Determinants of Tax Incentives for Foreign Direct Investors by arnold1


									                                             Web Appendix for
            Democracy, Autocracy, and Tax Incentives to Foreign Direct Investors:
                                        A Cross-National Analysis

This web-appendix reports a number of additional tests for verifying the robustness of the results

reported in Table 1 of the text. The sensitivity analyses are reported in Table 2.

        First, one may argue that in federalist countries, subnational governments often compete

with each other and with the central government in offering incentives. If democracies are more

likely to be decentralized than autocracies, the empirical finding that democracies offer lower levels

of incentives could be an artifact of fiscal decentralization and the competition between the central

and local governments and among the local governments.1 Models 1 and 2 test this possibility using

two different measures of federalism, respectively. The first measure federalism1 is coded 1 if a

country has a constitutionally guaranteed division of power between central and regional

governments with autonomy in at least one area for each level of governance and 0 otherwise,

following Treisman (2000). The second measure federalism2 is coded 1 if the provincial level

government has authority over taxing, spending, or legislating and 0 otherwise, taken from Beck et

al. (2001). Federalism1 is statistically insignificant in Model 1 while federalism2 is statistically significant

and positive in Model 2. The findings are reasonable. A general type of federalism is not associated

systematically with tax incentives. But fiscal decentralization itself brings about higher levels of

incentives, a result consistent with the theoretical proposition of Flochel and Madies (2002). More

important, with decentralization controlled for in both models, the effects of all the key variables

 In a theoretical model of tax competition between two overlapping governments (federal and local)
sharing the same tax base, Flochel and Madies (2002) show that interjurisdictional tax competition
reduces the global tax rate set by both federal and local governments. Tung and Cho (2001)
empirically show that areas in China that offer lower tax rates and higher tax incentives attract more

(rule of law, transition, democracy, FDI inflow, their interaction term, autocfdir, autocfdio and their inequality

test) remain robust and consistent with those in Table 1.

        Second, one may suspect that the labor endowment conditions may have either confounded

or motivated the regime type results in Table 1. On the one hand, many of the developing countries

in the sample are labor abundant. Because the abundant factor labor arguably benefits from FDI in

terms of more jobs, higher productivity and wage rates, labor may support tax incentives that

encourage FDI inflows. Democratic governments that seek labor’s political support will adopt more

generous incentive programs. Hence, the results on democracy in Table 1 should be stronger once

we control for labor endowment. On the other hand, in some democratic developing countries,

opposition to FDI may also come from labor unions, thus resulting in populist government policies.

The regime type effects in Table 1 may have originated from labor opposition in democracy. I need

to evaluate whether it is labor’s influence that drives the regime type results. Furthermore, MNCs

may be attracted to labor abundant countries because they can exploit low-cost labor and use these

countries as export platforms. This may motivate a type of compensating effect similar to that

produced by the rule of law. To assess whether labor endowment affects the findings in Table 1,

Model 3 employs the commonly-used measure, the labor-capital ratio, which is computed as a

country’s population between ages 15 and 64 divided by its total physical capital stock in 1987

constant dollars, log transformed to address skewed distribution.2 High values of labor-capital ratio

imply capital scarcity and labor abundance while low values suggest capital abundance and labor

scarcity. Model 3 shows that the labor-capital ratio is statistically significant and negative. Countries

  Data on the total physical capital stock in 1987 constant domestic price and the population
between ages 15 and 64 are collected from Nehru and Dhareshwar (1993). The total physical capital
stock in 1987 constant domestic price is converted into the 1987 constant US dollar. The labor-
capital ratio for 1990 is used in analysis because the 1990 data on physical capital stock are the most
recent from Nehru and Dhareshwar’s comprehensive database. While not perfect, the measure is a
reasonably good proxy, given data availability, because factor endowment conditions tend to be
stable over time.

that are abundant in labor tend to offer lower levels of tax incentives than those scarce in labor. The

result is plausible because abundant labor may not be effectively organized to lobby for more

incentives but scarce domestic capital may put up more effective opposition. Or low labor cost

eliminates the need for more generous incentives to compensate the investors. The most important

findings in this test are that the effects of all the key independent variables remain robust and

consistent as those in Table 1, showing that they are not an artifact of alternative mechanisms

related to labor. The labor-capital ratio result also is consistent with an earlier observation that rival

host firms are not the only source of opposition to incentives to foreign firms.

        Third, one reviewer suggests that some countries with abundant natural resource

endowment such as oil and minerals offer tax incentives to attract large amounts of FDI. The

governments of these countries often are autocratic and suffer from credibility deficit due to past

expropriations of foreign assets (Kobrin 1984). The rule of law effect on tax incentives in Table 1

may arguably be driven by the presence of natural resource endowment. I measure natural resource

endowment using the logged resource variable from Jensen (2003), which is the share of fuel and

mineral exports in merchandise exports. Model 4 shows that the resource variable is negative but

statistically insignificant. In contrast, the rule of law variable remains negative and highly significant.

