Leaning Right and Learning From the Left

Abstract

There is an increased focus in comparative politics and international relations on how choices of governments are dependent on choices made by other governments. The authors argue that although the relationship between policy choices across countries is often labeled as either diffusion or competition, in many cases the theoretical mechanisms underpinning these labels are unclear. In this article, the authors build a model of social learning with a specific application to the diffusion of corporate tax reductions. The model yields predictions that are differentiable from existing models of tax competition. Specifically, the authors argue that social learning is most likely in the wake of tax policy cuts by left governments. They test the model using an existing data set of corporate tax rate changes and an author-created data set of changes in tax legislation, covering 20 Organisation for Economic Co-operation and Development countries. The authors’ empirical findings show that social learning is an important determinant of corporate tax policy making.

Keywords

tax competition corporate taxation race to the bottom fiscal policy foreign direct investment (FDI)diffusion social learning

Corporate tax policy is at the intersection of economics and politics, where tax policy changes can be used to raise revenue, stimulate production and investment, and redistribute wealth and income. These decisions also have an international dimension, where other countries’ tax policies can influence a country’s tax policy decisions. For example, the admission of 10 Central and Eastern European countries with relatively low corporate tax rates into the European Union (EU) led to discussions in Western Europe on how to compete with these new entrants for foreign investment.¹

In this article we directly address the international dimensions of corporate tax policy, exploring how tax policy decisions diffuse across borders. We build a model of social learning and contrast it with existing theories of corporate tax competition. The model yields comparative statics that allow us to empirically differentiate our social learning explanation for clustered tax policy cuts from existing models of tax policy competition. Specifically, although models of tax policy competition focus on the role of large countries or competitors for investment as triggers for tax policy diffusion, we focus on the potential for global learning, when governments with long-held preferences for higher corporate taxes (left governments) legislate corporate tax reductions.

We stress that it is possible that multiple channels of diffusion affect corporate tax decisions and that social learning is not mutually exclusive from these other forms of diffusion, including competition. Our goal here, therefore, is not to eliminate other forms of diffusion as general explanations for tax policy convergence but rather to develop a model and empirical test that allows researchers to differentiate between instances of diffusion by social learning and diffusion by pure competition.

We test our theory using an existing data set of tax policy rates from Basinger and Hallerberg (2004) and a unique, author-created data set on the timing of tax policy cuts in 20 Organisation for Economic Co-operation and Development (OECD) countries from 1980 to 1998. This time period provides for direct comparison of our empirical results with past studies of tax competition and limits the focus to competition within the OECD, prior to the enlargement of the EU and the emergence of China as a major investment location. Our empirical results are in direct contrast to findings of across-the-board competition in response to global corporate tax reductions and instead show that clustered tax cuts are conditional on the partisanship of the first-mover government. Specifically, we find, consistent with our theory, that although tax cuts by right governments typically do not trigger global tax reductions, the same policy choice by left governments tends to lead to clustered in-kind policy responses by other governments. We attribute this apparent policy leadership by the left to the fact that left governments that opt for a policy that goes against their general preferences communicates important information about the viability of tax cuts. At the same time, tax cuts by right governments, whose policy programs tend to include tax cuts irrespective of economic climate and investor behavior, provide very little new information about the viability of tax cuts. Although competitive dynamics can influence tax policy, the temporally clustered corporate tax cuts in the OECD have originated, at least to some extent, in social learning, not tax competition.

Our theory does not claim that left governments do not cut taxes; they do, and the Scandinavian governments in the 1980s are a case in point.² Rather, we propose that when left governments choose to implement tax cuts, their action provides information to other governments, whereas the same is not true (at least not to the same extent) for right governments that cut taxes. For that reason, we argue that the diffusion of tax cuts after Reagan’s tax reductions in the early 1980s was not solely based on tax competition with the United States. The decision of left-leaning Scandinavian governments to cut taxes, at that time, triggered waves of tax cuts through social learning. In our empirical analysis we describe how to differentiate between these two theories. Our article proceeds as follows. In the next section, we provide an overview of the growing literature on diffusion in international and comparative political economy research. We construct a social learning model with a specific application to corporate tax policy in the second section, highlighting the observable implications and how they differ from existing models of tax competition. In the third section, we provide an overview of our data and method for testing our social learning model on tax policy reforms in advanced industrialized countries. We present our results in the fourth section and conclude with the fifth section.

Diffusion, Learning, and Competition in the Global Economy

In this article, we address the diffusion of tax policy across national borders. Although this topic is of substantive importance, it also fits into a broader literature in political science on the interdependence and diffusion of policy across a wide set of policy areas.³ For example, in the study of monetary policy scholars have pointed to periods of “beggar thy neighbor” policies where countries use currency devaluations to gain advantages over their trading partners.⁴ Other scholars have studied the spread of neoliberal economic reforms across countries, including privatization and pension reform (see Bonoli, 2007; Brooks, 2002; Drezner, 2005; Hansson & Olofsdotter, 2008; Henisz, Zelner, & Guillen, 2005; Meseguer, 2004; Steinmo, 2002). Diffusion of policy can be driven by a number of theoretical mechanisms. Differentiating among these mechanisms has important academic and substantive implications (Simmons, Dobbin, & Garrett, 2006).⁵ In this article, we clearly differentiate between one often invoked mechanism of diffusion, competition, and an alternative model of diffusion, social learning. We directly explore the diffusion of corporate tax policy by means of social learning, thereby building on the vast public finance literature examining the spatial determinants of tax rates (see Davies & Voget, 2009; Ganghof & Eccleston, 2004; Garrett, 1998; Swank, 1998; Wildasin, 1989; Wilson, 1986, 1999; Wilson & Wildasin, 2004).

Specifically, a substantial number of scholars have argued that the reduction in corporate tax rates is driven by a “race to the bottom,” a form of diffusion largely driven by competitive dynamics (e.g., Andrews, 1994; Cerny, 1990; Kurzer, 1993).⁶ Recent scholarship in political science, however, has amended the purely competitive model and has focused on how domestic political institutions mitigate competitive dynamics (e.g., Franzese & Hays, 2007, 2008; Ganghof & Eccleston, 2004; Garrett, 1998; Swank, 1998; Swank & Steimo, 2002).⁷ For example, this literature has shown that the existence of veto players can limit the ability of governments to respond to tax competition (Hallerberg & Basinger, 1998) and that electoral institutions can influence how governments respond to tax cuts in other countries (Hays, 2003). Basinger and Hallerberg (2004) propose a formal model of tax policy competition that accounts for domestic political constraints. They argue that tax rates are greatly influenced by domestic politics, including levels of partisanship and veto players, which can limit tax policy cuts. Empirically, they find strong evidence for corporate tax competition, yet domestic political factors still heavily influence policy choice.

