Unlock tax planning solutions to

maximize your investment results.
achieve your financial goals.
comply with regulations.

Leverage a proprietary dataset from 190+ countries, Artificial Intelligence, and a network of top local experts.


Why Tax Incentives may be an ineffective tool to encouraging Investment? The role of Investment Climate

Stefan Van Parys,*

Sebastian James

Abstract:

In this paper we first analyze theoretically how the investment climate affects the impact of the corporate tax rate on investment. We do so in a model where the tax revenues are used to improve the investment climate. We find that if the investment climate is very effective at enhancing the productivity of capital, capital can react positively to a rise of the tax rate. However, this result is only true if a higher tax rate yields extra tax revenues. We also determine the conditions under which an improvement of the investment climate increases the opposition of capital to a rise in the tax rate. This is only the case if the investment climate is very complementary to capital or if the complementarity of the investment climate is much higher at low levels of the investment climate than at high levels of the investment climate. In the end, whether an improved investment climate increases or decreases the opposition of capital to a rise in the tax rate is an empirical question.

We test the model using a cross section of 80 countries ranging from very bad to very good investment climate countries. With FDI as a percentage of GDP as the dependent variable, a unique series of Marginal Effective Tax Rates as the tax variable and several investment climate variables are used to find two important results. First, an improvement of the investment climate increases the opposition of capital to a rise in the tax rate. Second, in the worst investment climate countries capital reacts not negatively to a rise in the tax rate.

These results lead to the important policy implication that for bad investment climate countries it is ineffective to lower the tax rate to compensate for the bad investment climate. Instead, they should focus on improving the basic investment climate.

Keywords: tax competition, investment climate, developing countries, Foreign Direct

Investment

1. Introduction

The neoclassical investment theory, pioneered by Jorgenson (1963), predicts that lowering the tax burden – through a drop in the user cost of capital – increases investment. This prediction has been tested many times[1]. The results of the empirical research on the relationship between investment and taxation are mixed, depending on how the data are defined, which periods and regions are covered, the methodology used, etc (Devereux 2006). In their Meta analysis De Mooij and Ederveen (2003) find a median tax rate elasticity of foreign capital of -3.3, indicating that the majority of studies find a negative relationship. Their Meta analysis offers interesting insights in how conceptual issues explain the heterogeneity of evidence. However, in this paper we investigate how differences in the investment climate (IC) between countries can moderate the relationship between investment and taxation. More precisely we focus on that part of the investment climate that can be directly affected by government policy, and that is crucial for the productivity of capital. Concretely, we examine the impact of public goods such as the regulatory quality, the rule of law, basic infrastructure, etc. on the effect of corporate taxation on investment.

We want to provide policy makers and policy advisers with better insights on the extent to which the investment climate moderates the relationship between investment and tax. Two quotes from policy notes of the OECD and the World Bank demonstrate that this is a relevant topic:

“There is a broad agreement that a low host country tax burden cannot compensate for a generally weak or unattractive FDI environment… Tax is but one element and cannot compensate for weak non-tax conditions” (OECD 2007).

“…tax incentives are a poor instrument for compensating for negative factors in a country’s investment climate” (World Bank, Morisset 2003).

The rationale behind it is that lowering the tax rate is effective in attracting investment if the investment climate is good, but not if the investment climate is bad. These statements are more based on anecdotic evidence and surveys than on theory and empirical evidence. The literature section of this paper shows that the little existing empirical evidence on this subject seems to be mixed. Common to the empirical papers is that the theoretical framework underlying the interaction between investment climate variables and tax are rather weak. Only Bénassy-Quéré et al. (2007) present a theoretical model but they do not completely analyze the behavior of the interaction term, i.e. the sensitivity of the elasticity of capital to taxation, to the investment climate.

In this paper we start with providing a theoretical framework that explains (i) under which circumstances capital reacts positively or negatively to a drop of the tax rate, and (ii) under which conditions the investment climate has a negative or positive impact on the relationship between capital and the tax rate. We start from the tax competition model of Zodrow and Mieszkowski (1986) (Z-M). Crucial to their model is the assumption that the public good (read investment climate) enhances the productivity of capital and that the public good is financed by the corporate tax. Following Sinn (2003) and Dhillon et al. (2007), we find that capital can react positively or negatively to a rise in the corporate tax rate depending on how effective the public good is at enhancing capital’s productivity. In addition, we find that this crucially depends on whether a higher corporate tax rate increases the tax revenue, i.e. on whether a country is on the rising or falling side of the Laffer curve. Whether the investment climate has a positive or negative impact on the relationship between capital and the tax rate depends on: (i) how complementary the public good is to capital, (ii) whether a higher corporate tax rate leads to higher or lower tax revenues and (iii) the rate at which the effectiveness of the public good at enhancing the productivity decreases as the level of the public good increases.

Whether the derivative of capital to the corporate tax rate is negative or positive and whether the impact of the investment climate on this derivative is negative or positive is an empirical question. We test the model imposing two important restrictions to our data selection. First, we select investment climate variables that correspond to two assumptions of the theoretical model: they are complementary to investment, and their outcome can improve if the government disposes of higher tax revenues. Second, we select a sample of countries ranging from countries with very low levels of the investment climate to countries with very high levels, allowing for variations in the complementarity of the investment climate variables along with their level. The unique dataset of marginal effective tax rates (METRs) from Chen and Mintz (2008) of 80 countries, including low and highly developed countries, allows us to do so. We use FDI inflows as a percentage of GDP as the dependent variable.

We find that in countries with a very poor basic investment climate, capital reacts positively on a rise of the tax rate. The reaction gets more negative as the investment climate improves. As a result, if the basic investment climate is poor, governments should focus on improving the investment climate and not on lowering the tax burden which is ineffective in such a situation.

Section 2 reviews the literature. Section 3 sets out the theoretical model. Section 4 describes the data and the methodology. Section 5 presents the results, which are further discussed in section 6. Section 7 concludes.

2. Literature review

The little empirical evidence on the impact of the investment climate on the effect of corporate taxes on investment consists of two kinds: (i) studies that split samples in better and worse investment climate (IC) countries and that estimate separate elasticities of capital to taxation for each sample, and (ii) studies that explicitly interact IC variables with the tax variable in investment equations[2]. Table 1 gives an overview of both kinds of studies. We learn that the evidence is mixed.

Demekas et al. (2007) study FDI flows to 16 South, Central and Eastern European countries and find that FDI is more sensitive to taxation in high GDP per capita countries than in low GDP per capita countries. At the same time, FDI is more sensitive to infrastructure in low GDP per capita countries. If one would assume that lower GDP per capita countries are worse IC countries, the findings of Demekas et al. (2007) would confirm the OECD and World Bank policy advice that for bad IC countries it is more effective to work first on the basic investment climate before lowering tax rates. However, a couple of other studies seem to contradict this, at first sight. In a study of FDI flows to 18 EU countries, Bénassy-Quéré et al. (2007) find that FDI is more sensitive in the sample of low public capital countries than in the sample with high public capital countries. However, given the selection of EU host countries, it is likely that the variance in public capital is rather low and that all sample countries provide the most vital IC needs for investors. Therefore, their results should be interpreted as results within a group of countries with a generally good investment climate. Turning to studies using interaction variables, we note that Mutti and Grubert (2004) for US firms and Azemar and Delios (2008) for Japanese firms, find that the location decision of multinationals to developed and developing countries is more sensitive to tax when GDP is lower, GDP per capita is lower, host countries are less developed, have less public goods or a worse quality of public governance. If one would assume that these variables are positively correlated with the IC, these studies suggest that the investors’ location decision is more likely to react to a drop in the tax rate in worse than in better IC countries. Bellak et al. (2007) find that the FDI flow to 8 CEEC countries is more sensitive to taxation if the infrastructure is poorer, but the coefficient of their interaction term is not significant. Gorg et al. (2009), on the other hand, study 18 OECD countries and find a higher sensitivity of FDI flows to taxation if social expenditures (as a share of GDP) are higher. A final empirical study that contributes to the discussion, be it from a different angle, is a study of Hines (2009) who finds that tax havens are typically countries with high-quality governance institutions. According to him tax havens are likely to be unsuccessful in the absence of high- quality governance.

These mixed results force us to be very careful interpreting them before drawing policy implications about the impact of the investment climate on the effectiveness of lowering the tax burden. The results depend on a number of important conceptual settings, but a general framework to evaluate them is missing. The results depend on the kind of IC variable that is used and on the level of variance of the IC level across the sample countries. For example, in studies with only developed countries, with a generally good IC, a worse IC does not mean a bad IC. Furthermore, some investment climate variables are more crucial or complementary to investment than others.

