? Economic Journal, 117(523), 1310-1335.

Tiebout, C.M. (1956). A pure theory of local expenditures. Journal of Political Economy

64, 416– 424.

United Nations Conference on Trade and Development. (2008). Foreign direct investment database (online). http://www.unctad.org/Templates/Page.asp?intItemID=1923&lang=1 Wilson, John. (1999). Theories of taxcompetition. National Tax Journal, 52(2), 269-304. Wilson, John and Wildasin. (2004). Capital tax competition: bane or boon. Journal of Public Economics, 88, 1065– 1091

World Bank. (2008). Data and Statistics (online). http://go.worldbank.org/D7SN0B8YU0

Zodrow, G.R., Mieszkowski, P. (1986). Pigou, Tiebout, property taxation, and the underprovision of local public goods. Journal of Urban Economics, 19, 356– 370.

Table 1. Variable description and source

Variable name

Variable description and source

Ln(FDI)

Natural log of the inflow of foreign direct investment (US dollars, m illions). Source: UNCTAD.

Ln(Affiliates)

Natural log of the number of affiliates of American MNCs.

Source: BEA.

TH proximity

Inverse of the natural log of the distance (in kilometers) to the closest tax haven (as defined by Dharmapala and Hines, 2006, see Appendix 1). We use the distance calculated with the latitudes and longitudes of the most important cities and agglomerations. Source: Mayer and Zignago (2006).

Ln(FDI in closest TH)

Natural log of the inflow of foreign direct investment in closest tax haven (US dollars, millions). Source: UNCTAD.

Ln(Affiliates in closest TH)

Natural log of number of affiliates of American MNCs in closest tax haven. Source: BEA.

Ln(exchange rate)

Natural log of exchange rate. Source: WDI.

Ln(GDP initial)

Natural log of the initial level of real GDP per capita (constant

2000 US dollars). Source: WDI.

Ln(openness)

Natural log of exports plus imports divided by real GDP.

Source: WDI.

Ln(population)

Natural log of population. Source: WDI.

Institutions

Average of six different indicators of governance (rule of law, control of corruption, voice and accountability, government effectiveness, political stability, and regulatory quality). Source: Kaufmann, Kraay, and Mastruzzi (2007).

 

Variablename

Variabledescriptionandsource

Ln(FDI)

Natural log of the inflow of foreign direct investment (US dollars, m illions). Source: UNCTAD.

Ln(Affiliates)

Naturallogof thenumberof affiliatesof AmericanMNCs.

Source: BEA.

TH proximity

Inverse of the natural log of the distance (in kilometers) to the closest tax haven (as defined by Dharmapala and Hines, 2006, see Appendix 1). We use the distance calculated with the latitudes and longitudes of the most important cities and agglomerations. Source: Mayer and Zignago (2006).

Ln(FDI in closest TH)

Natural log of the inflow of foreign direct investment in closest tax haven (US dollars, millions). Source: UNCTAD.

Ln(Affiliates in closest TH)

Natural log of number of affiliates of American MNCs in closest tax haven. Source: BEA.

Ln(exchangerate)

Naturallogof exchangerate.Source:WDI.

Ln(GDP initial)

Natural log of the initial level of real GDP per capita (constant

2000 US dollars). Source: WDI.

Ln(openness)

Natural log of exports plus imports divided by real GDP.

Source: WDI.

Ln(population)

Naturallogof population.Source:WDI.

Institutions

Average of six different indicators of governance (rule of law, control of corruption, voice and accountability, government effectiveness, political stability, and regulatory quality). Source: Kaufmann, Kraay, and Mastruzzi (2007).

Table 2. FDI in tax havens summary statistics

Mean

Max

Min

Std. Dev.

Ln(FDI)

5.147

10.412

1.468

2.433

TH proximity

0.187

0.244

0.126

0.039

Ln(FDI in closest TH)

5.318

10.412

0.000

3.046

Ln(GDP initial)

9.700

15.809

6.849

2.003

Ln(population)

13.114

15.808

10.631

1.656

Ln(exchange rate)

1.125

7.332

-1.050

1.939

Ln(openness)

4.876

5.792

4.040

0.352

Institutions

0.677

1.796

-0.369

0.591

Table 3. American MNCs activity in tax havens summary statistics

Mean

Max

Min

Std. Dev.