The effect of the rule of law is not an artifact of natural resource endowment. The effects of other

regime type variables remain robust as in Table 1, with one minor exception. The autocracy without

restriction over foreign entry now offers a significantly higher level of incentives than the semi-

regime country.

        Fourth, governments often sign bilateral investment treaties with property rights guarantees

for foreign investors and through this channel, communicate expectations and information about

capital account openness (Simmons and Elkins 2004). These treaties therefore may provide an

alternative explanation for the effect of the rule of law variable. Model 5 includes the total number

of bilateral investment treaties (totalbit) for each country, collected from the Country Commercial

Guide. The effect of totalbit is statistically significant and positive. One interpretation of this result

is that the number of bilateral investment treaties by a country signals its desire to actively pursue

foreign capital under the influence of policy diffusion, thus positively associated with its level of tax

incentives. The results for the key variables are consistent with those in Table 1, with one minor

exception. The democracy at mean level of FDI inflows now offers the same level of incentives as

the semi-regime country.

        Fifth, as noted, the dependent variable includes only incentives that apply to all products,

which is most appropriate for testing the theoretical argument. One may ask whether the results in

Table 1 are sensitive to the alternative measure that also includes incentives that apply to imported

or exported goods only. Model 6 presents this analysis. The results for the key variables are

consistent with those in Table 1, except for some interesting differences. These differences suggest

that both democracies and autocracies are selective in choosing tax incentives. First, the democratic

country at mean level inflows adopts incentives of all types at the same level as the semi-regime

country. Even though rising FDI inflows above the mean level still raise societal opposition, the

level at which societal pressures affect policymaking is much higher for these narrow, selective tax

incentives. Democratic governments may take more time designing incentive programs that cause

less societal opposition. Second, the autocracy that retains restrictions over foreign entry now offers

the same level of incentives as the semi-regime country. The difference between the two types of

autocracies is much weaker with respect to the alternative dependent variable measure. This is

consistent with the notion that even authoritarian regimes that have conflicts of interest with foreign

capital may use tax incentives selectively to strengthen their own interest. While they oppose general

tax incentives that may hurt their own interest, they remain strategic and flexible in utilizing foreign

capital to their own advantage.

          Finally, the dependent variable measure may be argued to suffer from the same problem as

many other quantitative indicators of policy outcomes. Within each of the six incentive types, there

is only one dichotomous choice. The coding of 1 or 0 is based on the presence or absence of a

particular type of incentive by a host country. A generous corporate tax break, for example, is coded

as 1, and so is a small corporate tax break. One implication is that a package of 3 types of incentives

with larger magnitude, for example, could be at the same level on the latent continuum of incentives

as a package of 4 or even 5 types of incentives with smaller magnitude. The possibility is more

probable for adjacent scores and the high end values on the index. Despite the weakness, one

justification for using the ordinal index is that different types of incentives appear to be

complements with positive correlations between them in the sample (with correlation mostly

between 0.1 and 0.4). More types imply higher levels on the latent scale. A second justification is

that an incentive offer that is generous only in one type may be weakened by limitations and

restrictions on other dimensions, especially when it is up to the bureaucrat to interpret the laws and

regulations. As such, a package of more types of incentives is likely to be an offer at a higher level

on the latent scale, because less interference and restriction by the government is involved. Still, it is

worth investigating the robustness of the results in Table 1 to the possibility that adjacent index

values (such as 1 and 2) or high values on the index reflect equally attractive offers. I therefore

collapse the 0-6 scale into a 0-2 scale (merging 1 and 2 into 1, and merging values 3 and above into

2).3 The results are presented in Model 7. It is reassuring that the effects of all the key variables

remain consistent with those in Table 1.

    The results are not sensitive to other alternative coding schemes.


Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh. 2001. “New tools

in comparative political economy: The Database of Political Institutions.” World Bank Economic

Review 15(September): 165-76.

Flochel, Lourent and Thierry Madies. 2002. “Interjurisdictional Tax Competition in a Federal

System of Overlapping Revenue Maximizing Governments.” International Tax and Public Finance 9:


Nehru, Vikram and Ashok Dhareshwar. 1993. “A New Database on Physical Capital Stock: Sources,

Methodology and Results.” Rivista de Analisis Economico 8(1): 37-59.

Treisman, Daniel. 2000. “Decentralization and Inflation: Commitment, Collective Action, or

Continuity?” American Political Science Review 94(December): 837-57.

Tung, Samuel and Stella Cho. 2001. “Determinants of Regional Investment Decisions in China: An

Econometric Model of Tax Incentive Policy.” Review of Quantitative Finance and Accounting 17: 167–85.