Table 1: literature overview.

Author Period-Countries Dep var Sample split/ Interaction Results
Demekas et al. (2007) 1995-2003

FDI from 24 (EU 15

+9) to 16 South, Central and Eastern European countries

FDI flow Sample split: GDPpc > $ 5887

GDPpc < $ 5887

– FDI more sensitive to STR in high GDPpc countries

– FDI more sensitive to infrastructure in low GDPpc countries

Bénassy-Quéré et al. (2007) 1994-2003

US FDI to 18 EU

countries

FDI flow Sample split:

high and low public capital countries

FDI more sensitive to tax in low public capital countries
Mutti and

Gubert (2004)

1982/1989/1994

US MNCs to 47 (for activity) or 60 a (for location) countries

Activity (GPO), and location decision Interaction: AETR*GDP AETR*GDPpc Activity and location more sensitive to tax when:

– GDP higher

– GDPpc hihger

Azemar and

Delios (2008)

1990-2000

Japanese firms to

66 developed and developing

countries

Location decision Interaction: STR*(L)DCb STR*GDPpc STR*public goods STR*quality of

public governance

Location decision more sensitive to STR in:

– less developed countries

– low GDPpc countries

– low public good countries

– low quality of public governance countries

Bellak et al. (2009) 1995-2004

FDI from 7 western

Bilateral FDI

flow

Interaction: BEATR*infra Higher sensitivity of FDI to tax if infrastructure low, but not
countries to 8 CEEC

countries

significant
Gorg et al. (2009) 1984-1998

18 OECD countries

FDI flow Interaction: EMTR*soc exp (% GDP) Higher sensitivity of FDI to tax if higher social expenditures

a: Only countries with >= 5 inv projects selected

b: dummy for less developed and developed countries

3. Theoretical analysis

The review of the empirical literature showed that the evidence on the sign of the derivative of investment to taxation is mixed and that the sign of the impact of investment climate on the sensitivity of investment to taxation is also ambiguous. In this section we present a theoretical framework that explains the impact of the investment climate on the derivative of capital to the corporate tax rate. Throughout this section we talk about public goods rather than the IC. An IC variable satisfies the definition of a public good in the sense of the model if it is non rival and non excludable, under direct control of the government, and if higher tax revenues can improve the investment climate outcome. We interpret this definition rather broadly by considering variables such as the security of property rights, the regulatory quality, the rule of law, etc. as public good variables.

3.1 The basic tax competition model

The analysis is based on the basic tax competition model of Zodrow and Mieszkowski (1986), henceforth Z-M. In the model citizens play the role of consumer and producer. Together, the citizens of a jurisdiction possess the amount of capital K- . The producer behaves competitively and takes government policies as given. He maximizes profits by choosing the appropriate level of capital K. Next to capital, also the public good G is a useful production factor. The production function for producing the private numéraire good is denoted by F (K, G) . The production function is characterized by:

– decreasing marginal productivity:

– complementarity of capital and the public good:

– decreasing complementarity of capital and the public good with rising G and K:

Capital is perfectly mobile between jurisdictions and it follows from international arbitrage that the single world price for capital is r. Next to the world price of capital, the firm also faces a capital tax rate t on capital. The firm maximizes profits when marginal costs equal marginal

revenue:

As a consumer the citizen draws utility from consumption of the private good. The output available for consumption of the private good by citizens is given by the consumers’ budget

constraint:

Citizens receive the profits from the firm, and the revenue from their endowment of capital. Government’s revenue is determined by the tax rate and the jurisdiction’s tax base capital, since t is a destination based tax. The government produces public goods according to the

budget constraint:

The government chooses the optimal tax rate to maximize consumption of the representative citizen under the constraint of equations (1), (2) and (3).

One can derive the change in capital demand within a jurisdiction due to a change in the tax rate, by differentiating equations (1) and (3) and find:

3.2 The interpretation of dK/dt

The denominator of (4) can be called the overall marginal productivity of capital (See Dhillon et al. (2007)). When capital goes up, this affects the productivity of capital FK in two ways. First, it decreases the productivity of capital because of the decreasing marginal productivity (FKK<0). Second, a rise in capital also causes a rise in the public good, by a factor t (according to equation (3)). Now, a rise in the public good increases the productivity of capital because they are complementary (FKG>0). As a result, there are two opposite forces on the productivity of capital. Z-M assume that the overall marginal productivity of capital is negative. This assumption rules out the possibility of ever increasing demand for capital since the overall productivity rises due to the provision of public goods[3].

Given the negative denominator, the overall sign of dK/dt depends on the sign of the numerator. Z-M assume that the numerator is always positive, irrespective of the level of public goods[4]. However, we follow the argument of Dhillon et al. (2007) and Sinn (2003) that this assumption should be relaxed to allow the numerator to be negative. The sign of the numerator depends then on the magnitude of FKG, i.e. the extent to which the public good enhances the marginal productivity of capital or the extent to which the public good is complementary to capital. Now, it is possible that the marginal value of the public good – through productivity enhancement – is higher than the marginal cost of the tax rate. In that case dK/dt is positive[5]. The basic intuition that we retain is that, the more public goods contribute to the productivity

of capital, the less capital is opposed[6 ] to a rise in the tax rate

However, to know the complete impact of the complementarity of the public good and capital, one may not solely rely on the analysis of the numerator. FKG also appears in the denominator. When taking the total derivative of dK/dt to FKG, we find that a higher complementarity of the public good and capital can also lead to a higher opposition of capital to a rise in the tax rate.

Formally,

This expression is only positive if (−KFKK − t) is positive. We prove in the appendix A1 that (−KFKK − t) is only positive if an increase in the tax rate – through a rise of the tax revenue leads to an increase in the public good (dG/dt>0). This is not the case if a rise in the tax rate does not make up for the fall of the tax base due to the rise of the tax rate (dG/dt<0)7 8. Thus,

This outcome is illustrated graphically in appendix B1.

If a country finds itself at the declining (right) end of the Laffer curve, a rise in the tax rate leads to lower revenues and – in our model – a lower level of the public good. Especially when analyzing countries with a high tax burden, this is an important consideration.

Proposition 1a: Capital is less opposed to a rise in the tax rate if the complementarity of the public good and capital is higher, conditional on the fact that the rise in the tax rate leads to more public goods.

Next to the complementarity of the public good and capital (FKG), also the rate at which the productivity of capital diminishes (FKK) is important for dK/dt. The derivative of dK/dt to FKK is:

This expression is positive if (1 − KFKG ) is positive, i.e. if dK/dt is negative. Thus,


Intuitively, a higher rate of decreasing productivity of capital, decreases the sensitivity of capital to the tax rate: if the productivity of capital rises less with the amount of capital, the demand of capital is less sensitive to changes in the user cost of capital, including the tax rate. As a result, the change in capital (dK) gets lower. For dK/dt, it means that a higher rate of decreasing productivity of capital brings dK/dt closer to zero: if dK/dt is positive it lowers dK/dt, if dK/dt is negative it increases dK/dt.

Propostion 1b: capital is less sensitive to a rise in the tax rate, if the rate at which the productivity of capital decreases is higher. As a result, capital is more (less) opposed to a rise in the tax rate, as the productivity of capital decreases at a higher rate, if dK/dt is positive (negative).

This outcome is illustrated graphically in appendix B2.

3.3 The interpretation of (dK/dt)/dG

We analyze the total impact of the level of the public good G on dK/dt, by analyzing the impact of the level of the public good (i) on the complementarity of the public good and capital and (ii) on the rate at which the productivity of capital decreases.

Concerning the complementarity of the public good and capital, we assumed in our tax competition model that the complementarity decreases when the level of the public good goes up (FKGG<0). The more public goods, the less an extra unit of the public good enhances the productivity of capital. Combining this with propostion 1a – higher complementarity of public good and capital lead to higher opposition of capital to a higher tax rate – one could prematurely conclude that capital always gets more opposed to a rise in the tax rate (dK/dt gets lower) if the level of public goods is higher, because an extra public good becomes less productivity enhancing. This is where the analysis of Dhillon et al. (2007) and Bénassy-Quéré et al. (2007) stops. Again, we find that this is only valid if a higher tax rate – though higher tax revenues – leads to more public goods, i.e. that a country is on the rising side of the Laffer curve.