Ln(Affiliates)

2.353

6.314

0.000

2.165

TH proximity

0.183

0.244

0.123

0.041

Ln(Affiliates in closest TH)

3.233

6.314

0.000

2.468

Ln(GDP initial)

9.602

15.773

4.958

2.228

Ln(population)

13.298

15.822

10.652

1.694

Ln(exchange rate)

1.311

7.318

-1.065

2.021

Ln(openness)

4.841

5.953

3.860

0.413

Institutions

0.559

1.830

-1.596

0.784

Table 4. Ordinary Least Squares Estim a tes: FDI five-year averages (N=57) 1

Independent Variable (1) (2) (3) (4)

TH proximity

-11.605 ***

(4.416)

-11.852 ***

(4.672)

Ln(FDI in closest TH)

-0.068 ***

(0.024)

-0.087 **

(0.036)

Ln(GDP initial)

1.183 ***

(0.071)

0.933 ***

(0.061)

1.114 ***

(0.066)

0.857 ***

0.085)

Ln(population)

0.501 ***

(0.149)

0.529 ***

(0.145)

0.662 ***

(0.138)

0.708 ***

(0.134)

Ln(exchange rate)

-1.118 ***

(0.125)

-0.797 ***

(0.076)

-1.158 ***

(0.150)

-0.836 ***

(0.085)

Ln(openness)

1.375 ***

(0.141)

1.423 ***

(0.209)

1.018 ***

(0.112)

1.034 ***

(0.050)

Institutions

0.704 ***

(0.256)

0.753 ***

(0.283)

Constant

2.41E-11

(5.26E-11)

2.50E-11

(5.43E-11)

1.79E-11

(4.15E-11)

1.81E-11

(4.23E-11)

R-squared 0.772 0.785 0.749 0.762

1 Robust standard errors are in parenthesis.

*, **, and *** indicate significance at 10, 5 and 1 percent level, respectively.

Table 5. Spatial Error Model Estimates: FDI five-year averages (N=57) 1

Independent Variable (1) (2) (3) (4)

TH proximity

-11.465*** (4.628)

-11.606*** (4.315)

Ln(FDI in closest TH)

-0.067

(0.097)

-0.083

(0.099)

Ln(GDP initial)

1.189***

(0.183)

0.942***

(0.229)

1.138***

(0.188)

0.881*** (0.258)

Ln(population)

0.503***

(0.086)

0.536***

(0.083)

0.657***

(0.145)

0.702*** (0.149)

Ln(exchange rate)

-1.122***

(0.161)

-0.799***

(0.217)

-1.171***

(0.169)

-0.850*** (0.210)

Ln(openness)

1.390***

(0.362)

1.458***

(0.354)

1.082**

(0.462)

1.094**

(0.475)

Institutions

0.714**

(0.338)

0.740*

(0.416)

Constant

0.001

0.002

0.005

0.004

(0.158)

(0.158)

(0.176)

(0.169)

Lambda (λWε)

0.031

0.071

0.082

0.077

(0.210)

(0.199)

(0.211)

(0.215)

Log-likelihood

Wald statistic (Chi-square)

-86.870

0.021

-85.240

0.127

-89.625

0.151

-88.043

0.128

1 Robust standard errors are in parenthesis.

*, **, and *** indicate significance at 10, 5 and 1 percent level, respectively.