Table 2 Effect of Political Regime Type on Level of Tax Incentives: Sensitivity Analyses

Variable              Model 1       Model 2    Model 3    Model 4     Model 5    Model 6    Model 7
Rule of law           -0.416**      -0.455**   -0.561**   -0.512**    -0.730**   -0.935**   -0.574**
                      (0.210)       (0.204)    (0.180)    (0.277)     (0.143)    (0.227)    (0.237)
Transition            -0.139        -0.104     0.152      -0.775      -0.076     -0.277     -0.394
                      (0.743)       (0.686)    (1.069)    (0.541)     (0.608)    (0.775)    (0.688)
Democracy             -0.554**      -0.672**   -1.136**   -0.707**    -0.368     0.049      -0.280*
                      (0.218)       (0.174)    (0.425)    (0.429)     (0.342)    (0.281)    (0.206)
Democracy*Inflow      -0.233**      -0.260**   -0.350**   -0.248**    -0.222**   -0.118**   -0.129**
                      (0.047)       (0.031)    (0.064)    (0.070)     (0.048)    (0.047)    (0.055)
Inflow                0.213**       0.227**    0.228**    0.236**     0.232**    0.150**    0.119**
                      (0.024)       (0.015)    (0.047)    (0.041)     (0.018)    (0.031)    (0.021)
Autocfdir             -1.777**      -1.860**   -2.731**   -1.659**    -2.130**   -0.016     -1.120**
                      (0.851)       (0.716)    (1.599)    (0.772)     (0.762)    (0.747)    (0.678)
Autocfdio             0.523         0.813      0.295      1.021**     0.602      0.401      0.685
                      (0.875)       (0.887)    (1.350)    (0.569)     (0.771)    (0.588)    (0.975)
Autocfdir < Autocfdio 6.10**        36.32**    59.70**    17.77**     36.43**    2.29*      5.86**

Federalism1             0.288       --         --         --          --         --         --
Federalism2             --          0.793**    --         --          --         --         --
Labor/Capital           --          --         -0.216**   --          --         --         --
Resource                --          --         --         -0.050      --         --         --
Totalbit                --          --         --         --          0.023**    --         --
External pressure      1.061**     1.046**     0.868       1.161**    1.401**    0.649**    1.132**
                       (0.131)     (0.192)     (0.854)     (0.147)    (0.276)    (0.109)    (0.323)
External pressure2     -0.034** -0.033** -0.029            -0.037**   -0.044**   -0.018**   -0.036**
                       (0.004)     (0.006)     (0.023)     (0.005)    (0.008)    (0.003)    (0.010)
Development            0.127       0.047       -0.071      -0.037     0.172      0.494**    0.262
                       (0.189)     (0.131)     (0.367)     (0.336)    (0.305)    (0.298)    (0.291)
Size                   0.026       0.028       0.195       0.051      -0.098     -0.186**   0.122
                       (0.176)     (0.108)     (0.153)     (0.075)    (0.156)    (0.077)    (0.142)
μ0                     6.717       5.892       8.750       5.721      6.916      3.018      10.780
                       (3.095)     (3.167)     (10.272) (2.550)       (1.817)    (2.056)    (2.866)
μ1                     7.280       6.478       9.272       6.432      7.495      3.349      12.539
                       (3.174)     (3.217)     (10.349) (2.488)       (1.957)    (2.075)    (2.975)
μ2                     8.547       7.789       10.685      7.655      8.807      4.479      --
                       (3.399)     (3.463)     (10.036) (2.390)       (2.062)    (1.943)
μ3                     9.458       8.699       12.067      8.779      9.750      5.597      --
                       (3.551)     (3.637)     (9.921)     (2.051)    (2.197)    (1.836)
μ4                     10.034      9.270       12.793      11.473     10.352     6.116      --
                       (3.559)     (3.633)     (10.349) (2.211)       (2.221)    (2.009)
μ5                     11.501      10.871      14.013      --         11.980     7.266      --
                       (3.439)     (3.405)     (9.940)                (2.112)    (1.866)
N                      51          51          42          42         51         51         51
** p<0.05; * p<0.1. Robust standard errors in parentheses.

                List of Countries in Estimation Sample
Albania                 Ghana                            Peru
Algeria                 Guatemala                        Philippines
Argentina               Guinea                           Poland
Bangladesh              Honduras                         Romania
Bolivia                 Hungary                          Russia
Botswana                India                            Singapore
Brazil                  Indonesia                        Slovak Rep
Bulgaria                Jamaica                          South Africa
Chile                   Jordan                           Sri Lanka
China                   Kenya                            Thailand
Colombia                Korea                            Trinidad Tobago
Costa Rica              Malawi                           Turkey
Cote d'Ivoire           Malaysia                         Uganda
Czech Rep               Mexico                           Uruguay
Dom Rep                 Nicaragua                        Venezuela
Ecuador                 Pakistan                         Zimbabwe
Egypt                   Panama
El Salvador             Paraguay


To top