But, an increase in the level of public good provision also has an impact on the rate at which the productivity of capital decreases. The tax competition model assumes that the productivity of capital decreases faster at higher levels of the public good (FKKG<0). Combining this with propostion 1b – faster decreasing returns lead to lower sensitivity of capital to the tax rate – we get that capital becomes less sensitive to a rise in the tax rate at a higher level of the public good (dK gets closer to zero). If dK/dt is positive, less sensitive means more opposed to a rise in the tax rate (dK/dt gets lower), while if dK/dt is negative, less sensitive means less opposed (dK/dt gets higher).

Therefore, if a country is on the rising side of the Laffer curve (dG/dt>0), three possibilities with respect to the impact of the level of public goods on the derivative of capital to the tax rate arise. If dK/dt is positive, both forces work in the same direction: a higher level of the public good always makes capital more opposed to a rise in the tax rate (dK/dt gets lower). Formally, the total derivate of dK/dt to the level of public goods can be written as follows:

Indeed, this is negative when (−KFKK − t) is positive (rising side of Laffer curve) and (1 − FKG K ) is negative (initial dK/dt is positive). Note that the initial dK/dt can only be positive if the public good is very effective at enhancing the productivity of capital. Graphically, this result is illustrated in appendix B3.

If dK/dt is negative, both forces work in opposite directions. Then, the overall effect depends on the relative magnitude of FKGG versus FKKG. Only if the rate at which more public goods reduce the complementarity of public goods is high relative to the rate at which more public goods enhance the decreasing productivity of capital, will capital become more opposed to a rise in the tax rate at higher levels of the public good. In appendix A2, we show that the probability of getting a negative (dK/dt)/dG indeed decreases with FKGG and increases with FKKG. Graphically, this result is illustrated in appendix B4.

Formally, table 1 illustrates these possibilities in the first three lines. The last three lines illustrate the three possibilities when countries are at the decreasing end of the Laffer curve, i.e when increasing the tax rate leads to lower levels of the public good (dG/dt<0).

Table 2: The sign of (dK/dt)/dG

1)dG/ dt > 0

2)dG/ dt < 0

a)dK / dt > 0

b)dK / dt < 0

a)dK / dt < 0

b)dK / dt > 0

i)FKGG (−KFKK − t) − FKKG(1− FKG K) < 0

ii)FKGG (−KFKK − t) − FKKG(1− FKG K) > 0

i)FKGG (−KFKK − t) − FKKG(1− FKG K) < 0

ii)FKGG (−KFKK − t) − FKKG(1− FKG K) > 0

(dK / dt) / dG < 0

(dK / dt) / dG < 0

(dK / dt) / dG > 0

(dK / dt) / dG > 0

(dK / dt) / dG < 0

(dK / dt) / dG > 0

In short, based on our theoretical model we can formulate the following propositions:

Proposition 2a: if higher tax rates lead to more (less) public goods, and if dK/dt is positive (negative), capital always becomes more (less) opposed to a rise in the tax rate as the level of public goods increases.

Proposition 2b: if higher tax rates lead to more (less) public goods, and if dK/dt is negative (positive), capital is more (less) likely to become more opposed to a rise in the tax rate as the level of public goods increases:

as the rate at which the level of public goods diminishes the complementarity of public goods and capital is higher, i.e. FKGG more negative.

as the rate at which the level of public goods enhances the decreasing productivity of capital, is lower, i.e. FKKG less negative.

4. Empirical analysis

The theoretical analysis demonstrates that the sign of (dK/dt)/dG depends on how complementary the public good is, how the complementarity changes with the level of the public good, and on the position of the country on the Laffer curve. Ultimately, this makes the sign of (dK/dt)/dG an empirical question. To test the model we need data on capital or investment, the tax rate, and the public good.

In order to test the sign of (dK/dt)/dG properly there are two important restrictions to the data. First, the public good variable needs to satisfy the definition of the public good in the model. This means that it must be productivity enhancing or complementary to capital (FKG>0), and that the outcome of the public good variable must improve if the tax revenue increases (G=tK). Second, we need a sample of countries ranging from very bad to very good public good countries. We need enough variance in the public good variable in order to establish its impact on the relation of capital to the tax rate. We need a range of countries that allows the public good to be more complementary in poor investment climate countries than in good investment climate countries (FKGG<0).


Within these data restrictions we were able to find a dataset for 80 countries for the year 2006- 2007.

4.1 Data

As the investment variable we use FDI inflows as a percent of GDP (FDI). The FDI inflow data come from the World Bank’s World Development Indicators. We use the most recent available data of 2007. We divide the FDI inflows by GDP to make them more comparable across countries. FDI data are the best available investment data since we want to include developing countries with a poor IC. We believe that FDI is a good investment measure in the context of this paper because it is very footloose. As a result it fits well into our theoretical model. We recognize that capital stock variables would have been preferable when available.

The capital tax rate comes from Chen and Mintz (2008). They calculate a unique set of marginal effective tax rates (METR) on capital for 80 developed and developing countries. We use the rates of 2006 to give investment some time to react to the METR. The dataset is exceptional because, more than any other tax dataset, it includes countries with a very poor investment climate. There has been much discussion in the literature about which tax rate should be used in empirical analysis (e.g. Devereux and Griffith (1998), Devereux (2006), and de Mooij and Ederveen (2008)). Although every tax measure has its strengths and weaknesses, incremental investment decisions should depend on the marginal effective tax rate. Moreover the METR is a forward looking measure which should be generally preferred to backward looking measures. The advantage of a forward looking measure is that it is not based on data on realized profits and tax payments, which could introduce an important endogeneity bias (Devereux (2006)). In our theoretical model the average and marginal tax rate are the same. However, METR is the right measure since the model considers marginal investment decisions (see equation (1)). The METR is a summary measure of the amount of tax paid as a percentage of the pre-tax return on investments that are marginal, i.e. just sufficient to cover financing and tax costs. Next to corporate income taxes, the METR that we use also includes sales and excise taxes on capital purchases and capital-related taxes (Chen and Mintz 2008). We will also use the statutory corporate income tax rate (STR) from Chen and Mintz (2008) to see whether the results are sensitive for the tax measure.

The investment climate variables, in order to fit the theoretical model, need to enhance the productivity of capital and be financed by tax revenues. To find out which investment climate variables are important for investors, we rely on the 2005 World Development Report (World Bank 2005), which reports the major investment climate constraints for investors based on investment climate surveys in 53 countries. From their list, we filter those variables of which we believe that their outcome can improve as the tax revenue increases. This reduced list consists of: regulations and tax administration, courts and legal systems, electricity, transportation and telecommunications. The ideal variable would summarize the variables of this list. Since this is not easy to find we define nine IC variables that cover at least part of the load. The detailed definitions and sources are presented in table 3. We start with the World Bank’s Doing Business (DB) indicators, which provide ten quantitative measures of regulations[9]. Our first IC variable, DB10, is a score summarizing these ten indicators[10]. Because not all 10 sub-indicators equally satisfy our definition of public good, we also construct the variable DB6, leaving out three sub-indicators that are related to market regulations (see definition table 3) of which we believe the outcome depends less on tax revenues, and the ‘paying taxes’ sub-indicator because it overlaps with our tax variable. Next to the DB indicators, we use the ‘regulatory quality’ (REQU) and the ‘rule of law’ (RULA) indicators from the Kaufmann, Kraay and Mastruzzi (2009) governance indicators. From the Economic Freedom of the World index we take the sub-component ‘legal structure and security of property rights’ (LEGPROP), which is assembled from the International Country Risk Guide, the Global Competitiveness report and the World Bank’s DB project. The Heritage Foundation offers us the ‘investment freedom’ indicator (INVFREE), which scrutinizes each country’s policies towards the free flow of investment capital. As an infrastructure variable, we use the ‘basic requirements’ indicator (BASREQ) from the Global Competitiveness Report. Unfortunately, this variable is broader than the definition of the public good that we intend, incorporating also the quality of private institutions and infrastructure and macroeconomic stability, something to keep in mind when interpreting the results[11].

For ease of interpretation, all IC variables are defined such that a higher value means a better IC. For ease of comparison, we standardize the IC variables. This will help us interpreting the results of the interaction variable across different regressions with different IC variables. Note that several IC variables might overlap, summarizing indicators from common sources. This is apparent from table 4 that shows the correlations between the 7 variables. Still, most indicators also seem to partly cover different loads. Also METR is included in the table. The investment climate variables are very little correlated with our tax measure.

With the investment, tax and investment climate variable at hand we can perform a first descriptive analysis of the interplay between them. Table 5 lists the countries and ranks them based on the DB10 score, with the worst IC countries on top. The list ranges from very poor to very good IC countries. In figure 1 we set out the data points for FDI and METR of these countries in a scatter plot. We divide the sample in two groups: the top half and the bottom half of the list[12]. This simple scatter already discloses an interesting pattern. FDI in the lower IC countries seems to be less responsive to difference in the METR than FDI in the higher IC countries. This suggests that a worse investment climate reduces the sensitivity of capital to changes in the tax rate.