Table 6. Ordinary Least Squares Estimates: Annual American affiliates (N=133) 1

Independent Variable

(1)

(2)

(3)

(4)

TH proximity

-8.559***

(0.379)

-7.948***

(0.367)

Ln(Affiliates in closest TH)

0.100***

(0.017)

0.061***

(0.016)

Ln(GDP initial)

0.225***

(0.012)

-0.271***

(0.022)

0.127***

(0.008)

-0.343***

(0.018)

Ln(population)

0.867***

(0.017)

0.893***

0.011)

0.945***

(0.017)

0.964***

(0.011)

Ln(exchange rate)

-0.301***

(0.017)

0.325***

(0.025)

-0.232***

(0.010)

0.359***

(0.024)

Ln(openness)

1.357***

(0.046)

0.837***

(0.043)

1.163***

(0.033)

0.673***

(0.028)

Institutions

1.803***

(0.087)

1.772***

(0.072)

Constant

9.10E-10

8.27E-10

6.97E-12

-1.08E-11

(4.24E-11)

(3.65E-11)

(4.45E-12)

(6.40E-12)

R-squared 0.700 0.778 0.693 0.767

1 Robust standard errors are in parenthesis.

*, **, and *** indicate significance at 10, 5 and 1 percent level, respectively.

Table 7. Spatial Error Model Estimates: Annual American Affiliates (N=133) 1

Independent Variable

(1)

(2)

(3)

(4)

TH proximity

-8.664 ***

(3.087)

-8.330 ***

(2.448)

Ln(Affiliates in closest TH)

0.139 ***

(0.040)

0.090 ***

(0.033)

Ln(GDP initial)

0.234 ***

(0.089)

-0.378 ***

(0.105)

0.120 *

(0.068)

-0.447 *** (0.109)

Ln(population)

0.913 ***

(0.089)

1.060 ***

(0.062)

1.020 ***

(0.071)

1.146 ***

(0.057)

Ln(exchange rate)

-0.333 ***

(0.099)

0.495 ***

(0.165)

-0.252 ***

(0.088)

0.510 ***

(0.171)

Ln(openness)

1.929 ***

(0.360)

1.090 ***

(0.204)

1.818 ***

(0.329)

0.985 ***

(0.216)

Institutions

2.371 ***

(0.401)

2.259 ***

(0.416)

Constant

0.002

0.002

0.003

0.003

(0.157)

(0.172)

(0.169)

(0.174)

Lambda (λWε)

0.378

0.544

0.424

0.537

(0.147)

(0.108)

(0.139)

(0.113)

Log-likelihood

Wald statistic (Chi-square)

-206.864

6.628

-180.466

25.415

-207.573

9.306

-184.635

22.751

1 Robust standard errors are in parenthesis.

*, **, and *** indicate significance at 10, 5 and 1 percent level, respectively.

Appendix 1. Countries Class i fied as Tax havens by Darmaphala and Hines (2006)

Andorra

Channel Islands*

Lebanon

Netherlands Antilles

Anguilla

Cook Islands

Liberia

Panama

Antigua and Barbuda

Cyprus

Liechtenstein*

St. Kitts and Nevis

Bahamas

Dominica

Luxembourg

St. Lucia

Bahrain

Gibraltar

Macao

St. Vincent and Gren.

Barbados

Grenada

Maldives

Singapore

Belize

Hong Kong

Malta

Switzerland

Bermuda

Ireland

Marshall Islands

Turks and Caicos

British Virgin Islands

Isle of Man*

Monaco*

Vanuatu

Cayman Islands

Jordan

Montserrat

*Distance data from Mayer and Zignago (2006) is not available.

Appendix 2. Sample of Tax Havens use in Estimates*

Antigua and Barbuda

Grenada

Liberia i

St. Kitts & Nevis

Bahrain

Barbados ii

Hong Kong

Ireland

Luxembourg

Macao

St. Lucia

St. Vincent and Gren.

Belize

Jordan

Malta

Switzerland

Dominica

Lebanon

Panama

Vanuatu

 

 

i Not included in esti m ates of FDI inflows

ii Not included in estimates using American MNC affiliates.

*Both samples include 19 countries total.

