Capital taxation, investment, growth, and welfare

Abstract

This paper formulates a model of economic growth to study the effects of broad capital taxation (of profits, dividends, and capital gains) on macroeconomic outcomes in small open economies. A framework of exogenous growth permits modeling countries in transition to a country-specific steady state and to discern steady-state and transitory effects of shocks on economic outcomes. The chosen framework is amenable to structural estimation and, in view of the parsimony of the model, fits data on 79 countries over the period 1996–2011 well. The counterfactual analysis based on the estimated model suggests that capital-tax reductions induce positive effects on output and the capital stock (per unit of effective labor) that are economically significant and are accommodated within time windows of 5 years without much further economic response after that. The responses of economic aggregates are found to be relatively strongest to changes in corporate-profit-tax rates and weaker for dividend and capital-gains taxes.

Appendices

Appendix 1: Current-value Hamiltonian formalizing the household problem

Writing the maximization problem of households for generic period t as a current-value Hamiltonian, we obtain

$$\begin{aligned} H = u(c) + \underbrace{\mu ~\exp ^{-(\rho - n) t}}_{a} \left( (1-\tau _d) d^d e^d + (1-\tau _g) \dot{q} e^d + r e^{\star d} q^{\star } + w + f - c \right) , \end{aligned}$$

where $\mu$ and a are the present-value and current-value Lagrange multipliers of wealth, respectively. The resulting first-order conditions (FOCs) for the choice variables of interest to households are

$$\begin{aligned} \frac{\partial H}{\partial c} = 0 \quad \rightarrow \quad u_c(c) = a, \end{aligned}$$

$$\begin{aligned} \frac{\partial H}{\partial e^d} = (\rho - n) a - \dot{a} \quad \rightarrow \quad (1-\tau _d) \frac{d^d}{q} + (1-\tau _g) \frac{\dot{q}}{q} = \rho - n - \frac{\dot{a}}{a} \end{aligned}$$

(16)

and

$$\begin{aligned} \frac{\partial H}{\partial e^{\star d}} = (\rho -n) a - \dot{a} \quad \rightarrow \quad r = \rho - n - \frac{\dot{a}}{a}. \end{aligned}$$

(17)

In Eq. (16) we assume, as in Turnovsky and Bianconi (1992), that individuals take the dividend yields on their equity as given, which allows us to express the FOCs in terms of dividend yield, $\frac{d^d}{q}$, and the growth-rate of the value of equity, $\frac{\dot{q}}{q}$. Equation (17) is the long-run (steady-state) arbitrage condition. In equilibrium, the rate of return on investment in the domestic country, on the left-hand side of Eq. (16), has to match the net rate of return on investment in the rest of the world, r.

Appendix 2: Present-discounted value of the firm

Total dividend payments are paid either to domestic individuals, $D^d$, or to foreign ones, $D^f$, according to their equity, $E^d$ and $E^f$, respectively. The total value of equity in the economy is $V = q E$. The change in the total value of equity in the domestic economy equals the change in the value of equity plus the change in equity:

$$\begin{aligned} \frac{\partial V}{\partial t} = \dot{V} = \dot{q} E + q \dot{E}. \end{aligned}$$

(18)

Total investment has to be equal to retained earnings, R, plus the change in equity, $q \dot{E}$:

$$\begin{aligned} I = \dot{K} + \delta K = q \dot{E} + R. \end{aligned}$$

Combining Eqs. (1) and (18) with (B), we obtain a first-order differential equation for the change in total value of equity:

$$\begin{aligned} \dot{V} = \frac{r}{1-\tau _g} V - \left( \frac{\tau _g-\tau _d}{1-\tau _g}\right) D - \gamma , \end{aligned}$$

(19)

where $\gamma \equiv (1 - \tau _p) \Pi - I$ are the net profits less investment.