To properly test the relationship between investment and the tax rate we will account for a number of control variables. Particularly in a cross country setting it is important to control for factors that might be correlated with our variables of interest (tax and the IC) and cause spurious relations if omitted. The definitions and sources of the control variables are detailed in table 6. For the selection of control variables we follow among others Djankov et al. (2008) and Azemar and Delios (2008). GDP per capita (GDPPC) controls for agglomeration effects since average income is higher in countries in agglomerated regions. It can also points to the productivity level and wages. Population (POP) points to the market potential within the borders of the country[13]. Inflation might influence investment through its impact on the cost of capital (Auerbach and Jorgenson 1980). We use the average inflation over the past five years (INFL). OPEN measures the extent to which the country’s economic activity is export oriented. As an alternative for OPEN we also use TRADE, an ‘Economic Freedom of the World’ index that measures how free countries are from tariff and non tariff barriers to trade. The geography variables capture how isolated a country is. We use the distance (DIST) of a country’s capital to the closest of three economic agglomerations (New York, Tokyo and Amsterdam) and we

control for the fact that a country is landlocked (LANDLOCK) and not European. Both variables come from Gallup et al. (1999). Next, we check additionally for the degree of liberalization of the labor market (LABMAR) and the credit market (CREDMAR). Finally, we control for political stability (POLSTAB) and the degree of democracy and freedom of speech (VOICE).

Table 7 shows the descriptive statistics of all the variables. Note that the normalized IC variables result in zero means and standard deviation equal to one. Luxemburg is left out of the sample because it is an outlier in terms of FDI as a percentage of GDP. As a result, we have at most 79 observations. FDI is from the year 2007. The other variables are from the year 2006, except the Doing Business variables DB10 and DB6, which are from 2007 for availability reasons.

4.2 Econometric specification

In order to establish the link between the tax rate and capital and the impact of the IC on this link we estimate the following specification to explain investment:

TAX denotes the capital tax rate, IC is the investment climate, X is a vector of control variables and u the error term. We are primarily interested in the total derivative of investment to the tax rate, β + δIC , and the impact of the IC on the derivative of investment to the tax rate, i.e. the sign and magnitude of interaction coefficient δ .The interpretation of β is made straightforward because we normalized the IC variables. Since the mean of IC is zero, β is the derivative of FDI to the tax rate when the IC of a country is equal to the sample’s average IC. We use one year time lags for all right hand side variables[14]. This is to reduce possible reverse causality and to give FDI a year to react to changes of the variables.

We start testing specification (6) without the interaction term as a benchmark. Then we add the interaction term. Next, we perform robustness checks by adding additional control variables that might be correlated with TAX or IC and FDI. We also use the STR instead of the METR as the tax variable to see whether the results still hold. Finally, we split the sample in below and above median IC countries, as an alternative to including the interaction term.

5. Results

5.1 Basic results

We start with the estimation of specification (6) without interaction term. Table 8 presents the results with the basic control variables. The dependent variable is FDI as a percentage of GDP, the tax rate is the METR, and the IC variable that differs from column to column is mentioned on top of each column. All variables are lagged by one year. We observe that the METR has a highly significant negative impact on FDI as a percentage of GDP, whatever IC variable is controlled for. The β coefficient ranges between -.155 and -.180. On average a 10 percent

point lower METR results in a 1.55 to 1.80 percent point rise of the FDI rate. The IC variables have the expected positive impact on the FDI rate but the coefficients are not significant, although the regulatory quality (column 3) and investment freedom (column 6) come close. Concerning the basic control variables, GDPPC, POP and INFL are not significant at all. Only OPEN is significant in all equations with the expected positive sign: more trade oriented economies attract a higher share of FDI on GDP.

Table 8 is an interesting benchmark to analyze the results with interaction terms in table 9. The only difference between the two tables is the inclusion of METR*IC, which also differs over the nine columns. Adding the interaction term visibly increases the R-squared in all columns. Focusing first on the β coefficient on METR, we observe no much change compared to table 8.

In some columns the coefficient is a bit higher, in others a bit lower, but the high significance remains. That the magnitude of β differs little between table 8 and 9 is not a surprise. The β coefficient must be interpreted as the total derivative of FDI to the METR when the IC is equal to zero, i.e. at the sample average IC. Turning to the coefficient δ on the IC variables, we observe a more important change. For an METR of zero, the IC has a significant positive impact on the FDI rate at least at the 5% level except for the IC variables ‘basic requirements’ where it is still significant at the 10% level. At the lowest end, a 1 standard deviation increase of the DB6 score increases the FDI rate with 3.50 percent points. At the highest end, a 1 standard deviation improvement of the regulatory quality increases FDI with 5.24 percent points.

Next, we are interested in the impact of the IC on the relationship between FDI and the METR. The interaction term’s coefficient is highly significantly negative for all IC variables. The negative sign corresponds to a negative (dK/dt)/dG in our theoretical model. Assuming that higher tax rates yield higher tax revenues, the negative sign must be due to the high complementarity of the IC with capital (high FKG) or due to a much higher complementarity of the IC at low levels of the IC than at high levels of the IC (a very negative FKGG).

With the coefficient on the interaction, we can now calculate the total derivative of FDI to METR. The first two columns of table 10 give the minimum and maximum value of the total derivative, β + δIC , for each IC variable, given the sample maximum and minimum level of the

IC variable and the point estimates of β and δ . We find that the FDI rate can react positively to a rise in the METR rate in the worst IC countries, and negatively in the better IC countries.

The third column of table 10 indicates how many sample countries have a positive total derivative for each IC variable, based on the point estimates. It ranges from four countries when using DB6 to 18 countries when the IC variable is INVFREE. Figure 2 shows how the total derivative (vertical axis) changes with the IC (horizontal axis) for four IC variables[15]. It shows that, at low levels of the IC, the derivative can be positive and turns negative when a certain level of the IC is reached. Finding positive derivatives for low levels of the IC does not mean yet that they are significantly positive. In column four to six of table 10 we replace the point estimate of β with its value at the upper end of the 95% confidence interval. The number of countries with a positive derivative declines importantly to between zero and four. Consequently, we cannot conclude that the derivative of capital to the corporate tax rate is significantly positive for the worst investment climate countries. However, we can say that the derivative becomes nonnegative. This is contrary to the message of table 8 – without the interaction term – of an overall negative impact of the corporate tax rate on FDI.

5.2 Robustness checks

The first set of robustness checks focuses on adding other control variables to exclude possible spurious correlations between METR or IC and FDI, due to omitted variables. We leave out the basic control variables that were not significant at all in tables 8 and 9. Only POP is retained. We also exclude OPEN because it is highly correlated with most of the additional control variables[16].

In table 11 we replace the variable OPEN with the variable TRADE[17], which includes more

regulatory measures of trade freedom. With TRADE the R-squared slightly improves. Compared to table 9, the coefficients on METR are slightly lower, but the significance is comparable. Also the coefficients on the IC variables drop a bit, sometimes reducing the significance from the 1% to the 5% level. The interaction coefficients are very similar. Overall, the results are robust to the inclusion of the variable TRADE. The variable TRADE itself enters highly significantly, even more so than OPEN.

In table 12 we check whether it is the geographical location of the sample countries rather than the tax rate and the investment climate that drives FDI. In most columns the air distance (AIRDIST) to one of three of the world’s biggest economic agglomerations enters just or almost significantly negative, giving support to the gravity model of international investment. Being landlocked (LANDLOCK) apparently plays no role for the FDI rate. Controlling for the geographical location has no qualitative impact on our three variables of interest.

Table 13 shows whether the liberal character of the credit and the labor market interferes with the relationship between tax and the investment climate and FDI. Neither variable enters the equation significantly, even though they have the expected positive sign. As a result they have no impact on the variables of interest.

Finally, in table 14 we add political variables. As expected political stability (POLSTAB) is important for FDI. It pops up positively and (almost) significantly in all cases. Note that this variable is highly correlated with any of the IC variables[18]. Still, the coefficients on IC and the interaction terms remain significantly positive. Finally, the voice and accountability index (VOICE) does not enter significantly.