We develop a theoretical model of policy learning. One assumption that we make in our model is that governments are operating in an environment of imperfect information on the specific impact of tax changes on investment (the elasticity of capital to taxes). Scholarship, such as Kiymaz and Taylor (2000), makes this assumption at the individual project level, where “hosts cannot be sure of how much to offer the firm to locate in its borders, largely because it cannot precisely estimate the political, social, and cultural costs that the firm would face in foreign production” (p. 55). Thus, although governments can make reductions in taxation to increase investment, existing scholarship has debated the magnitude and overall effectiveness of corporate tax reductions in increasing investment.⁸

One benefit of our simple theoretical framework is that it allows for other types of learning. For example, uncertainty could lead governments to “experiment” with tax reductions, cutting corporate taxes and then examining how FDI responds to tax cuts. Alternatively, governments could learn from the tax experiences of others by observing how FDI responds to tax reductions in other countries (see, e.g., Baturo & Gray, 2009; Volden, Ting, & Carpenter, 2008). These are plausible alternatives, or complementary models of learning, yet we stress that our empirical identification strategy is to differentiate our model of social learning from (pure) competition. We specifically address how our hypothesized patterns of statutory corporate tax reductions are consistent with our social learning model and inconsistent with arguments about common shocks or tax competition. We present our model in the following section.

Biased Learning and Globalization

We begin by discussing the two components of our theoretical approach and then explain how these two approaches can be combined to further our understanding of national (tax) policy making. The first approach goes under various names, but social learning is probably the most widely used and most descriptive among them. Common to all social learning approaches is the idea of “pure informational externalities” (Gale, 1996). This means that individuals can learn from each other’s behavior, but their payoffs are not affected by the actions of the other individuals.

Here, we focus on a particular class of social learning models that has been referred to as cascade or herding models (hereafter, simply referred to as cascade models). In the typical cascade model, individuals receive a private signal, for instance, on the viability of corporate tax cuts, and then are presented with a binary choice, that is, to cut the corporate tax rate or not (Banerjee, 1992; Bikhchandani, Hirshleifer, & Welch, 1992, 1998). In the most basic version of the model, individuals choose sequentially, with the order determined exogenously. The model generates two key findings. First, individuals take into consideration the actions of earlier individuals. Second, the use of information tends to be inefficient.

We apply the general idea of cascade models to the question of how national governments make tax policy decisions in a globalized world.⁹ We believe that the basic dynamic of cascade models applies to national governments’ tax decisions, but with additional twists, some of which have been discussed in the existing literature, and one of which, to the best of our knowledge, has not yet been considered in the context of cascade models.

The first important part of the model is ideology. Specifically, we assume that right governments are more likely to enact corporate tax cuts.¹⁰ The fact that right and left governments have different preferences on taxation is often the very definition of right and left governments (Laver & Hunt, 1992) and a core assumption of Basinger and Hallerberg’s (2004) tournament model.¹¹ Right governments have lower costs of cutting corporate taxes and are more likely to enact corporate tax cuts, ceteris paribus, than left governments. Building on this simple assumption and an imperfect information environment, we show how partisanship can facilitate learning.

The second important part concerns the value of biased advice (Calvert, 1985).¹² We show that a national government can—knowing about the other government’s ideology and the general process of political advice (i.e., the value of biased advice)—infer information about another government’s domestic policy-making process and use that information to optimize its own policy. Take the example of a left government that adopts a conservative policy. Other governments, both left and right, that observe the first government can infer from this that the left government must have obtained new and credible information, since absent new information and given the preferences of its core constituencies, the left government would have had no incentive to adopt the conservative policy.¹³

In sum, we argue that Calvert’s biased adviser model has implications for international politics, in that governments that reverse long-held policy positions facilitate social learning and diffusion across countries. In the cascade literature, such governments would be referred to as fashion leaders. What is interesting about this model is that by adding an international level to Calvert’s biased advisor story, Calvert’s result is expanded.

Combining the cascade model with the biased advisor model and applying the modified theory to our question about national governments’ tax policy positions yields the following prediction. Tax cuts by right governments provide very little information to other countries, whereas a tax cut by a left government allows for social learning to take place.

The assumption that left governments are less likely to enact tax cuts also illustrates how common exogenous shocks affect tax policy. For example, in Basinger and Hallerberg’s (2004) model an increase in the number of right governments in the world (an exogenous shock) increases the “effort” exerted by governments to cut taxes. Yet their model still predicts that right governments are more likely to reduce corporate taxes than left governments.¹⁴ This is an important point for both our model and our empirical analysis. Common shocks, such as a change in the mobility of capital, can lead many or even all countries to cut corporate taxes. However, we argue that a specific pattern of partisan learning can be differentiated from both standard models of tax competition and responses to common shocks.

The somewhat counterintuitive conclusion here is that left governments lead the way on tax cuts. Moreover, our model yields the following important empirical prediction: In line with the standard logic of cascade or herding models, the larger the number of left governments cutting taxes, the stronger the herding effects, in that the dynamic is more difficult to reverse (even if herding is based on incorrect information). The underlying logic of this key empirical prediction will be further developed in the following paragraphs where we outline the mathematical model.¹⁵

In the model, there are two types of governments, right and left governments, that have to decide whether to cut taxes or not, considering that elasticity of capital to taxes is either higher or unchanged.¹⁶ When θ, the state of nature indicating capital elasticity, is equal to 1, then there has been an increase in capital elasticity. When θ = 0, then capital elasticity is unchanged. Both types of governments, left and right, have the opportunity, when called on, to take action, by either cutting their corporate tax rate, x = 1, or maintaining the current corporate tax rate, x = 0.¹⁷ Therefore, government t gets to decide on a course of action in period t.

When governments are called on to act, they receive a privately observed signal that provides probabilistic information on the elasticity of capital. The core of this model, as with any social learning model, concerns the evolution of beliefs about the state of nature. When it is an agent’s turn to act, she or he combines, via Bayesian updating, the public knowledge about the probability of increased capital elasticity, i.e., θ = 1, with her or his private knowledge, the source of which is the agent’s private signal. Specifically, in each period, the public belief is updated by the following formula.¹⁸

λ_{t + 1} = λ_{t} + ν_{t}, with ν_{t} = log (\frac{P (x_{t} | θ = 1)}{P (x_{t} | θ = 0)})

(1)

In words, this formula states that the public belief about capital elasticity at time t + 1 is derived by combining the public belief at time t and the information conveyed by the action of government t in time period t. The updating of the public belief is one crucial place where we diverge from the standard model in the literature. Although the literature that we are familiar with would distinguish governments only by their actions, that is, tax cut or no tax cut, we introduce the idea that it matters what type of government—left or right—it was that took the action.¹⁹

As evidenced by Equation 1, we model the updating of the public belief by observing the actions (i.e., tax policy) of other actors (i.e., governments). The ratio ν _t is to be interpreted as the information conveyed by the actions, specifically in terms of the probability of a particular state of nature given the observed action. We show at the end of this section that, under some general conditions, tax cuts by left governments lead to more favorable public beliefs about increases in capital elasticity than is the case for tax cuts by right governments. Tax cuts by left governments, who are normally loath to cut taxes, provide new information.