We define

$$\begin{aligned} E \equiv E^d + E^f \quad \text {and} \quad D \equiv D^d + D^f \end{aligned}$$

and take the derivative of $V=qE$ to obtain

$$\begin{aligned} \frac{\partial V}{\partial t} = \dot{V} = \dot{q} E + q \dot{E}, \end{aligned}$$

(20)

substitute R from equation (B) in Eq. (1), and solve for $q \dot{E}$:

$$\begin{aligned} q \dot{E} = D + \dot{K} + \delta K - (1 - \tau _p) \Pi . \end{aligned}$$

(21)

Substituting (21) in Eq. (20) results in

$$\begin{aligned} \dot{V} = \dot{q} E + D + \dot{K} + \delta K - (1 - \tau _p) \Pi , \end{aligned}$$

which, after defining $\gamma \equiv (1 - \tau _p) \Pi - \dot{K} - \delta K$, may be written as

$$\begin{aligned} \dot{V} = \dot{q} E + D - \gamma . \end{aligned}$$

(22)

Solve (16) for $\dot{q}$ using the arbitrage condition from the FOCs in Eqs. (16) and (17) to obtain

$$\begin{aligned} \dot{q} = \frac{r}{1-\tau _g} q - \frac{1-\tau _d}{1-\tau _g} \frac{D^d}{E^d}. \end{aligned}$$

Substituting this in Eq. (22) and using $V\equiv qE$ as well as the fact that the dividend yields are the same for domestic and foreign individuals, $\frac{D^d}{V^d} = \frac{D^f}{V^f}$, which implies that $E^d D^f = E^f D^d$, obtains

$$\begin{aligned} \begin{aligned} \dot{V}&= \frac{r}{1-\tau _g}V - \frac{1-\tau _d}{1-\tau _g} \frac{D^d}{E^d} (E^d + E^f) + D - \gamma \\&= \frac{r}{1-\tau _g} V - \frac{1-\tau _d}{1-\tau _g} \frac{D^d E^d + D^f E^d}{E^d} + D - \gamma \\&= \frac{r}{1-\tau _g} V - \frac{1-\tau _d}{1-\tau _g} D + D - \gamma \\&= \frac{r}{1-\tau _g} V + \left( 1- \frac{1-\tau _d}{1-\tau _g}\right) D - \gamma \\&= \frac{r}{1-\tau _g} V - \left( \frac{t_g-\tau _d}{1-\tau _g}\right) D - \gamma , \\ \end{aligned} \end{aligned}$$

which gives Eq. (19).

Assuming that the firm distributes a share $\phi \in (0,1)$ as dividends and retains $1 - \phi$ of the profits, dividend payments are defined as:

$$\begin{aligned} D = \phi (1-\tau _p) \Pi . \end{aligned}$$

Substituting this in Eq. (19) yields,

$$\begin{aligned} \dot{V} = \frac{r}{1-\tau _g} V - (1-\tau ) \left( F(A,K,L) - wL - I \psi \right) + I, \end{aligned}$$

(23)

where we define for notational convenience $1- \tau \equiv \frac{(\phi (1-\tau _d) + (1-\phi ) (1 - \tau _g)}{1-\tau _g} (1-\tau _p)$. In the steady state, $\dot{V} = 0$ and Eq. (23) implies that the value of the firm then satisfies

$$\begin{aligned} V = \frac{\gamma (1-\tau _g)+ (\tau _g-\tau _d)D}{r}. \end{aligned}$$

(24)

From the individual FOCs in Eqs. (16) and (17) it follows that in the steady state with $\dot{q} = 0$ the firm value has to be

$$\begin{aligned} V = \frac{(1-\tau _d)D}{r}. \end{aligned}$$

(25)

Both Eqs. (24) and (25) have to hold simultaneously so that in equilibrium the individual valuation of the firm is consistent with its optimal behavior, which is the case if $D=\gamma$. As the dividends are $D = \phi (1-\tau _p) \Pi$, this implies that all investments are financed by retained earnings.³⁴ We assume that the dividend/payout ratio is constant and the representative firm does not optimize its capital structure with regard to the capital-gains and dividend tax rates. For an individual firm the model would suggest that the dividend/payout ratio depends on the capital-gains and dividend tax rates and, consequently, the dividend- payout ratio would be either zero or unity for an individual firm.³⁵