The second robustness check verifies whether the results are also robust for the tax rate measure. Although we are convinced that METR is the best measure given our theoretical model, it is interesting to check whether the results also hold when using the statutory corporate tax rate (STR). Table 15 gives the answer. For easy comparison, we include the same basic control variables as in table 9. Surprisingly, compared to table 9 the R-squared is higher in all columns, suggesting that the STR explains the variance in the FDI rate better than the METR. It is possible that investment decisions are rather based on STR because it is more visible and easily comparable. The absolute magnitudes of the coefficients on the first three variables of interest are all higher. The derivative of the FDI rate to the STR is more negative at the average IC, ranging from -0.238 to -0.318. The IC has a more positive effect, although significance disappears in column 7 for the ‘basic requirements’ index. Generally, the coefficient on the interaction term remains significantly negative. We conclude that qualitatively the results are almost the same switching from STR to METR as the tax variable.

Finally, we repeat the analysis of table 8 without the interaction variables, but splitting the sample into countries with an IC below or equal to the median IC and countries above the median IC. Table 16 shows the results. For clarity only the coefficients for METR and the IC variable are presented[19]. The results confirm that there is an important difference in the relationship between FDI and METR for low and high IC countries. A higher METR has a significant negative impact on FDI for the high IC countries, for all IC measures. However, reducing the METR has no significant impact on FDI in the low IC countries. Turning to the IC variable, we find that the point estimates are always higher and have a bigger t-value in the low IC countries than in the high IC countries. They are significant using DB10 and REGQUA as the IC variable[20]. This indicates a higher importance of improving the investment climate to attract FDI in bad investment climate countries than in good investment climate countries.

6. Discussion

Let us start by interpreting two important findings in the results in the light of our theoretical model. First, we find a robust negative interaction coefficient: a worse investment climate means a lower opposition of capital to a rise of the tax rate. According to our theoretical model this can occur in three situations (line 1, 2 and 5 of table 2). Assuming that the bulk of countries in our sample is at the increasing left hand side of the Laffer curve, a negative sign only occurs when the IC is highly effective at enhancing the productivity of capital (proposition 2a) or when the effectiveness of the IC at enhancing the productivity is much higher at low level of the IC than at high levels (proposition 2b). Second, we find that the derivative of capital to the tax rate can become nonnegative or even positive in the worst IC countries. The theoretical model shows that this can only be the case when there is a high complementarity of the IC to capital (propostition 1a).

From a policy perspective, these are very important results. The nonnegative reaction of investment to a rise in the tax rate in the worst IC countries means that these countries should focus on improving the IC, rather than decreasing the tax rate to compensate for the poor investment climate. Decreasing the tax rate does not attract capital. This supports the view of policy advisers that “a low host country tax burden cannot compensate for a generally weak or unattractive FDI environment”. A general policy advice based on the results of table 8 without interaction term would say that any country can attract investment by lowering its tax rate. In contrast, based on the results of table 9 with interaction terms, the policy advice on corporate taxation should depend on the level of the investment climate in a country. In addition, this paper munitions the policy advice that a better investment climate is of significant importance to attract investment.

The empirical analysis shows the importance of the empirical design to find this result. To be in line with the theoretical model it is important to use a measure of the IC that corresponds to the definition of the public good. The measure must be comprehensive enough to incorporate those aspects of the investment climate that enhance the productivity of capital, but also concise enough to only include those aspects of the IC over which the government has direct control and for which higher tax revenue results in a better IC outcome. We also believe that for a proper empirical analysis of the impact of the IC on the sensitivity of capital to the tax rate, it is crucial to have a sample of observations that ranges from very low to very high IC countries. This is necessary to exploit the possibility that the IC is much more complementary at very low levels of the IC than at very high levels of the IC. In this respect, a cross section analysis is useful. For example, in a within country analysis the change in complementarity of the IC is likely to be too small.

The selection of the IC measure and of the sample might explain the mixed results on interaction terms in the literature as mentioned in the literature review. Studies that seem to contradict our results include less poor IC countries. For example, Mutti and Grubert (2004) select countries in which at least five investment projects have taken place, which could exclude the worst IC countries. Azemar and Delios (2008) include many developing countries but only 45 of their 66 sample countries are common to our sample. Ours includes less Latin American and Caribbean countries, while theirs includes less African and Central and Eastern European countries. The studies by Bénassy-Quéré et al. (2007) and Bellak et al. (2009) include only EU and CEEC countries respectively.

7. Conclusion

In this paper we provide a theoretical analysis to better understand the conditions when capital is more or less opposed to a rise in the tax rate. We find that the derivative of capital to the tax rate can be either positive or negative depending on how complementary the public good created with the tax revenue is to capital, and on whether a country finds itself on the increasing or decreasing side of the Laffer curve. We also find the conditions under which an improved investment climate has a positive or negative impact on the opposition of capital to a rise in the tax rate.

We then test the model empirically and find two major results. First, in the worst investment climate countries capital does not react to a rise in the tax rate or even reacts positively. Second, the reaction of capital to a rise in the tax rate becomes (more) negative as the investment climate improves.

These results have important policy implications. Very poor investment climate countries should not compensate for their poor investment climate by lowering the tax burden on capital. Instead, improving the investment climate is a more effective strategy. As a result a general policy advice to countries that decreasing the tax burden is effective in attracting investment is wrong. The advice should depend on the level of the investment climate of a country.


References

Auerbach, A., Jorgenson, D., 1980. Inflation-Proof Depreciation of Assets. Harvard Business Review. September/October, 113-118.

Azemar, C., Delios, A., 2008. Tax Competition and FDI: The Special Case of Developing Countries. Journal of Japanese International Economics. 22, 85-105.

Bellak, C., Leibrecht, M., Damijan, J.P., 2009. Infrastructure Endowment and Corporate Income Taxes as Determinants of Foreign Direct Investment in Central and Eastern European Countries. The World Economy. 32(2), 267-290.

Bénassy-Quéré, A., Gobalraja, N., Trannoy, A., 2007. Tax and Public Input Competition. Economic Policy. 19, 385–430.

Chen, D., Mintz, J., 2008. Taxing Business Investments: A New Ranking of Effective Tax Rates on Capital. FIAS World Bank, Washington, DC.

Demekas, D.G., Horváth, B., Ribakova, E., Wu Y., 2007. Foreign Direct Investment in European

Transition Economies – the Role of Policies. Journal of Comparative Economics. 35, 369 -386.

De Mooij, R.A., Ederveen, S., 2003. Taxation and Foreign Direct Investment: A Synthesis of

Empirical Research. International Tax and Public Finance. 10(6), 673-93.

De Mooij, R.A., Ederveen, S., 2008. Corporate Tax Elasticities: A Reader’s Guide to Empirical

Findings. Oxford Review of Economic Policy. 24(4), 680-697.

Devereux, M.P., Griffith, R., 1998. The Taxation of Discrete Investment Choices. IFS Working Paper 98/16.

Devereux, M.P., 2006. The Impact of Taxation on the Location of Capital, Firms and Profit: A Survey of Empirical Evidence. Working Paper from Oxford University Centre for Business Taxation, No. 702.

Dhillon, A., Wooders, M., Zissimos, B., 2006. Tax competition reconsidered. Journal of Public Economic Theory. 9(3), 391-423.

Djankov, S., Ganser, T., McLiesh, C., Ramalho, R., Schleifer, A., 2008. The Effect of Corporate

Taxes on Investment and Enterpreneurship. NBER Working Paper, No. 13756.

Feld, L.P., Heckmeyer, J.H., 2009. FDI and Taxation: A Meta-Study. CESifo Working Paper, No. 2540.

Gallup, J.D., Sachs, J.D., Mellinger, A., 1999. Geography and economic development. CID Working Paper, No. 1.

Gastanaga, V.M. , Nugent, J.B,. Pashamova, B., 1998. Host Country Reforms and FDI Inflows: How Much Difference Do They Make?. World Development. 26(7), 1299-1314.

Gorg, H., Molana, H., Montagna, C., 2009. Foreign Direct Investment, Tax Competition and Social Expenditure. International Review of Economics and Finance. 18, 31-37.

Grubert, H., Mutti , J., 2004. Empirical Asymmetries in Foreign Direct Investment and Taxation. Journal of International Economics. 62, 337 – 358.

Hines, J.R., 1999. Lessons from Behavioral Responses to International Taxation. National Tax Journal. 52, 305 – 322.

Hines, J.R., Dhammika, D., 2009. Which Countries become Tax Havens?. Journal of Public Economics. 93, 1058-1068

Jorgenson, D.W., 1963. Capital Theory and Investment Behavior. American Economic Review.53, 247-259.

Kaufmann, D., Kraay, A., Mastruzzi, M., 2009. Governance Matters VIII: Aggregate and Individual Governance Indicators 1996-2008. World Bank Policy Research Working Paper, No.4978.