We generate a set of simple simulations that show the trajectory of public beliefs under different scenarios. The simulations are based on a model with 10 governments, 5 of which are left governments and 5 of which are right governments (in the following paragraphs right governments are denoted by R and left governments by L). In a first step, we randomly generate the following sequence of those 10 governments.

(L 1, R 1, L 2, L 3, L 4, R 2, R 3, R 4, L 5, R 5)

We then set the log likelihood ratio of the initial public belief, λ₀, equal to zero, which corresponds to a starting public belief of µ₀ = 0.5.²⁰ We also assume the following updating probabilities: p_L = 0.4, p′ _L = 0.8, p_C = 0.49, and p′ _C = 0.51. Here 1–p_L refers to the probability that a tax cut by a left government accurately reflects increased capital elasticity, 1–p′ _L that a left tax cut does not accurately reflect changes in capital elasticity, 1–p_C that a right government’s tax cut happens in an environment of increasing capital elasticity, and 1–p′ _C that a right tax cut occurs despite unchanged capital elasticity. We then reverse the partisanship of the sequence, replacing right governments with left governments and vice versa. We graph the evolution of public beliefs in Figure 1, with lighter lines in the five panels representing the evolution of public beliefs for the first sequence and darker lines the evolution of public beliefs for the second sequence.

Figure 1.

Evolution of public beliefs

We then estimate how tax cuts early in the sequence affect public beliefs. In the first panel of Figure 1 (lighter line), it is assumed that only R1 cuts taxes (c on the x-axis), with all of the other governments, both left and right, opting not to cut taxes (n on the x-axis). In the second panel (again, the lighter line and going clockwise), R1 and R2 are the only governments to cut taxes. This pattern continues until, in the fifth panel, all right governments are assumed to cut taxes, but none of the left governments.

Comparing the dark and lighter lines, we can see that increasing the number of right governments cutting taxes (lighter lines) only minimally increases the public belief that capital is more elastic (from 0.027 in the first panel to 0.037 in the last panel). However, increasing the number of left governments cutting taxes (dark lines) greatly increases the public belief that capital is more elastic (from 0.13 in the first panel to 0.99 in the last panel). This simulation illustrates the predictions from our theory. In the next section, we test the following two hypotheses:

Hypothesis 1: The more left governments cut corporate taxes, the more likely other governments, both left and right, are to cut their own corporate tax rates.

Hypothesis 2: The number of right governments cutting corporate taxes has no effect on tax policy decisions of other governments, both left and right.

There are some additional hypotheses that emerge from our model. Although we do not empirically test those additional hypotheses, we briefly mention them here for the sake of completeness. First, no tax cuts by right governments can lead to a slight decrease in the beliefs that tax cuts are conducive to attracting foreign capital. Second, just like left governments lead the way on tax cuts, right governments should lead the way on tax increases. Here, however, we are primarily concerned with the issue of tax cuts, not least because of the rare occurrence of tax increases.

One limitation of most existing empirical studies on learning and competition is that the presence of common shocks can complicate statistical identification. In the context of corporate tax policy, countries may cut taxes at the same time, suggesting competition or learning when, in fact, tax cuts could be driven by a common exogenous shock. For example, a decrease in the world supply of FDI (i.e., a recession in the countries with the most outbound FDI) could lead all governments to change tax policy.

Our model and empirical analysis can address this issue of common shocks. First, our theory directly models public beliefs about taxation and explores changes in these public beliefs. An exogenous change that is common knowledge to all countries would directly affect public beliefs. Yet our key assumption, a prominent one in much of the political economy literature, is that right governments are more likely to reduce corporate taxes than left governments. Common shocks would not produce a pattern of left governments leading the way in tax cuts, followed by all other governments. We also directly address this issue of common shocks in our empirical analysis.

Learning or Competition: Data and Method

Following Hypotheses 1 and 2, we explicitly test for partisan differences in learning by introducing measures of tax cuts by both left and right governments. Thus, our interest is in exploring whether tax cuts by right or left governments trigger tax cuts by other governments. We hypothesize, in accordance with the predictions generated at the end of the previous section, that learning takes place when left governments cut taxes. We stress that this pattern of diffusion is inconsistent with existing arguments about (pure) tax policy competition.

We use two data sets to test our theoretical model. The first is the data set used by Basinger and Hallerberg (2004), who take corporate tax rate changes as the dependent variable. Although we believe that these rate changes are important to multinationals making investment decisions, the focus of most of the social science literature is on the tax policy reforms of governments. Despite the fact that rate changes and policy reforms may seem equivalent, tax policy reforms occur when the legislature passes a corresponding law, whereas rate changes occur when these rates are applied to corporate taxes. In our empirical analysis, we focus on the most transparent, and often most political, tax policies, nominal corporate tax rates.

Using a number of data sources we coded announced tax policy reforms in 18 OECD countries.²¹ Our data cover the period 1980 to 1998. This time period allows us to focus on tax diffusion within the OECD (since during this time period all of the main competitors for FDI are within the OECD) and also provides for direct comparability with Basinger and Hallerberg’s (2004) analysis and results.²² A number of contemporary examples of tax policy cuts help to motivate this new data set. In many cases, governments pass a single law that affects corporate tax rates, yet the year that it is applied varies. Austria’s 2004 tax reform proposed lowering corporate tax rates from 34% to 25%, taking effect in 2005. Thus, the policy reform was enacted in 2004, and the rate change became effective in 2005. In contrast, Ireland passed a corporate tax reform in December 1997 that increased corporate tax rates from 10% to 12.5%, yet the change did not go into effect until 2003. Other countries, such as Canada, have recently (and historically) passed tax policy reforms where rate changes are phased in over a period of years. Perhaps the largest corporate tax reform in Canada’s history was introduced in 2000, phasing in corporate tax rate reductions over 5 years. Thus, a single tax policy reform is associated with five tax rate changes. In these cases, if one assumed that tax reforms were passed the year before a tax rate change came into effect (a rate change in say 1985 is assumed to have been passed through a tax reform in 1984), this would lead to a correct classification of the Austrian cut, an incorrect coding of the Irish tax reform (coding the reform as having occurred in 2002 when, in fact, the law was passed in 1997), and multiple incorrect codings of the Canadian tax cut (rather than one law change in 2000, one would code five legislative changes). Thus, there are major differences between when a tax rate comes into effect and when a tax policy reform is passed.