We may integrate Eq. (23) to obtain the present value of a firm:

$$\begin{aligned} V(0) = \int _0^{\infty } \left( (1-\tau ) \left( F(A,K,L) - wL - I \psi \right) - I \right) \exp \left( -\int ^t_0 \frac{r}{1 - \tau _g} dm \right) \mathrm{d}t. \end{aligned}$$

Appendix 3: $\tilde{k}$ and $\lambda$

The FOCs corresponding to the Hamiltonian in Eq. (3) are

$$\begin{aligned} \frac{\partial J}{\partial I}= & {} 0 \rightarrow \frac{\eta -1}{b (1-\tau )} = \frac{\hat{i}}{\hat{k}}, \end{aligned}$$

(26)

$$\begin{aligned} \frac{\partial J}{\partial K}= & {} \eta \left( \frac{r}{1-\tau _g} \right) -\dot{\eta } \rightarrow \dot{\eta } = \left( \frac{r}{1-\tau _g} + \delta \right) \eta - (1-\tau ) \left( f_{\hat{k}}(\hat{k}) + \frac{b}{2} \left( \frac{\hat{i}}{\hat{k}}\right) ^2 \right) , \nonumber \\ \end{aligned}$$

(27)

where we use the homotheticity of the production function and the capital-adjustment-cost function to write all expressions in terms of units of effective labor.

Steady-state equilibrium

Using $\dot{\hat{k}} = \hat{i} - (x + n +\delta ) \hat{k}$ in Eq. (26), the change of $\hat{k}$ is expressed as

$$\begin{aligned} \dot{\hat{k}} = \left( \frac{\eta -1}{b (1-\tau )} - (x + n + \delta ) \right) \hat{k}. \end{aligned}$$

(28)

Since $\dot{\hat{k}} = 0$ in the steady state, the solution to the above equation for the steady-state value, $\tilde{\eta }$, is

$$\begin{aligned} \tilde{\eta } = 1 + b ( x + n + \delta )(1-\tau ), \end{aligned}$$

which is independent of $\hat{k}$ and represents a horizontal locus in a phase diagram with $\hat{k}$ on the abscissa. Similarly, we rewrite Eq. (27) and substitute Eq. (26) to obtain

$$\begin{aligned} \dot{\eta } = \left( \frac{r}{1-\tau _g} + \delta \right) \eta - (1-\tau ) f_{\hat{k}}(\hat{k}) -\frac{(\eta - 1)^2}{2b (1-\tau )}. \end{aligned}$$

(29)

The two differential Eqs. (28) and (29) may now be used to construct the phase diagram of the dynamic system in $\hat{k}$-$\eta$-space. The $\dot{\hat{k}} = 0$ locus represents Eq. (28), whereas the $\dot{\eta } = 0$ locus represents Eq. (29). The latter is downward-sloping near the steady-state, if $\frac{r}{1-\tau _g} > x + n$. To show this, we use that $\dot{\eta } = 0$ and $\tilde{\eta }$ from Eq. (5) in Eq. (29) and apply the implicit function theorem to obtain $\frac{\partial \eta }{\partial \hat{k}} = \frac{(1 - \tau ) f_{\hat{k}\hat{k}}(\hat{k})}{\frac{r}{1-\tau _g} - (x + n)}$. As $f_{\hat{k}\hat{k}}(\hat{k}) < 0$, we have $\frac{\partial \eta }{\partial \hat{k}} < 0$, if and only if $\frac{r}{1-\tau _g} > x + n$. Hence, the $\dot{\eta } = 0$-locus is downward sloping around the steady state. Using that in $\dot{\eta } = 0$ and substituting Eq. (5) yields the steady-state value of capital per unit of effective labor, $\tilde{\hat{k}}$, as:

$$\begin{aligned} \tilde{\hat{k}}= \left( \frac{1}{\alpha } \left( \left( \frac{r}{1-\tau _g} + \delta \right) \frac{1}{1-\tau } + (x + n + \delta )b \left( \frac{r}{1-\tau _g}+\frac{\delta }{2} - \frac{x}{2} - \frac{n}{2} \right) \right) \right) ^{\frac{1}{\alpha -1}}. \end{aligned}$$