Klemm, A., Van Parys, S., 2009. Empirical Evidence on the Effects of Tax Incentives. IMF Working Paper WP/09/136.

Morisset, J., 2003. Public Policy for the Private Sector: Tax Incentive. World Bank Note, No. 253. OECD, 2007. Tax Effects on Foreign Direct Investment, Recent Evidence and Policy Analysis. OECD Tax Policy Studies, No.17, Paris.

Sinn, H.-W., 2003. The New Systems Competition, Blackwell, Oxford.

Squalli, J., Wilson, K., 2006. A New Approach to Measuring Trade Openness. EPRU Working Paper 06-07.

World Bank, 2005. A Better Investment Climate for Everyone. World Development Report 2005, Washington, DC.

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 3: Definitions of IC variables.

Variable Definition Source
DB10 Doing Business score 10. This is a score calculated by the authors based on the scores of a country on the 10 Doing Business sub-indicators. DB indicators – World Bank- authors’ calculation
DB6 Doing Business score 6. This is a score calculated like DB10 but excluding

4 DB sub-indicators: employing workers, getting credit, paying taxes and trading across borders.

DB indicators – World Bank- authors’ calculation
REGQUA Regulatory quality. Captures the perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development. Kaufmann et al. (2009)
RULAW Rule of law. Captures the perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence. Kaufmann et al. (2009)
LEGPROP Legal structure and security of property rights. Captures judicial independence, presence of impartial courts, protection of property rights, military interference in rule of law and the political process, integrity of the legal system, legal enforcement of contracts, and regulatory restrictions on the sale of real property Economic Freedom of the World
INVFREE Investment freedom. Measures whether there is a foreign investment code that defines the country’s investment laws and procedures; whether the government encourages foreign investment through fair and equitable treatment of investors; whether there are restrictions on access to foreign exchange; whether foreign firms are treated the same as domestic firms under the law; whether the government imposes restrictions on payments, transfers, and capital transactions; and whether specific industries are closed to foreign investment. Heritage Foundation
BASREQ Basic requirements. Measures the quality of institutions, infrastructure, macroeconomic stability, and the quality of health and primary education Global Com- petitiveness Report

Table 4: Correlations of IC variables and METR.

METR DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
DB10 -0.08

1

DB6 -0.09 0.94

1

REGQUA -0.07 0.83 0.73

1

RULAW 0.03 0.81 0.72 0.94

1

LEGPROP 0.04 0.79 0.72 0.87 0.92

1

INVFREE -0.12 0.59 0.55 0.78 0.68 0.57

1

BASREQ 0.04 0.80 0.72 0.85 0.90 0.91 0.52

1


Table 5: List of 80 countries ranked by DB10.

DB10 rank Country DB10 rank Country DB10 rank Country

172 Sierra Leone 99 Kazakhstan 30 France

169 Chad 97 Croatia 29 Latvia

164 Bolivia 95 Zambia 27 Korea

160 Ukraine 93 Poland 26 South Africa

152 Brazil 88 Pakistan 25 Mauritius

151 Uzbekistan 83 Argentina 24 Switzerland

149 India 81 Serbia 23 Thailand

144 Uganda 80 Italy 21 Netherlands

137 Russian Federation 78 Jamaica 18 Finland

133 Nigeria 76 Romania 17 Germany

132 Iran 74 Mexico 16 Belgium

130 Indonesia 70 Ghana 15 Japan

129 Rwanda 66 Tunisia 14 Sweden

127 Morocco 65 China 13 Georgia

126 Ethiopia 63 Luxembourg 12 Iceland

125 Tanzania 62 Botswana 11 Malaysia

124 Bangladesh 57 Trinidad and Tobago 10 Norway

122 Ecuador 56 Slovakia 9 Australia

119 Egypt 53 Hungary 8 Denmark

116 Lesotho 52 Spain 7 Ireland

115 Costa rica 45 Fiji 6 United Kingdom

112 Kenya 43 Turkey 5 Canada

111 Madagascar 39 Portugal 4 United States

108 Greece 37 Peru 3 Hong Kong

105 Jordan 34 Bulgaria 2 Singapore

104 Vietnam 32 Chile 1 New Zealand

102 Czech republic 31 Austria


Table 6: Definitions of control variables.

Variable Definition Source
Basic:
GDPPC GDP per capita (thousands of current USD). WDI-World Bank
POP Population (millions). WDI-World Bank
INFL Inflation. Average percentage inflation in period 2002-2006. WDI-World Bank
OPEN Openness. A measure that combines trade intensity and the relative importance of a country’s trade level to total world tradea. This measure is proposed by Squalli and Wilson (2006). WDI-World Bank and authors’ calculations
Trade freedom:
TRADE Freedom to international trade. Index that measures taxes on international trade, regulatory trade barriers, size of trade sector relative to expected, black market exchange rates, and international capital market control. Economic Freedom of the World
Geography:
DIST Closest air distance to New York, Tokyo or Amsterdam (thousand of km). Gallup et al. (1999)
LANDLOCK Landlocked dummy. Dummy equal to one if country is landlocked and not in Europe. Gallup et al. (1999)
Market liberalization:
CREDMAR Credit market regulations. Index accounting for the ownership of banks, foreign bank competition, private sector credit and interest rate controls. Economic Freedom of

the World

LABMAR Labor market regulations. Index accounting for minimum wages, hiring and firing regulations, centralized collective bargaining, mandated cost of hiring, mandated cost of worker dismissal, and conscription. Economic Freedom of

the World

Political stability:
POLSTAB Political stability and absence of violence. Captures perceptions of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including politically motivated violence and terrorism. Kaufmann et al. (2009)
VOICE Voice and accountability. Captures perceptions of the extent to which a country’s citizens are able to participate in selecting their government, as well as freedom of expression, freedom of association, and free media. Kaufmann et al. (2009)

a: The measure is defined as: i

n

where X and M are exports and imports of country i, and n is the number of countries in the world.


Table 7: Descriptive statistics of all variables.

Variable Obs Mean Std. Dev. Min Max
FDI (% of GDP)

78

6.314 5.614 -0.232 26.895
METR

79

20.24 10.66 -6.00 46.00
STR

79

26.99 7.17 10.00 45.00
DB10

79

0

1

-2.202 2.311
DB6

79

0

1

-2.018 2.622
REGQUA

79

0

1

-2.201 1.595
RULAW

79

0

1

-1.630 1.727
LEGPROP

78

0

1

-2.413 1.698
INVFREE

78

0

1

-2.300 1.732
BASREQ

74

0

1

-2.158 1.752
GDPPC

79

14.77 17.55 0.20 72.33
POP

79

70.27 193.74 0.30 1311.02
INFL

79

0.044 0.027 -0.007 0.102
OPEN

79

2.032 3.782 0.003 20.196
TRADE

78

7.004 0.874 4.390 9.500
AIRDIST

77

3.719 2.813 0.140 9.590
LANDLOCK

77

0.130 0.338

0

1

CREDMAR

78

8.339 1.183 5.190 9.980
LABMAR

78

5.963 1.365 2.520 8.410
POLSTAB

79

0.073 0.901 -2.124 1.588
VOICE

79

0.282 0.950 -1.896 1.588

Table 8: Estimation results with basic control variables and no interaction variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
METR -0.172*** -0.175*** -0.155** -0.173*** -0.177*** -0.176*** -0.180***
(-2.84) (-2.93) (-2.54) (-2.88) (-2.93) (-2.84) (-2.88)
IC 0.767 0.788 1.695 1.129 0.207 1.156 -0.650
(0.89) (0.99) (1.56) (0.91) (0.21) (1.56) (-0.53)
GDPPC -0.037 -0.036 -0.073 -0.061 -0.023 -0.037 0.011
(-0.74) (-0.75) (-1.29) (-0.92) (-0.39) (-0.82) (0.18)
POP -0.003 -0.002 -0.003 -0.003 -0.003 -0.002 -0.003
(-0.87) (-0.72) (-0.88) (-0.93) (-0.99) (-0.66) (-0.95)
INFL -7.949 -10.222 4.996 -3.623 -7.879 -5.043 -7.754
(-0.28) (-0.37) (0.17) (-0.12) (-0.28) (-0.18) (-0.24)
OPEN 0.475** 0.474** 0.476*** 0.508*** 0.527*** 0.444** 0.579***
(2.54) (2.56) (2.68) (2.85) (2.92) (2.42) (2.97)
Cst 9.980*** 10.109*** 9.629*** 10.130*** 9.863*** 10.021*** 9.304***
(4.45) (4.49) (4.33) (4.49) (4.27) (4.48) (3.83)
Observations