[*]Acknowledgement: We thank Josephine Huang for excellent research assistance. Send correspondence to Luisa Blanco at lblanco@pepperdine.edu

[1] As described below, the tax havens in the sample vary slightly across specifications. FDI estimates use 5 year averages from 1991 to 2005 and the American MNC affiliates estimates use annual data from 1999 to

[2] Dharmapala and Hines (2006) present empirical evidence supporting the argument that tax havens have good institutions such as voice and accountability, rule law, government effectiveness, political stability and control of corruption.

[3 ] A pp en d ix 1 sh ow s t h e cou n tries i d e n tified as tax h a v e n s b y D h ar m a p ala a n d H i n es (200 6 ).

[4 ] Averages are constructed for available observation, yielding 3 observations per country and 57 observations total. Other tax havens were not included in the sample due to data unavailability.

[5] 5 year averages of the natural logarithm of FDI were constructed with the available observations from 1991 to 2005.

[6] Mayer and Zignago (2006) use the great circle formula to calculate the distance from the most important cities and agglomerations of one country to another. See their note on the calculations athttp://www.cepii.fr/distance/noticedist_en.pdf. Distance is missing for 4 of the tax haven countries identified in Appendix 1. Our approach to determine the impact of proximity to a tax haven is similar to Rose’s and Spiegel (2006) approach.

[7] Blon i gen ( 2 00 5 ) prese n ts a c o m p rehensi v e l iterat u re re v iew o n t h e d etermi n a n ts of FDI.

[8] Although several empirical analyses use the current level of GDP as a control variable, we use the initial level of GDP because FDI and GDP may be simultaneously determined.

[9] Time fixed effects are eliminated by subtracting the mean of each variable calculated for each country- year. Country fixed effects could not be included since the distance to a tax havens is time invariant.

[10] See Anselin (1999) and Anselin et al. (2008) for reviews of spatial econometric models. Elhorst (2003) discusses f i xed effects spatial err o r m odels.

[11] In t h is part of t h e analysis W is co m posed by 9 m atrices of di m ension 19 x 1 9 .

[12] We use the distance from the most important cities and agglomerations of one country to another (provided by Mayer and Z i g n ago, 20 0 6) t o construct o u r weighti n g m at r i x , W.

[13] Our estimates show robust standard errors since the estimator of variance uses the Huber/White estimator instead of the traditional calculation.

[14] See LeSage (1999) for a good review on how to test for the presence of spatially correlated errors.

[15] See Table 1 for a description of how this variable is constructed. In this part of the analysis, this index is constructed as the average for the whole period using observations between 1996 and 2005 (observation for earlier years are not available).

[16] Sample period was selected based on data availability. 19 countries were included, where selection was based on data availability. See Appendix 2 for a list of countries included. Here we have 7 observations per country, with a total of 133 observations.

[17] The weighting matrix W is constructed as mentioned above. For this part of the analysis, W is composed by 49 matrices of dimension 19×19.

[18] This directory is provided by the Uniworld Business Publications, where the 19 th edition published in 2 007 is t h e latest e d iti o n.

[19] In Hong Kong there are 158 firms that have an affiliate, in Singapore 154, in Switzerland 125, and Ireland 111. These four destinations are important since in the next most attractive destination, Luxemburg, there is only presence of 37 firms.

Previously published by the Pepperdine University – School of Public Policy, November 2008

Abstract

We investigate competition between tax havens and how this competition is related to geographic distribution. We study the extent to which proximity to the nearest tax haven affects foreign direct investment and the number of American affiliates in a tax haven. Empirical results show that distance to the nearest tax haven is positively related to FDI inflows and the number of American affiliates in tax havens. These findings suggest that tax havens compete with each other in a potentially harmful manner. Interestingly, we also find evidence of positive spillovers: the number of American affiliates in a tax haven and its closest tax haven neighbor is positively related. This finding suggests that agglomeration benefits are important for firms and that the nature of the competition between tax havens changes once there is a subsidiary in the nearby tax haven.

JEL classification codes: H87, F21

Keywords: Tax havens, Tax Competition, Foreign Direct Investment, Spatial

Econometrics

  • Was this helpful ?
  • Yes   No
 Assistant Professor, Pepperdine University School of Public Policy, Malibu, USA