Dynamics We are interested in the dynamics of the capital stock, investment, and output after changes in capital taxation. Therefore, we linearize the two dynamic Eqs. (28) and (29) around their steady states to obtain an analytic expression of the speed of convergence. Considering

$$\begin{aligned} \left( \begin{array}{l} \dot{\hat{k}} \\ \dot{\eta } \end{array} \right) = \begin{bmatrix} 0&\quad \frac{\tilde{\hat{k}}}{b(1-\tau )} \\ - (1-\tau ) f_{\hat{k}\hat{k}}(\tilde{\hat{k}})&\quad \frac{r}{1-\tau _g} + \delta - \left( \frac{\tilde{\eta } - 1}{b(1-\tau )}\right) \end{bmatrix} \times \left[ \begin{array}{c} (\hat{k} - \tilde{\hat{k}}) \\ (\eta - \tilde{\eta }) \end{array} \right] , \end{aligned}$$

after substituting $\tilde{\eta }$ from Eq. (5) and $\tilde{\hat{k}}$ from Eq. (4), the two eigenvalues, $\lambda _{1,2}$, are

$$\begin{aligned} \lambda _{1,2} = \frac{\zeta }{2} \pm \left( \left( \frac{\zeta }{2} \right) ^2 + \frac{(1-\alpha )}{b} \left( \left( \zeta + \omega \right) \frac{1}{1-\tau } + \omega b \left( \zeta +\frac{\omega }{2} \right) \right) \right) ^{\frac{1}{2}}. \end{aligned}$$

where $\zeta = \frac{r}{1-\tau _g} - (x+n)$ and $\omega = x + n + \delta$.

Appendix 4: Transfers

We write the government’s budget constraint as a function of capital:

$$\begin{aligned} \begin{aligned} F&= \tau _p \Pi + \tau _d D + \tau _g \dot{q} E, \\&= \tau _p \Pi + \tau _d \phi (1-\tau _p) \Pi + \tau _g E \left( \frac{r}{1-\tau _g} q - \frac{1-\tau _d}{1-\tau _g} \frac{D}{E} \right) , \\&= (\tau _p + \tau _d \phi (1-\tau _p) - \frac{\tau _g}{1-\tau _g} (1-\tau _d)(1-\tau _p) \phi ) \Pi + \frac{\tau _g}{1-\tau _g} rK, \end{aligned} \end{aligned}$$

where total transfers F have to be equal to total tax revenues. Tax revenues per unit of effective labor, $\hat{f}$, are given by

$$\begin{aligned} \hat{f}= & {} (\tau _p + \tau _d \phi (1-\tau _p) - \frac{\tau _g}{1-\tau _g} (1-\tau _d)(1-\tau _p) \phi ) \left( \alpha \hat{k}^{\alpha } - \frac{b}{2} \left( \frac{\hat{i}}{\hat{k}} \right) ^2 \hat{k} \right) \nonumber \\&+\, \frac{\tau _g}{1-\tau _g} r \hat{k}. \end{aligned}$$

(30)

Appendix 5: Taxes

1.1 Corporate-profit-tax rates ($\tau _p$)

For corporate-profit taxes, we utilize the maximum tax rate levied at the national level on corporate profit in a country of residence. In federal states, the total corporate tax rate is calculated as the weighted average of the local (sub-national) taxes combined with federal tax rates (e.g., for Germany or Canada as reported by the OECD) or the tax rate prevailing in the economic center (e.g.,for Switzerland, where the rates of the canton of Zurich are taken).

The primary sources for corporate-profit-tax rates are the following:

• Ernest and Young Worldwide Corporate Tax Guide 1998–2012

• Coopers and Lybrand International Tax Summaries 1996–1997

• International Bureau of Fiscal Documentation Global Corporate Tax Handbook 2007–2012

• Price Waterhouse Coopers Corporate Taxes - Worldwide Summaries 1999–2000, 2001–2003, 2012–2013

• OECD www.taxfoundation.org

1.2 Capital-gains-tax rates ($\tau _g$)

For corporate capital-gains taxes, we utilize the maximum tax rate levied at the national level on corporate capital gains in a country of residence.