78

78

78

78

77

77

73

R-squared 0.28 0.28 0.29 0.28 0.27 0.29 0.27
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 9: Estimation results with basic control variables and with interaction variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
METR -0.166*** -0.183*** -0.161*** -0.184*** -0.202*** -0.144** -0.185***
(-2.86) (-3.15) (-2.79) (-3.22) (-3.51) (-2.44) (-3.09)
IC 4.136*** 3.504** 5.235*** 4.768*** 4.396*** 4.816*** 3.550*
(2.74) (2.54) (3.43) (2.79) (2.67) (3.58) (1.79)
METR*IC -0.140*** -0.127** -0.166*** -0.156*** -0.173*** -0.180*** -0.155**
(-2.67) (-2.37) (-3.14) (-2.94) (-3.10) (-3.19) (-2.65)
GDPPC -0.024 -0.021 -0.050 -0.052 -0.014 -0.026 0.008
(-0.51) (-0.44) (-0.92) (-0.82) (-0.25) (-0.60) (0.14)
POP -0.003 -0.003 -0.003 -0.003 -0.002 -0.005 -0.002
(-0.91) (-1.04) (-1.08) (-1.02) (-0.72) (-1.39) (-0.74)
INFL 3.067 -6.641 16.743 10.995 10.757 -2.162 18.826
(0.11) (-0.25) (0.60) (0.39) (0.39) (-0.08) (0.57)
OPEN 0.343* 0.377** 0.381** 0.432** 0.456*** 0.317* 0.455**
(1.84) (2.05) (2.24) (2.53) (2.65) (1.79) (2.37)
Cst 9.387*** 10.063*** 9.028*** 9.826*** 9.619*** 9.287*** 8.665***
(4.34) (4.62) (4.29) (4.58) (4.41) (4.38) (3.71)
Observations

78

78

78

78

77

77

73

R-squared 0.34 0.33 0.38 0.36 0.36 0.38 0.34
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 10: Minimum and maximum total derivative of FDI to METR, and number of countries

with positive derivative.

IC variable (β+δIC)min (β+δIC)max

countries with (β+δIC)>0

(β+95+δIC)min (β+95+δIC)max

countries with

(β+95+δIC)>0

DB10 -0.490 0.142

7

-0.605 0.027

2

DB6 -0.516 0.073

4

-0.632 -0.042

0

REGQUA -0.426 0.204

14

-0.541 0.090

2

RULAW -0.453 0.070

6

-0.567 -0.043

0

LEGPROP -0.496 0.215

10

-0.610 0.101

4

INVFREE -0.456 0.270

18

-0.573 0.153

1

BASREQ -0.457 0.149

9

-0.576 0.030

1

β+95: β at the upper end of the 95% confidence interval

Table 11: Estimation results including TRADE as control variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
METR -0.151*** -0.159*** -0.158*** -0.163*** -0.175*** -0.127** -0.148**
(-2.73) (-2.86) (-2.87) (-2.90) (-3.05) (-2.20) (-2.53)
IC 3.401** 3.199** 3.664** 3.312** 3.353** 4.335*** 2.866*
(2.44) (2.42) (2.61) (2.45) (2.28) (3.20) (1.98)
METR*IC -0.137*** -0.129** -0.162*** -0.151*** -0.156*** -0.183*** -0.145***
(-2.73) (-2.50) (-3.12) (-2.94) (-2.88) (-3.30) (-2.71)
POP -0.002 -0.003 -0.003 -0.003 -0.001 -0.004 -0.002
(-0.77) (-0.93) (-0.93) (-0.83) (-0.49) (-1.30) (-0.57)
TRADE 1.861** 1.918** 1.965** 2.174*** 2.219*** 1.566* 2.704***
(2.32) (2.56) (2.36) (2.88) (2.87) (1.97) (2.99)
Cst -3.462 -3.652 -4.003 -5.208 -5.344 -1.853 -9.350
(-0.59) (-0.65) (-0.65) (-0.92) (-0.91) (-0.31) (-1.35)
Observations

77

77

77

77

77

76

72

R-squared 0.35 0.35 0.37 0.36 0.36 0.39 0.37
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 12: Estimation results including geographical variables as control variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
METR -0.171*** -0.183*** -0.183*** -0.198*** -0.213*** -0.139** -0.187***
(-3.05) (-3.23) (-3.27) (-3.40) (-3.63) (-2.42) (-3.11)
IC 5.131*** 4.829*** 5.233*** 4.765*** 5.032*** 5.736*** 4.521***
(4.11) (3.83) (4.16) (3.58) (3.54) (4.76) (3.31)
METR*IC -0.182*** -0.176*** -0.197*** -0.189*** -0.206*** -0.213*** -0.198***
(-3.58) (-3.28) (-3.68) (-3.44) (-3.62) (-3.89) (-3.51)
POP -0.001 -0.002 -0.002 -0.002 -0.000 -0.003 -0.001
(-0.44) (-0.66) (-0.54) (-0.48) (-0.13) (-0.99) (-0.34)
AIRDIST -0.406* -0.438* -0.303 -0.354 -0.405* -0.223 -0.409
(-1.86) (-1.98) (-1.33) (-1.53) (-1.72) (-1.05) (-1.64)
LANDLOCK 0.545 0.242 0.142 -0.298 -0.005 0.110 -0.725
(0.29) (0.13) (0.07) (-0.16) (-0.00) (0.06) (-0.34)
Cst 11.191*** 11.645*** 11.146*** 11.898*** 12.378*** 9.893*** 12.056***
(7.91) (8.06) (7.73) (8.10) (8.30) (6.71) (7.99)
Observations

76

76

76

76

75

75

72

R-squared 0.33 0.31 0.33 0.30 0.30 0.38 0.30
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 13: Estimation results including market variables as control variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
METR -0.155*** -0.167*** -0.166*** -0.178*** -0.196*** -0.128** -0.177***
(-2.72) (-2.92) (-2.98) (-3.10) (-3.41) (-2.20) (-2.98)
IC 4.682*** 4.459*** 5.150*** 4.767*** 5.125*** 5.583*** 4.636***
(3.76) (3.63) (4.20) (3.78) (3.82) (4.68) (3.48)
METR*IC -0.189*** -0.183*** -0.213*** -0.204*** -0.219*** -0.219*** -0.206***
(-3.67) (-3.36) (-4.01) (-3.82) (-4.01) (-4.00) (-3.68)
POP -0.001 -0.002 -0.001 -0.000 0.001 -0.003 0.001
(-0.25) (-0.52) (-0.35) (-0.15) (0.33) (-0.86) (0.23)
CREDMAR 0.857 0.867 0.764 0.899 0.953 0.566 1.107
(1.29) (1.36) (1.24) (1.49) (1.53) (1.11) (1.49)
LABMAR 0.422 0.437 0.524 0.541 0.544 0.526 0.436
(0.94) (0.97) (1.20) (1.23) (1.25) (1.23) (0.90)
Cst -0.158 -0.026 0.278 -0.531 -0.694 1.163 -1.609
(-0.03) (-0.00) (0.05) (-0.10) (-0.13) (0.25) (-0.25)
Observations

77

77

77

77

77

76

72

R-squared 0.33 0.32 0.36 0.33 0.34 0.39 0.33
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 14: Estimation results including political variables as control variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
METR -0.181*** -0.194*** -0.187*** -0.205*** -0.218*** -0.151** -0.204***
(-3.27) (-3.43) (-3.40) (-3.62) (-3.82) (-2.63) (-3.51)
IC 4.373*** 3.834*** 5.540*** 3.766** 3.698** 5.258*** 3.526**
(3.21) (2.81) (3.41) (2.25) (2.19) (4.04) (2.27)
METR*IC -0.156*** -0.141** -0.170*** -0.161*** -0.173*** -0.179*** -0.162***
(-3.11) (-2.64) (-3.18) (-2.94) (-3.10) (-3.12) (-2.93)
POP -0.001 -0.001 -0.002 -0.000 0.001 -0.003 0.000
(-0.22) (-0.33) (-0.48) (-0.11) (0.28) (-0.78) (0.14)
POLSTAB 1.595 1.578 0.870 1.736 2.160* 1.374 2.158*
(1.55) (1.51) (0.78) (1.43) (1.76) (1.48) (1.87)
VOICE -0.878 -0.723 -1.447 -0.533 -0.611 -1.276 -0.818
(-1.01) (-0.82) (-1.36) (-0.54) (-0.67) (-1.31) (-0.86)
Cst 10.134*** 10.360*** 10.536*** 10.678*** 10.897*** 9.715*** 10.799***
(8.32) (8.37) (8.70) (8.35) (8.55) (7.65) (8.24)
Observations