In an online appendix (available at http://nathanjensen.wustl.edu/) we present the descriptive data on the timing of tax policy cuts (coded by us) and the observed rate changes. We compare years in which there is a change in the rate of corporate taxation and years of statutory corporate tax reform as coded by us.²³ To be clear, if a country has a rate change from, for example, 45% in year t to 40% in year t + 1, we code this as a tax rate change in year t + 1. For our own data, we code tax policy cuts when major pieces of tax legislation are passed (as explained above).

Of the 360 total cases in the data set, there are a total of 295 agreements in coding and 65 cases of disagreement. The large number of cases of agreement is the result of the large number of zeros (no tax cuts) in the data. Comparing our data on observed tax policy cuts and tax rate changes, there are only 14 cases of agreement (3.9% of the cases). We find 39 cases of Type I error, where the rate change data indicate tax policy cuts, yet we find no evidence of an actual tax law being enacted. We also find 26 cases of Type II error, where our data indicate a major tax policy reform, yet there is no rate change. In our empirical analysis we use both data sets, yet we stress that models of policy choice are best tested using data on policy changes, not implementation (i.e., tax rates). In models of learning and competition that make arguments about temporal patterns, we believe that identifying the date of the policy change is necessary to ensure accuracy of the empirical analysis.

In our empirical analysis we rely on two types of models. In the first set of models we estimate the determinants of tax rate changes. We replicate the work of Basinger and Hallerberg (2004), using a standard ordinary least squares (OLS) model with panel-corrected standard errors. Our dependent variable is the change in the corporate tax rate (Change in Central Tax Rate). The key independent variable is our measure of corporate tax policy cuts in the n − 1 other countries lagged by one year (World Tax Law Changes (t − 1)). The model takes the following form.

In the second set of models we focus solely on our data collected on tax policy cuts and do not use data on tax rates. We estimate the probability of tax law changes in two standard time-series cross-sectional logit models of the following form.

We also estimate a hierarchical model, in which we allow the intercept to randomly vary across countries. We believe that a hierarchical model is the most appropriate one in this context, given (a) the hierarchical structure of our data and (b) the built-in ability of this model to test for, rather than assume, the degree of country-level variance beyond that accounted for by the fixed effects.

The key independent variables are the government’s partisanship (Partisanship) as measured by Laver and Hunt (1992) and coded by Tsebelis (2002) and the ideological distance between veto players (Ideological Distance).²⁴ Partisanship is coded on a 0–1 continuum, using the Basinger and Hallerberg (2004) standardization of the Laver and Hunt (1992) expert survey measure, where left governments (coded as 0) are more likely to resist corporate tax cuts because of the higher domestic costs associated with this policy change for left governments. Ideological Distance is the distance between veto players, based on the Laver and Hunt (1992) expert survey and constructed by Tsebelis (2002). The measure captures not only the number of veto players but also the ideological distance between these veto players. We expect left governments and those with large ideological distances between veto players to find it more difficult to pass tax policy reforms.

Following Basinger and Hallerberg (2004), we include a number of control variables. First, we control for the lagged corporate tax rate level (Tax Rate (t − 1)). The expectation is that countries with higher rates are more likely to respond to tax policy competition. Second, we include Basinger and Hallerberg’s economic controls, including the lagged level of economic growth

Δ_{(t, t - 1)} c e n t r a l r a t e_{i j} = X_{i j} β + Z_{j} ζ + ε_{i j}, for i = 1, \dots, 299 and j = 1, \dots, 20

(2)

(Growth Rate (t − 1)) and the lagged level of inflation (Inflation Rate (t − 1)). Basinger and Hallerberg argue that countries with higher levels of economic growth are more easily able to reduce corporate taxes and still maintain adequate tax revenue. The effect of inflation is more complicated, where higher

PR (t a x l a w c u t_{i}) = L o g i t^{- 1} (X_{i} β), for i = 1, \dots, 299

(3)

levels of inflation can lead to either more tax revenues (through growth in profits) or a decline in effective tax rates. In their empirical analysis, Basinger and Hallerberg find weak support for the argument that higher levels of inflation are associated with more tax reform. Finally, we include the variable Capital Controls (Simmons, 1999), which measures capital controls on an 8-point scale using the International Monetary Fund International Financial

\begin{array}{l} PR (t a x l a w c u t_{i}) = L o g i t^{- 1} (α_{j [i]} X_{i} β), for i = 1, \dots, 299 \\ α_{j} \sim N (μ_{α}, σ_{α}^{2}), for j = 1, \dots, 20 \end{array}

(4)

Statistics.Countries with lower levels of control on capital flows are expected to more likely respond to other countries’ tax policy reforms. We provide details on all of these variables in the online appendix (see Table 2; available at http://nathanjensen.wustl.edu/).

Basinger and Hallerberg (2004) also use a number of weights to generate variables representing the international environment. These include averages of the rest of the world for capital controls, ideological distance, and changes in capital taxation in competitor countries. In this replication we take the simple weight of the n – 1 countries in the world, where for a country (say Germany) Capital Controls (World), Ideological Distance (World), and Competitor Taxation Changes (t − 1) are the mean values of the other 19 OECD countries. Competitor Taxation Changes (t − 1) is the key variable measuring competition.

Our own empirical analysis examines only the impact of observing tax cuts in other countries, and not the size of the tax cuts. Incorporating the size of the tax cuts into our analysis would require additional assumptions, such as how to code tax cuts that are phased in over a series of years. Our theoretical model does not provide predictions on how a single policy reform, phased over a number of years, affects public knowledge, as compared to a tax reform that is implemented in one year. Rather than resort to a subjective ranking of tax policy reforms, we focus on our objective coding scheme of tax reforms that is directly derived from our theoretical model. In the next section, we present our empirical analysis of determinants of tax policy reforms in the OECD.

