Chapter

5 Income Taxes and Investment Some Empirical Relationships for Developing Countries

Author(s):
Ved Gandhi, Liam Ebrill, Parthasarathi Shome, Luis Manas Anton, Jitendra Modi, Fernando Sanchez-Ugarte, and George Mackenzie
Published Date:
June 1987
Share
  • ShareShare
Show Summary Details
Author(s)
Liam P. Ebrill

There has been growing interest in the issue of how supply-side policies might be used to affect the growth rates of developing countries. In particular, it has been suggested that taxes may well be powerful instruments of development policy. Central to this position is the presumption that tax policy can influence the level of investment. In a neoclassical framework, this presumption is sustained by the argument that taxes can be employed to reduce the cost of capital where the latter is viewed as the principal determinant of the level of fixed business investment. This chapter is written in the tradition of that neoclassical analysis. As a preliminary to answering the question of how best to direct tax policy so as to stimulate investment, one objective of this chapter is to quantify the impact of existing tax systems on the cost of capital in a selection of developing countries.

Of course, other (nonneoclassical)factors such as liquidity and acceleration effects and the availability of foreign exchange may be important determinants of investment behavior. Indeed, the available empirical evidence for developing countries indicates that these latter factors are of great importance (for example, Billsborrow (1977), Sundararajan and Thakur (1980), and Tun Wai and Wong (1982)). Accordingly, in the empirical work presented below, some of these nonneoclassical elements are accommodated. This permits an evaluation of the relative importance of a variety of factors in the determination of capital formation in developing countries. Since there are a number of potential influences on investment, care is taken to ensure that the countries studied below are chosen with a view to covering a broad range of developing country experience.

This chapter first presents the theory behind the cost-of-capital approach. Subsequent sections consider the relevance of this approach for developing countries, show how the cost of capital in these countries may be sensitive to changes in anticipated inflation, discuss the estimates of the cost of capital for the selected sample of countries, and present cross-section regression results on the determination of investment levels in developing countries.

I. The Theoretical Framework

Background to Cost-of-Capital Calculations

We begin with some comments that concentrate on the salient features of all cost-of-capital calculations.

The most obvious tax influencing the cost of capital facing corporations is the corporate income tax. The manner in which this tax exercises its influence is complicated. Corporate taxes are levied on corporate income after the allowable costs incurred in the process of generating revenue have been deducted. These costs are relatively easy to calculate as far as outlays on labor and raw materials are concerned. They are much more difficult to calculate, however, for capital inputs, because capital is a durable good whose rate of depreciation is difficult to gauge.

In much of the earlier theoretical work it was assumed that corporate taxes were, by and large, borne by capital.1 This assumption is accurate if the deduction accorded capital is inadequate. Proceeding from the assumption that capital bears corporate taxes, authors then analyzed the general equilibrium implications of such taxes. This work, exemplified by Harberger (1962), was very important not only because of its specific conclusions concerning the impact of the corporate income tax but also because of its demonstration that general equilibrium effects of taxes could and should be documented. (For a review of the original Harberger model and more recent extensions, see McLure (1975) and Ballentine and McLure(1980).)

Recent literature, which concentrates less on the general equilibrium effects of corporate taxes and more on the issue of how the corporate income tax actually affects the cost of capital, is more relevant for our purposes.2 Stiglitz (1973) points out that the effects of a statutory corporate income tax rate on the cost of capital cannot be viewed in isolation from other elements of the tax system such as specific provisions of the corporate tax code and the coexistence of the corporate tax with a personal income tax structure. In general, a tax system which, between its interest deductibility provisions and its depreciation allowances permits a deduction whose present value is equal to the cost of a capital investment, is equivalent to a tax system with an immediate write-off provision. This, in turn, implies that such a tax system would be equivalent to a pure profits tax in the sense that, within the context of the model, no marginal conditions are affected. The ability to accommodate the full range of possible tax structures suggested by this literature is what makes the cost-of-capital formula specified below a useful device.

For institutional reasons, the effects of corporate income taxes are frequently complicated by inflation. There is a tendency in many tax systems to tax the nominal rather than the real income of capital. As a result, the tax system interacts with changes in the expected rate of inflation. For corporate taxes, this interaction is frequently the product of some combination of the following factors: the taxation of nominal capital gains and nominal interest income, the use of depreciation allowances based on historic cost, the full deductibility of nominal interest expense, and the requirement that inventory valuations be determined on a first in, first out (FIFO) basis. In a general equilibrium context, this problem is compounded by the fact that tax systems rarely treat all capital assets in the same manner. For example, owner-occupied and rental housing frequently receive a favorable tax treatment. In the case of the former, not only is there no taxation of the implicit rental income that accrues to the owner-occupier but also some amount of mortgage interest payments and property tax payments can sometimes be deducted against other income tax liabilities. This implies that one of the effects of an increase in the rate of inflation is for investors to shift their demand from corporate equity to owner-occupied housing. Thus, while changes in the expected rate of inflation might not be expected to result in large changes in total investment, there could be significant changes in the composition of that aggregate.