78

78

78

78

77

77

73

R-squared 0.33 0.31 0.35 0.31 0.32 0.38 0.32
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 15: Estimation results with STR as tax variable.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Dependent variable: FDI FDI FDI FDI FDI FDI FDI
IC variable: DB10 DB6 REGQUA RULAW LEGPROP INVFREE BASREQ
STR -0.247*** -0.255*** -0.249*** -0.276*** -0.312*** -0.238*** -0.318***
(-3.15) (-3.24) (-3.18) (-3.52) (-3.89) (-2.86) (-3.73)
IC 7.588*** 6.919*** 7.162*** 6.863*** 5.739** 6.984*** 3.669
(3.30) (2.99) (3.47) (2.85) (2.21) (2.84) (1.13)
STR*IC -0.231*** -0.217*** -0.218*** -0.208*** -0.191** -0.208** -0.154*
(-3.30) (-2.91) (-3.11) (-2.86) (-2.55) (-2.41) (-1.73)
GDPPC -0.043 -0.045 -0.057 -0.068 -0.040 -0.048 0.007
(-0.95) (-1.00) (-1.09) (-1.10) (-0.71) (-1.14) (0.11)
POP -0.005* -0.005* -0.005* -0.005* -0.005* -0.004 -0.005*
(-1.67) (-1.67) (-1.69) (-1.76) (-1.89) (-1.44) (-1.82)
INFL 2.285 -9.215 8.814 2.352 -1.773 -1.180 -2.475
(0.09) (-0.36) (0.32) (0.09) (-0.07) (-0.04) (-0.08)
OPEN 0.395** 0.417** 0.492*** 0.528*** 0.529*** 0.414** 0.570***
(2.25) (2.38) (3.00) (3.19) (3.16) (2.40) (3.02)
Cst 12.884*** 13.607*** 12.739*** 13.956*** 14.732*** 12.990*** 14.158***
(4.50) (4.74) (4.46) (4.87) (5.11) (4.32) (4.58)
Observations

78

78

78

78

77

77

73

R-squared 0.40 0.38 0.40 0.38 0.37 0.38 0.35
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Table 16: estimation results with sample split according to median IC value.

IC variable:

DB10

DB6

REGQUA

RULAW

IC<=median IC>median IC<=median IC>median IC<=median IC>median IC<=median IC>median
METR -0.001 -0.355*** -0.050 -0.217* 0.020 -0.302*** -0.039 -0.320***
(-0.01) (-3.20) (-0.98) (-1.86) (0.31) (-2.77) (-0.50) (-3.52)
IC 1.963* -0.946 0.398 -1.169 3.086** 4.280 3.486 3.277
(1.76) (-0.46) (0.39) (-0.63) (2.28) (1.39) (1.52) (1.15)
Observations

39

39

38

40

40

38

39

39

R-squared 0.20 0.39 0.17 0.39 0.19 0.48 0.14 0.48
IC variable: LEGPROP

INVFREE

BASREQ

IC<=median IC>median IC<=median IC>median IC<=median IC>median
METR -0.056 -0.249** -0.010 -0.439*** -0.038 -0.326***
(-0.72) (-2.33) (-0.16) (-3.48) (-0.45) (-3.33)
IC 1.491 0.058 1.451 -2.470 0.763 0.489
(0.99) (0.02) (1.23) (-0.75) (0.29) (0.19)
Observations

38

39

45

32

36

37

R-squared 0.12 0.44 0.09 0.53 0.09 0.47
t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Appendix


B1: The impact of FKG on dK/dt (FKK constant)

Rise in tax rate (dt):

=> (t+r) shifts up

=> FK shifts up when dG/dt>0 (left panel), FK shifts down when dG/dt<0 (right panel) Left graph: dG/dt>0; green case: low FKG, red case: high FKG

Right graph: dG/dt<0; green case: low FKG, red case: high FKG

=> higher FKG, higher (lower) dK/dt if dG/dt>0 (dG/dt<0).

B2: The impact of FKK on dK/dt (FKG constant)

Rise in tax rate (dt):

=> (t+r) shifts up

=> FK shifts up (assuming dG/dt>0)

=> FK steeper (assuming dG/dt>0), depending on FKK Left graph: dK/dt <0; green case: high FKK, red case: low FKK Right graph: dK/dt >0; green case: high FKK, red case: low FKK


B3: (dK/dt)/dG <0 (if dK/dt>0, and if higher tax means more G)

Rise in tax rate (dt):

=> (t+r) shifts up

=> FK shifts up

=> FK steeper

Light blue case: low G (high FKG, high FKK); dark blue case: high G (low FKG, low FKK)

=> higher G, lower dK/dt: (dK/dt)/dG<0

B4: (dK/dt)/dG not determined (if dK/dt<0, and if higher tax means more G)

Rise in tax rate (dt):

=> (t+r) shifts up

=> FK shifts up

=> FK steeper

Left graph: low FKGG; light blue case: low G, dark blue case: high G

Right graph: high FKGG (almost zero); light blue case: low G, dark blue case: high G

=> lower FKGG (compared to FKKG), more likely negative (dK/dt)/dG

* Corresponding author: Address: Tweekerkenstraat 2, 9000 GENT, Belgium Tel: +32 92 64 34 79; cell phone: +32 485 92 73 37; fax: +32 92 64 35 99 Email: stefan.vanparys@ugent.be

[1] For excellent overviews see Devereux (2006), Feld and Heckmayer (2009), Hines (1999), and OECD (2007).

[2] A third option could be to look at separate studies focusing on good IC countries and bad IC countries. However, comparing results of studies that use good IC countries versus studies that use bad IC countries would be dangerous because of differences in data definitions, estimation methodology, etc. A few studies of which we know that only focus on poor investment climate countries, assuming for a second that developing countries are poor IC countries, are Gastanaga et al. (1998) and Klemm and Van Parys (2009).

[3] It is equivalent to saying that the marginal product of capital exhibits decreasing returns to scale as a function of capital and the public good (Benassy-Quere et al. 2007).

[4] As a result, dK/dt is always negative: raising the tax rate leads to capital flight. From this result follows a race to the bottom with countries undercutting each other’s tax rate, resulting in underprovision of the public good.

[5] This would then lead to a race to the top with overprovision of public goods.

[6] We prefer to use the term ‘opposed’ rather than ‘sensitive’ because we allow dK/dT to be positive. If dK/dT is positive, ‘more opposed’ still means a lower dK/dT, while ‘more sensitive’ would mean a higher dK/dT. A third option could be to look at separate studies focusing on good IC countries and bad IC countries. However, comparing results of studies that use good IC countries versus studies that use bad IC countries would be dangerous because of differences in data definitions, estimation methodology, etc. A few studies of which we know that only focus on poor investment climate countries, assuming for a second that developing countries are poor IC countries, are Gastanaga et al. (1998) and Klemm and Van Parys (2009).

[9] starting a business, dealing with construction permits, employing workers, registering property, getting credit, protecting investors, paying taxes, trading across borders, enforcing contracts and closing a business.

[10] We prefer to calculate a score to the DB ranking because an ordinal variable is less accurate to reflect the variance across countries. The calculation of the score is based on the scores of all sub-indicator that together constitute the World Bank DB rank. In the calculation of the score of every of the ten DB sub-indicators gets an equal weight and within every sub-indicator every sub-sub-indicator gets an equal weight. This methodology corresponds to the methodology of calculating the Doing Business ranking (see www.doingbusiness.org ) but using the indicator scores instead of the ranks. The correlation between our DB score and the DB rank is -.91. The correlation between the rank based on our score and the World Bank DB score is .96.

[11] We would have liked to also cover the critical investment variable ‘transportation’. But no data are available. The WDI provides data on the share of roads paved, but only 47 of our sample countries are covered.

[12] We exclude Luxemburg because it is an outlier in terms of FDI as a percentage of GDP

[13] Since FDI is already divided by GDP, we do not control for the level of GDP anymore. We tried out all regressions below including GDP, but it was never significant.

[14] Except for DB10 and DB6 for which we only have 2007.

[15] We left out the other three for clearness. We included the top 2 and the bottom 2 in terms of countries with a positive total derivative of FDI to METR.

[16] We also did the regression including OPEN but this did not qualitatively alter the results on the variables of interest.

[17]The correlation between them is 0.61

[18] Between 0.60 and 0.84

[19] If the number of observations below or equal to the IC is different from the number of observation above the IC, several countries have the median IC value.

[20] Note that the low number of observations reduces the efficiency of the estimation.

Previously published by the Ghent UniversityDepartment of Economics


To top