Learning From the Left: Empirical Evidence

We present the empirical results from the replication of Basinger and Hallerberg (2004) in Model 1 of Table 1.²⁵ We also test a model of tax rate change by substituting their variable for tax competition, Competitor Taxation Changes (t − 1), with our own variables on tax policy cuts. Our measures, World Tax Law Changes (t − 1; all governments), Right Tax Law Changes (t − 1), and Left Tax Law Changes (t − 1) are the number of tax policy cuts in the other countries (excluding the observed country) in year t − 1. The intuition for both models is that countries are responding to tax policy cuts in other countries.²⁶ We expect the coefficients on World Tax Law Changes (t − 1), Right Tax Law Changes (t − 1), and Left Tax Law Changes (t − 1) to be positive since tax policy cuts are coded as 1 and the absence of tax policy cuts as 0. In both models tax rate changes and tax policy cuts are associated with tax cuts.

Table 1.

Ordinary Least Squares Models of Tax Rate Changes

Variable	Model 1	Model 2	Model 3
Partisanship	−3.021^† (1.703)	−2.504 (1.651)	−2.896^† (1.705)
Partisanship (World)	0.136 (7.339)	8.145 (8.878)	1.535 (8.308)
Capital Controls	0.143 (1.729)	0.169 (1.683)	0.282 (1.734)
Capital Controls (World)	14.403** (3.860)	13.843** (3.949)	13.927** (3.982)
Ideological Distance	0.095 (1.489)	0.498 (1.508)	0.385 (1.491)
Ideological Distance (World)	−7.591 (6.844)	−1.310 (8.469)	−0.625 (8.661)
Tax Rate (t − 1)	−0.265** (0.064)	−0.264** (0.063)	−0.257** (0.064)
Growth Rate (t − 1)	−0.103 (0.084)	−0.090 (0.086)	−0.132 (0.083)
Inflation Rate (t − 1)	−0.081** (0.007)	−0.084** (0.007)	−0.083** (0.007)
Competitor Taxation Changes (t − 1)	0.438** (0.146)
World Tax Law Changes (t − 1)		−0.256** (0.075)
Right Tax Law Changes (t − 1)			0.005 (0.143)
Left Tax Law Changes (t − 1)			−0.706** (0.224)
Constant	11.115* (4.772)	5.572 (5.734)	8.918^† (5.592)
Country dummies	Yes	Yes	Yes
N	299	299	299
R ²	.28	.29	.28

Panel-corrected standard errors in parentheses.

Significance levels: ^†10%. *5%. **1%.

Our control variables are in line with conventional work on the determinants of tax policy cuts. We are more likely to observe tax cuts in years when other countries have low levels of restrictions on flows of international capital. Also, high tax countries and countries with high levels of inflation are more likely to reform corporate taxes.

In Model 2 we use our variable counting the number of tax policy reforms in the n – 1 other countries (World Tax Law Changes (t − 1)). Models 1 and 2 show that tax changes, whether measured as rate reductions or tax policy cuts, are associated with rate reductions in other countries.

In Model 3 we differentiate between tax policy cuts by left governments and right governments. Similar to Model 2, we include counts of the number of tax policy cuts by right governments, Right Tax Law Changes (t − 1), and the number of tax policy cuts by left governments, Left Tax Law Changes (t − 1).²⁷ The results from this model indicate that tax cuts are driven by the tax policy cuts of left governments. The negative coefficient on Left Tax Law Changes (t − 1) indicates that a tax policy cut by a left government leads to a reduction in the corporate tax rates of the n – 1 other countries.

Dichotomizing government partisanship to be either “left” or “right” for the tax law changes variables might introduce measurement error, since, for example, slightly left-of-center governments are coded the same as far-left governments. We address this issue by reestimating our models using a measure that weights the tax law changes by the level of partisanship, and substituting these new measures for the tax law changes variables constructed with the dichotomous partisanship variable. Our results are similar across the specifications and do not depend on whether we use the dichotomous measure or the weighted measure of partisanship.²⁸ Thus, irrespective of how we measure partisanship, the effects of the tax law changes variables provide strong support for both Hypothesis 1 and Hypothesis 2.

One potential criticism of our empirical analysis is that left governments cutting taxes may be significantly different from right governments choosing to cut taxes. For example, although most left governments are loath to cut corporate taxes, a small set of left governments may see themselves as very competitive in attracting FDI. Thus, our coding of left tax policy cuts could be proxying for effectiveness in attracting international capital.

Of the 40 country-years with tax law changes lowering corporate tax rates, right governments enacted 29 and left governments enacted 11. Given that the major diffusions of tax policy were driven by these left government policy changes, we explored if there were significant differences between country-years when left governments enacted tax reforms and (a) the rest of the country-years in the sample as well as (b) country-years when right governments enacted tax reforms. We examined if there were statistically significant differences in the amount of capital attracted by these governments (FDI as a percentage of GDP), the size of the domestic market (log of GDP), and the countries’ level of development (log of GDP per capita).²⁹ All of these factors are associated with being a more attractive investment environment. We find no statistically significant differences between left governments cutting taxes and the rest of the governments in the sample, or between left governments cutting taxes and right governments cutting taxes. We are confident that our variable that counts the number of left governments cutting taxes is not proxying for competitive governments cutting taxes. We also tested these models by weighting tax cuts by country GDP, where cuts in large countries are more likely to lead to competitive pressures to reduce taxes. The weighted results are similar to the nonweighted results. Tax cuts by left governments lead to tax cuts in other countries, whereas tax cuts by right governments have no impact on tax cuts in other countries.

In Table 2 we present the results of two logit models using the passage of tax policy cuts as the dependent variable and including country dummy variables (Models 4 and 5).³⁰ Note that the coefficients on Left Tax Law Changes (t − 1) were negative in Table 1 (cuts by left governments lead to reductions in tax rates), whereas in this set of regressions we expect a positive coefficient. We expect corporate tax policy cuts in left countries to positively affect other countries’ propensity to enact tax policy cuts.

Table 2.

Logit Models of Tax Law Change

Variable	Model 4	Model 5	Model 6
Partisanship	2.304 (2.096)	2.386 (2.141)	1.081 (1.432)
Partisanship (World)	−3.620 (11.796)	1.539 (11.157)	3.897 (10.380)
Capital Controls	0.802 (2.632)	0.919 (2.667)	−1.430 (1.656)
Capital Controls (World)	−1.085 (4.609)	−0.871 (4.674)	−1.654 (4.130)
Ideological Distance	−2.652 (1.986)	−2.716 (2.028)	−2.082 (1.304)
Ideological Distance (World)	−11.103 (10.216)	−15.371 (10.572)	−14.380 (9.909)
Tax Rate (t − 1)	0.024 (0.043)	0.016 (0.044)	0.048^† (0.029)
Growth Rate (t − 1)	−0.002 (0.109)	0.022 (0.108)	−0.011 (0.100)
Inflation Rate (t − 1)	0.025 (0.030)	0.025 (0.030)	0.015 (0.015)
World Tax Law Changes (t − 1)	0.151^† (0.087)
Right Tax Law Changes (t − 1)		−0.231 (0.189)	−0.157 (0.171)
Left Tax Law Changes (t − 1)		0.735* (0.307)	0.558* (0.274)
Constant	−2.791 (7.626)	−4.657 (7.275)	−3.935 (6.808)
Country-level SD (ML)			≈ 0
Country dummies	Yes	Yes	Yes
N	299	299	299
AIC	227.3	226.4	218.0

Standard errors in parentheses. Country-Level SD refers to the standard deviation of the country-level error in the hierarchical model. AIC = Akaike information criterion. Loosely speaking, the lower the AIC the better the model fit.