The primary sources for corporate capital-gains-tax rates are the following:

• Ernest and Young Worldwide Corporate Tax Guide 1998–2012

• Coopers and Lybrand International Tax Summaries 1996–1997

• International Bureau of Fiscal Documentation Global Corporate Tax Handbook 2007–2012

• Price Waterhouse Coopers Corporate Taxes—Worldwide Summaries 1999–2000, 2001–2003, 2012–2013

1.3 Dividend-tax rates ($\tau _d$)

For corporate dividend taxes, we utilize the maximum tax rate levied at the national level on after-tax income, classified as dividends, which are distributed to (mostly corporate) shareholders. Shareholder taxation is not taken into account. If an imputation system (as, e.g., in Australia) is stated in the tax code, the tax rate on dividends is set to zero. In some cases, tax codes allow for differentiated dividend-tax rates conditional on holding requirements of the recipient. Then, we choose tax rates applied to dividends paid to corporations holding 10% or less of the distributing entity. In such cases, dividend taxation decreases with higher shares held by the receiving entity and the dividend-tax rates represent the upper bound upper bound.

Table 6 Capital taxes

Fig. 7

Corporate-profit tax. 79 countries, 1996–2011

The primary sources for corporate dividend-tax rates are the following:

• Ernest and Young Worldwide Corporate Tax Guide 1998–2012

• Coopers and Lybrand International Tax Summaries 1996–1997

• International Bureau of Fiscal Documentation Global Corporate Tax Handbook 2007–2012

• Price Waterhouse Coopers Corporate Taxes - Worldwide Summaries 1999–2000, 2001–2003, 2012–2013

The summary statistics for the three capital-taxes rates is shown in Table 6. Figures 7, 8 and 9 present the cross-country distribution of the tax rates for each year in the data, using whisker plots. The area around the median (a horizontal bar) indicated by a box refers to the interquartile range (IQR), whereas the extended lines, the whiskers, indicate values within a maximum of 1.5 times the IQR around the median. The corporate-profit-tax rates in Fig. 7 show a relatively high degree of variability over time, even at the median. The median capital-gains-tax rate in Fig. 8 decreases smoothly and only modestly over the sample period. The distribution of dividend-tax rates in Fig. 9 is skewed toward zero with the median being constant at zero throughout the sample period.

Fig. 8

Capital-gains tax. 79 countries, 1996–2011

Fig. 9

Dividend tax. 79 countries, 1996–2011

Appendix 6: Total asset holdings and home bias

See Table 7.

Table 7 Total assets holdings and home bias.

Appendix 7: Robustness

In this section, we analyze the robustness of the our baseline results in column (4) of Table 2. In the model, heterogeneity in terms of the speed-of-convergence parameter arises from macroeconomic fundamentals such as capital shares in output. If the capital-adjustment-cost parameter would be country specific, even more heterogeneity could be introduced. In this section, we check the persistence of our results when allowing capital-adjustment costs to vary systematically with GDP per capita, average capital taxes, and time. Moreover, we consider different measures of the dividend/payout ratio, $\phi$. We briefly discuss the effects of taking the effective tax rate instead of the statutory tax rate. Under the aforementioned specifications, estimated capital-adjustment costs do not significantly change and neither does the average convergence parameter $\lambda$ relative to the benchmark results. Lastly, we restrict our sample to OECD countries. All estimations are based on our preferred specification, namely column (4) in Table 2, which uses the first-differenced convergence equation and a control function; see Eq. (15). All sensitivity checks are summarized in Table 3.

Country heterogeneity

Capital-adjustment costs might vary systematically with GDP per capita and the general level of capital taxation in a country. To control for this kind of heterogeneity, we parameterize the adjustment-cost factor as follows: $b = \xi _1 + \xi _2 z$, where z is either GDP per capita or total capital tax, ($\tau _p + \tau _d + \tau _g$). We normalize the mean GDP per capita and the mean total capital tax to zero. Hence, we can interpret the estimate of $\xi _1$ as the average capital-adjustment cost parameter, $\overline{b}$, and $\xi _2$ as the effect of variable z on the dispersion of the capital-adjustment-cost parameter around the mean.