Significance levels: ^†10%. *5%. **1%.

Model 4 in Table 2 controls for the country-level factors from Table 1. As in Models 2 and 3 of Table 1, we include a count of tax policy laws reducing corporate taxes in the n – 1 other countries (World Tax Law Changes (t − 1); Model 4) and the number of tax policy laws passed by both left and right governments (Left Tax Law Changes (t − 1) and Right Tax Law Changes (t − 1); Model 5). Although these regressions use the same set of control variables as the previous OLS regressions, many of the control variables are no longer statistically significant. We believe that one reason for this is that the limited amount of tax policy reforms (11% of the observations) makes estimation difficult. Yet we note that our key empirical results on social learning from left governments are robust to different empirical specifications.³¹ As in Table 1, we find that although World Tax Law Changes (t − 1) has a positive and statistically significant effect on the tax policy decisions of governments, these results are driven by the tax policy decisions of left governments.

We have examined the robustness of these results under a number of alternative specifications. One important concern is the issue of temporal dependence where observations of tax cuts at time t may not be independent of tax cuts made in pervious years. Beck, Katz, and Tucker (1998) provide a methodology to deal with these issues of duration dependence in panel models with binary dependent variables. Following Beck et al., we estimate Models 4 and 5 by including a count of the years since the last tax cut and three smoothing splines.³² Our results on the relationship between left tax cuts and learning are robust, and there is no indication that duration dependence is an issue in our data.³³

A second set of robustness tests examines if our results are driven by omitted variable bias. Our empirical analysis is built on the specification of Basinger and Hallerberg (2004). Other influential articles cited in this article—Hays (2003), Swank and Steimo (2002), and Swank (2006), among others—include additional control variables. We test the robustness of Models 4 and 5 by including measures of unemployment, level of trade, level of FDI, and government budget balance. Moreover, we estimated these models using tax cuts weighted by GDP as our key independent variable.³⁴ Our core empirical results are unchanged.

A third robustness test builds on the work of Hays (2003) and Devereux, Lockwood, and Redoano (2008). Both argue that the existence of capital controls mediates tax competition. According to these models of tax competition, a government is likely to respond to tax cuts by other governments when capital controls are present. We test the interactive effect of capital controls and both the overall number of tax cuts and the number of tax cuts by right and left governments. We find that capital controls have no independent or mediating effect on responding to other countries’ tax cuts. Thus, we are even more confident that our empirical results capture effects of learning and not tax competition.

Finally, we estimate a hierarchical logit model, in which we regress tax policy cuts on the independent variables described earlier. This model accounts for country-level variance by including a random intercept.³⁵ The results, which are given in Table 2, Model 6, confirm our theoretical predictions. Left Tax Law Changes (t − 1) is positive and statistically significant, whereas Right Tax Law Changes (t − 1) is not statistically significant. The lagged tax rate, Tax Rate (t − 1), is again positive and marginally statistically significant, indicating that countries with higher tax rates are more likely to cut taxes than countries with already low tax rates. The standard deviation of the country-level errors (Country-level SD (ML) in Table 2) is very small, indicating that there is very little variation across countries beyond what is accounted for by the fixed effects.³⁶ This indicates that there is quite a large degree of pooling, suggesting that it is not necessary to include country fixed effects.³⁷

One criticism of diffusion studies is that common shocks can complicate statistical identification. For tax policy, a common shock that leads all governments to cut corporate taxes is also observationally equivalent to learning and competition. Our project mitigates this concern in that our social learning model has observational implications that are distinct from common shocks. We argue that it is the partisanship of the government that predicts the diffusion of tax policy. Our empirical results are consistent with our theoretical model and inconsistent with a common shock argument.³⁸

The substantive impact of tax policy cuts by left governments is quite large. In Figure 2, we present postestimation results generated with the R (R Development Core Team, 2009) package Zelig (Imai, King, & Lau, 2007a, 2007b, 2008). Specifically, Figure 2 shows the impact that a tax policy cut by a left government has on countries’ probabilities of implementing tax policy cuts. For the simulations, on which the results in Figure 2 are based, we set all of the variables, except for Right Tax Law Changes (t − 1) and Left Tax Law Changes (t − 1), at their means. The solid line in Figure 2 shows a situation in which neither a left nor a right government changes its tax laws.

Figure 2.

Marginal impact of left tax policy cuts on the probability of tax policy cuts

In the OECD sample, only 11% of the country-year observations are tax policy cuts. From our simulations, we find that the impact of one tax law change by a left government is quite substantial. During years when there were no tax law changes in the previous year, the expected probability of a tax law change is 8%. When a single left government chooses to cut tax policy in year t, the probability of each individual country implementing a tax policy cut in year t + 1 jumps to 14%, an increase of 75%. Although major tax policy cuts are still relatively uncommon, we find considerable evidence of learning (consistent with our theory).

Our empirical analysis suggests that social learning influences the observed patterns of temporally clustered tax policy cuts. Our replication of Basinger and Hallerberg (2004) indicates that tax rate reductions by left governments drive rate reductions in other countries, whereas rate reductions by right governments have no impact on tax rates. Our analysis of the timing of legislative tax reforms confirms this pattern: Legislative tax reforms by left governments, and only those by left governments, trigger global tax changes.

Conclusion

In this article we engage an important substantive debate about tax competition in the global economy by offering a new model of tax policy reform. Our social learning model, which is distinct from existing models of competition, suggests waves of corporate tax cuts are consistent with cross-country learning. In our empirical analysis we propose a research design strategy to analyze observed tax policy rate changes and the exact timing of legislative changes in corporate tax policy, which differentiates our social learning model from existing models of tax competition. We find that the waves of tax reform in the OECD are driven by partisan learning, specifically in the form of emulation of left government tax policy reforms. Although competitive dynamics may still be present, the observed patterns of tax policy reforms are inconsistent with existing models of (pure) tax competition.

Our theoretical model has empirical implications distinct from tax competition that could be integrated into broader models of competition. That is, we can imagine a more general model that includes both elements of social learning and competitive dynamics. There might also be other forms of learning that could work in parallel with social learning. We outline two potential extensions below. First, we make the assumption that right and left governments have different underlying preferences on tax policy. Although we believe this to be a defensible assumption, one could refine our theory to identify specific parties or individual politicians that are less likely to cut corporate taxes. This refinement would require a significant data collection effort, but the new data could easily be incorporated into our existing model.

Second, we would argue that there are alternative models of social learning that could also affect government policy. One potential extension of our work is to examine the emulation of other governments’ successful taxation strategies. This could take the form of simply adopting the tax policies that have proven the most successful in attracting FDI. Yet as we noted in previous sections, identifying the exact impact of tax policy on FDI is difficult. More importantly, politicians enact tax policy changes not only for economic reasons but also with clear electoral goals in mind. An examination of how corporate tax cuts affect the reelection prospects of incumbent politicians would be an interesting extension of our work. We would predict that left governments that gain legislative seats by enacting corporate tax reductions would lead to an emulation of these strategies by other governments.

Our key contribution has been the theoretical and empirical demonstration of social learning effects in international tax policy making. A sizeable literature spanning the fields of economics, finance, and political science has been arguing for pure competitive dynamics in international tax policy making. In this article, we have shown that corporate tax policy making does not follow this pattern of pure competition. Throughout the article we have maintained that social learning does not preclude underlying competitive dynamics. However, our evidence for social learning does show that purely competitive, “race-to-the-bottom”-type dynamics do not tell the whole story with respect to international tax policy making. Governments do not simply cut corporate taxes because other governments have done so. Rather, governments attempt to learn new information from the policy choices of other governments and imitate other governments only when new information points to the viability of the policy.

Footnotes

Acknowledgements

We have benefited from the comments and criticisms from members of the audience at numerous conferences and speaker series. We would like to thank Ken Benoit, Pablo Beramendi, Christian Bjørnskov, Randall Calvert, John Freeman, Matthew Gabel, Fabrizio Gilardi, Mark Hallerberg, Jude Hays, Layna Mosley, David Singer, Andrew Sobel, Duane Swank, and Craig Volden for useful comments and suggestions. We also thank Scott Basinger and Mark Hallerberg for sharing their data with us and Michael Dickerson and Thomas Zeitzoff for excellent research assistance.

The authors declared no potential conflicts of interest with respect to the authorship and/or publication of this article.

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Nathan Jensen would like to thank Washington University’s Weidenbaum Center on the Economy, Government, and Public Policy for financial support.

Notes

References

Andrews

(1994). Capital mobility and state autonomy: Toward a structural theory of international relation. International Studies Quarterly, 38, 193-218.

Bailey

Alimadhi

(2007). logit.mixed: Mixed effects logistic regression. In Imai

King

Lau

(Eds.), Zelig: Everyone’s statistical software. Retrieved from http://gking.harvard.edu/zelig

Banerjee

A. V.

(1992). A simple model of herd behavior. Quarterly Journal of Economics, 107, 797-817.

Basinger

S. J.

Hallerberg

(2004). Remodeling the competition for capital: How domestic politics erases the race to the bottom. American Political Science Review, 98, 261-276.

Baturo

Gray

(2009). Flatliners: Ideology and rational learning in the adoption of the flat tax. European Journal of Political Research, 48, 130-159.

Bearce

D. H.

(2007). Monetary divergence: Domestic policy autonomy in the post-Bretton Woods era. Ann Arbor, MI: University of Michigan Press.

Beck

Katz

J. N.

Tucker

(1998). Taking time seriously: Time-series-cross-sectional analysis with a binary dependent variable. American Journal of Political Science, 42, 1260-1288.

Bernhard

Broz

Clark

(2002). The political economy of monetary institutions: An introduction. International Organization, 56, 693-723.

Bernhard

Leblang

(1999). Democratic institutions and exchange rate commitments. International Organization, 53, 71-97.

10.

Bernhard

Leblang

(2006). Democratic processes and financial markets: Pricing politics. New York, NY: Cambridge University Press.

11.

Bikhchandani

Hirshleifer

Welch

(1992). A theory of fads, fashion, custom, and cultural change as informational cascades. Journal of Political Economy, 100, 992-1026.

12.

Bikhchandani

Hirshleifer

Welch

(1998). Learning from the behavior of others: Conformity, fads, and informational cascades. Journal of Economic Perspectives, 12, 151-170.

13.

Bonoli

(2007). Time matters: Postindustrialization, new social risks, and welfare state adaption in advanced industrialized democracies. Comparative Political Studies, 40, 495-520.

14.

Boskin

M. J.

McLure

C. E.

Jr. (1990). World tax reform: Case studies of developed and developing countries. San Francisco, CA: ICS Press.

15.

Brooks

S. M.

(2002). Social protection and economic integration: The politics of pension reform in an era of capital mobility. Comparative Political Studies, 35, 491-523.

16.

Calvert

R. L.

(1985). The value of biased information: A rational choice model of political advice. Journal of Politics, 47, 530-555.

17.

Cerny

P. G.

(1990). The changing architecture of politics. London, UK: SAGE.

18.

Chamley

C. P.

(2004). Rational herds: Economic models of social learning. New York, NY: Cambridge University Press.

19.

Davies

R. B.

Voget

(2009). Tax competition in an expanding European Union (Discussion Paper No. 276). Dublin, Ireland: Institute for International Integration Studies

20.

De Mooij

R. A.

Ederveen

(2003). Taxation and foreign direct investment: A synthesis of empirical research. International Tax and Public Finance, 10, 673-693.

21.

Devereux

M. P.

Lockwood

Redoano

(2008). Do countries compete over corporate tax rates? Journal of Public Economics, 92, 1210-1235.

22.

Drezner

D. W.

(2005). Globalization, harmonization, and competition: The different pathways to policy convergence. Journal of European Public Policy, 12, 841-859.

23.

Elkins

Simmons

(2005). On waves, clusters, and diffusion: A conceptual framework. Annals of the American Academy of Political and Social Sciences, 598, 33-51.

24.

Franzese

R. J.

Hays

J. C.

(2007). Spatial-econometric models of cross-sectional interdependence in political science panel and times-series-cross-sectional data. Political Analysis, 15, 140-164.

25.

Franzese

R. J.

Hays

J. C.

(2008). Interdependence in comparative politics: Substance, theory, empirics, substance. Comparative Political Studies, 41, 742-780.

26.

Frieden

J. A.

(1991). Invested interests: The politics of national economic policy in a world of global finance. International Organization, 45, 425-451.

27.

Frieden

J. A.

(2002). Real sources of European currency policy: Sectoral interests and European monetary integration. International Organization, 56, 831-860.