We estimate the parameterized capital-adjustment-cost parameter using our preferred specification. The results are given in columns (1) and (2) of Table 3. The average capital-adjustment costs are slightly higher than in our preferred specification, and countries with higher GDP per capita have lower capital-adjustment costs, while high-tax countries have higher capital-adjustment costs. But these effects are far from being statistically significant. Moreover, the average $\lambda$ is very close to the one without considering the effect of GDP per capita or total capital tax rates.

Time trends and pre-Financial-Crisis results

Technological progress or common, unobserved time-trending factors might reduce capital-adjustment costs over time. To control for this effect, we again parameterize b and take the year 1998 as a base year with $z=0$. The results for the corresponding estimation are given in column (3) of Table 3. The coefficient $\xi _1$ is much higher than in all other specifications, but as we take the year 1998 as a base year, we can only interpret it as the average capital-adjustment cost in that year. We find that capital-adjustment costs are falling with time, but this effect is not significant. We cannot reject the hypothesis that the average capital-adjustment costs in that specification are different from the ones in the benchmark specification in column (4) of Table 2.

Next, we restrict the sample to observations prior to the Financial Crisis (up until 2007) in column (4) of Table 3. According to the results, the average capital-adjustment-cost parameter is not significantly different from the benchmark estimate, and the convergence parameter $\lambda$ is almost identical to the reference estimate.

Dividend/payout ratio $\phi$

In our preferred specification we use the average observed $\phi$ over all countries. In column (5) of Table 3, we use the observed dividend/payout ratios for the subset of countries for which we have data. In column (6) we estimate $\phi$ as an independent variable. The results in terms of capital-adjustment costs, b, speed of convergence, $\lambda$, and Tobin’s q are all very similar to the benchmark estimates in column (4) of Table 2. Moreover, the estimated $\phi$ in column (6) of Table 3 is not significantly different from the dividend/payout ratio of about 0.61 we observe for a subset of 70 countries for which we have data.

To compute $\phi$, we took for each country the total dividends paid by each firm over the whole sample period from Standard and Poor’s Compustat data set and divided them by the total profits of each firm over the whole sample period. Compustat contains almost 24,000 companies in 70 countries between 1996 and 2012. We excluded firms where either dividend payments or income was missing or negative. Summary statistics for these variables are provided in Table 8.

Table 8 Dividend/payout ratio

The share of corporate after-tax profits which is paid as dividends to shareholders is calculated as:

$$\begin{aligned} \phi _j= \frac{\text {Total dividends}}{\text {Net income}} = \frac{{\sum ^{T}_{t=0} \sum ^{S_j}_{s_j=0} D_{s}(t)}}{{\sum ^{T}_{t=0} \sum ^{S_j}_{s_j=0} I_{s}(t)}}, \end{aligned}$$

where $s_j = 1,2,\ldots ,S_j$ indicate firms in country j and $T= 17$ years (1996–2012).

• Total dividends: To calculate total dividends we use the annual amount of dividends in current million US dollars from the corporate-income statement as stated in Standard and Poor’s Compustat database.

• Net income: To calculate net income we use annual income after all expenses in current million US dollars from the corporate-income statement as stated in Standard and Poor’s Compustat database.

OCED countries

Our baseline estimation in column (4) of Table 2 includes a broad set of countries. In column (8) of Table 3 we restrict our sample to OECD countries. Neither the capital-adjustment cost parameter b nor the speed-of-convergence parameter $\lambda$ is significant different from our baseline estimation.

Effective tax rates

In the previous estimations, we used the statutory tax rates for two reasons. First, ex ante effective tax rates include already a behavioral response of a model firm (e.g., its investment structure and its financing structure), and ex post effective tax rates (i.e., the actual tax revenues generated from the average firm relative to the tax base) depend even more on firm-level responses (e.g., due to the location decisions and tax avoidance through transfer pricing, profit shifting and debt shifting). Second, the effective average tax rates and statutory tax rates on corporate profits are highly correlated. Figure 10 plots the country rank of the statutory corporate-profit tax against the rank of the effective average corporate-profit tax 1998 and 2010. Countries that have high statutory tax rates have usually also high effective average tax rates, as is indicated by the clearly positive correlation. Moreover, we find comparable estimates to the preferred specification using ex ante effective tax rates on corporate profits instead of statutory ones as a further robustness check in column (7) in Table 3. The estimate for the capital-adjustment-cost parameter b is slightly lower, around 15, and the convergence parameter $\lambda$ is not significantly different from the one in our preferred specification with this specification.