28.

Gale

(1996). What have we learned from social learning? European Economic Review, 40, 617-628.

29.

Ganghof

Eccleston

(2004). Globalisation and the dilemmas of income taxation in Australia. Australian Journal of Political Science, 39, 519-534.

30.

Garrett

(1998). Partisan politics in the global economy. New York, NY: Cambridge University Press.

31.

Garrett

Lange

(1991). Political responses to interdependence: What’s “left” for the left? International Organization, 45, 539-564.

32.

Gelman

Hill

(2007). Data analysis using regression and multilevel/hierarchical models. New York, NY: Cambridge University Press.

33.

Hall

P. A.

Soskice

(2001). An introduction to varieties of capitalism. In Hall

P. A.

Soskice

(Eds.), Varieties of capitalism (pp. 1-68). New York, NY: Oxford University Press.

34.

Hallerberg

Basinger

S. J.

(1998). Internationalization and changes in tax policy in OECD countries: The importance of domestic veto players. Comparative Political Studies, 31, 321-352.

35.

Han

Leach

(2008). A bargaining model of tax competition. Journal of Public Economics, 92, 1122-1141.

36.

Hansson

Olofsdotter

(2008). Integration and the structure of public spending. Comparative Political Studies, 41, 1001-1027.

37.

Haufler

(2001). Taxation in a global economy. New York, NY: Cambridge University Press.

38.

Hays

J. C.

(2003). Globalization and capital taxation in consensus and majoritarian democracies. World Politics, 56, 79-113.

39.

Henisz

W. J.

Zelner

B. A.

Guillen

M. F.

(2005). The worldwide diffusion of market-oriented infrastructure reform. American Sociological Review, 70, 871-897.

40.

Imai

King

Lau

(2007a). logit: Logistic regression for dichotomous dependent variables. In Imai

King

Lau

(Eds.), Zelig: Everyone’s statistical software. Retrieved from http://gking.harvard.edu/zelig

41.

Imai

King

Lau

(Eds.). (2007b). Zelig: Everyone’s statistical software. Retrieved from http://gking.harvard.edu/zelig

42.

Imai

King

Lau

(2008). Toward a common framework for statistical analysis and development. Journal of Computational and Graphical Statistics, 17, 892-913.

43.

Jensen

N. M.

(2006). Nation-states and the multinational corporation. Princeton, NJ: Princeton University Press.

44.

King

Zeng

(1999a). Estimating absolute, relative, and attributable risks in case-control studies. Retrieved from http://gking.harvard.edu

45.

King

Zeng

(1999b). Logistic regression in rare events data. Retrieved from http://gking.harvard.edu

46.

King

McAfee

R. P.

Welling

(1993). Industrial blackmail: Dynamic tax competition and public investment. Canadian Journal of Economics, 80, 361-375.

47.

Kiymaz

Taylor

(2000). Competition for foreign direct investment when countries are not sure of site values. International Review of Economics and Finance, 9, 53-68.

48.

Kurzer

(1993). Business and banking: Political change and economic integration in Western Europe. Ithaca, NY: Cornell University Press.

49.

Laver

Hunt

W. B.

(1992). Policy and party competition. New York, NY: Routledge.

50.

Lohmann

(1994). The dynamics of informational cascades: The Monday demonstrations in Leipzig, East Germany, 1989–91. World Politics, 47, 42-101.

51.

Meseguer

(2004). What role for learning? The diffusion of privatization in OECD and Latin American countries. Journal of Public Policy, 24, 299-325.

52.

Organisation for Economic Co-operation and Development. (2004). Recent tax policy trends and reforms in OECD countries (OECD Tax Policy Studies). Paris, France: Author.

53.

Pechman

J. A.

(1988). World tax reform: A progress report. Washington, DC: Brookings Institution.

54.

Przeworski

Wallerstein

(1982). The structure of class conflict in democratic capitalist societies. American Political Science Review, 76, 215-238.

55.

R Development Core Team. (2009). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.

56.

Rodrik

(1997). Has globalization gone too far? Washington, DC: Institute for International Economics.

57.

Rodden

Rose-Ackerman

(1997). Does federalism preserve markets? Virginia Law Review, 83, 1521-1572.

58.

Simmons

B. A.

(1994). Who adjusts? Domestic sources of foreign economic policy during the interwar years. Princeton, NJ: Princeton University Press.

59.

Simmons

B. A.

(1999). The internationalization of capital. In Kitchelt

Lange

Marks

Stephens

J. D.

(Eds.), Continuity and change in contemporary capitalism (pp. 36-69). New York, NY: Cambridge University Press.

60.

Simmons

B. A.

Dobbin

Garrett

2006. Introduction: The international diffusion of liberalism. International Organization, 60, 781-810.

61.

Steinmo

(2002). Globalization and taxation: Challenges to the Swedish welfare state. Comparative Political Studies, 35, 839-862.

62.

Swank

(1992). Politics and the structural dependence of the state in democratic capitalist nations. American Political Science Review, 86, 38-54.

63.

Swank

(1998). Funding the welfare state: Globalization and the taxation of business in advanced market economies. Political Studies, 46, 671-692.

64.

Swank

(2006). Tax policy in an era of internationalization: Explaining the spread of neoliberalism. International Organization, 60, 847-882.

65.

Swank

Steimo

(2002). The new political economy of taxation in advanced capitalist democracies. American Journal of Political Science, 46, 642-655.

66.

Taxes: A race to the bottom. (2005, April 4). BusinessWeek. Retrieved from http://www.businessweek.com/magazine/content/05_14/b3927071_mz054.htm

67.

Tomz

King

Zeng

(1999). RELOGIT: Rare events logistic regression, version 1.1. Retrieved from http://gking.harvard.edu/

68.

Tsebelis

(2002). Veto players: How political institutions work. Princeton, NJ: Princeton University Press.

69.

Volden

Ting

M. M.

Carpenter

D. P.

(2008). A formal model of learning and policy diffusion. American Political Science Review, 102, 320-332.

70.

Wildasin

D. E.

(1989). Interjurisdictional capital mobility: Fiscal externality and a corrective subsidy. Journal of Urban Economics, 25, 193-212.

71.

Wilson

J. D.

(1986). A theory of interregional tax competition. Journal of Urban Economics, 25, 193-212.

72.

Wilson

J. D.

(1999). Theories of tax competition. National Tax Journal, 52, 269-304.

73.

Wilson

J. D.

Wildasin

D. E.

(2004). Capital tax competition: Bane or boon. Journal of Public Economics, 88, 1065-1091.

74.

World Bank. (2004). World development indicators (CD-ROM). Washington, DC: Author.