Fig. 10

Plot of rank of statutory corporate-profit tax against the rank of effective average corporate-profit tax rate, 1998–2010

Table 9 Summary results

Appendix 8: Numerical effects across countries

In this subsection, we describe how the four different tax policies affect a broad set of countries. Tables 9, 10, 11 and 12 show the numerical results in terms of short-run, medium-run, and long-run elasticities of output per unit of effective labor for some selected countries. We also present the country-specific speed-of-convergence parameter and the time it takes until the gap between the (observed) output and the steady-state output is less than 1E−7.

Table 9 gives the results for the general capital-tax increase by one percentage point. Gabon has the strongest decline in steady-state capital stocks (among the 79 countries), where the increase leads to more than a 3.5% decline in the long-run output. The effect is lowest for Iceland with a steady-state elasticity of −0.427. In terms of present-discounted utility, Thailand faces the biggest welfare loss, while Egypt faces the smallest one.

Table 10 Summary results

In Table 10, we consider only an increase of the corporate-profit tax by one percentage point. As discussed for the representative country, higher corporate-profit taxes might increase the consumption and the present-discounted utility in some countries. For the majority of countries a higher corporate-profit tax implies a higher utility. The model predicts that Egypt has the biggest gain, while the Philippines face a welfare loss after an increase of the corporate-profit tax by one percentage point. In terms of output, the Gabon again have the strongest decline by more than 1.9%, while Iceland has the lowest one.³⁶

Table 11 Summary results

Table 12 Summary of results

The effects of capital-gains taxation are shown in Table 11. The effect in terms of output is very small for Barbados, which comes from the fact that the capital-gains tax in Barbados is zero, while the dividend tax is 15%. Increasing the capital-gains tax (holding the dividend/payout ratio constant) decreases the distortions introduced by capital taxation and hence leads to only very small changes. On the other hand, the effects are strongest for Peru, which has zero dividend taxes and a 30% capital-gains tax. In the latter case, the distortions actually increase more as small differences between dividend and capital-gains taxes increase the capital stock. In terms of welfare (all) countries face negative effects, the strongest ones in Mexico and the weakest in Barbados.

Last, we present the effects for dividend taxation in Table 12. Output declines stronger under dividend taxation than under capital-gains taxation. The effects are strongest in Panama and weakest in Iceland. The welfare effects are mixed for dividend taxation: Some countries actually gain from higher dividend tax rates (e.g., Egypt), while most countries would loose.

In general, a marginal corporate-profit-tax policy has the strongest impact on output per unit of effective labor. Dividend taxation reduces output per unit of effective labor by more than capital-gains taxes in all countries but Argentina, Peru, and Tunisia. The welfare effects are negative for all countries when considering the general tax increase by one percentage point for all capital taxes. Almost all countries in our sample are predicted to gain from a higher corporate-profit-tax rates, while the results are mixed for dividend taxation, and higher capital-gains taxes are found to have negative welfare effects for all countries.

Fig. 11

Short-run elasticities of output per unit of effective labor after increasing all capital taxes by one percentage point. Base year 2008. 79 countries

Fig. 12

Speed-of-convergence parameter $\lambda$. Base year 2008. 79 countries

Appendix 9: Short-run elasticities and speed of convergence

Figure 11 shows the short-run elasticities of output per unit of effective labor after increasing all capital taxes by one percentage point for all 79 countries. The short-run effect of higher capital taxation is very strong in South American countries, South East Asian countries, African countries, and the Middle East. The country with the highest output elasticity is Gabon (−0.902). Most developed countries have rather low elasticities of output, with Iceland having the lowest one (−0.154).

Figure 12 illustrates the speed-of-convergence parameters for different countries. Most developed countries converge relatively slowly to their new steady state, while most developing countries converge rather fast. The short-run elasticities and the speed-of-convergence parameters $\lambda$, are negatively correlated with an unconditional coefficient of correlation of −0.65.