Any country’s corporate tax system could be modified to accommodate the effects of inflation. A common solution is to allow some form of accelerated depreciation allowances. Although such allowances will counter the effect of a given expected rate of inflation, the system nonetheless remains sensitive to changes in inflation. These and other aspects of this general problem are discussed in Feldstein (1980a, 1980b, and 1980c), and Ebrill and Possen (1982a and 1982b). Again, the cost-of-capital formula derived below is general enough to permit a consideration of these issues.

When empirically estimating the incentive effects of taxation on business investment, deriving a cost-of-capital estimate is by no means the only possible route to take. A more traditional approach takes as its starting point the actual taxes paid as a proportion of income. The presumption is that marginal tax rates in a given industry are not far from the ratio of actual taxes to capital income in that industry. This may not be too likely given that actual tax revenues may be the product of inframarginal supernormal profits due to the existence of noncompetitive market structures. Examples of well-known articles that rely on this approach are Harberger (1966), Shoven (1976), Fullerton and others (1981), and Fullerton and Gordon (1983).

The use of cost-of-capital calculations is actually quite new (King and Fullerton (1984)), Based on the neoclassical framework developed by Hall and Jorgenson (1971), they have come to be used with increased regularity (e.g., Hall (1981), Jorgenson and Sullivan (1981)). The underlying methodology of the approach used below is to consider a “hypothetical project” of a dollar invested in a particular asset to be used by a representative firm. The firm is assumed to be maximizing its value to the shareholders. Given the specifics of the tax code, the cost-of-capital formula is then derived permitting an evaluation of the impact of various tax provisions at both the corporate and the shareholder level on the actual cost of capital and, hence, on the incentive to invest.

Relevance of the Cost-of-Capital Approach to Developing Countries

The cost-of-capital calculation is concerned with the investment climate facing corporations. That its usefulness is limited by the fact that the corporate sectors in developing countries are often small is not necessarily true. First, the corporate sector, although small, could well be important at the margin (Bhagwati (1978)). Second, and more important, the cost of capital facing the corporate sector may serve as an excellent proxy for the cost of capital facing unincorporated businesses. It is sufficient for both sectors to determine their tax bases in a similar fashion. For example, in countries where historic cost depreciation is the convention, that convention will usually apply to both corporations and unincorporated businesses, and, further, will be reflected in the cost-of-capital calculation.3

A related issue is that the cost-of-capital calculation presumes the existence of competitive capital markets. The reality in many developing countries is much different. Capital markets are commonly either nonexistent or fragmented. This does not, in and of itself, negate the relevance of the cost-of-capital calculation. Even in the absence of well-developed markets, the distinction between debt and equity continues to be maintained both in the practice of financing marginal investments and in the specifics of many developing countries’ tax codes. Interest payments, whether on a bank loan or a bond, are often deductible against corporate tax liabilities while retained earnings are not. Further, capital gains, dividends, and interest payments frequently receive separate treatments at the personal income tax level. However, this reservation does suggest that liquidity variables may have an important role to play in determining investment flows.

Difficulties in the Use of Cost-of-Capital Calculations

The cost-of-capital approach is not without its problems. First, as pointed out by Bradford and Fullerton (1981), since the method is concerned with expected future tax liabilities (statutory corporate and personal income tax rates are normally assumed to remain constant), the choice of a discount rate is critical. Second, effective tax rate estimates are a function of the assumptions made concerning how anticipated inflation affects nominal interest rates. We assume below that allgross nominal returns rise with the rate of anticipated inflation.4

To these caveats raised by Bradford and Fullerton (1981), the most serious of which is the second, might be added a few further considerations. The value of depreciation allowances to the representative firm is gauged by the discrepancy, positive and/or negative, that exists over time between the depreciation allowances and economic depreciation, all expressed in present value terms. It is traditional to assume that economic depreciation is exponential, though there is evidence to suggest that this is not so (Feldstein and Rothschild (1974)). To the extent that economic depreciation is not exponential, one is valuing depreciation allowances against the wrong benchmark. Nonetheless, we will continue to assume that economic depreciation follows a path of exponential decay since alternative assumptions have their own difficulties.

Finally, a number of assumptions are needed to ensure that the cost-of-capital approach can be viewed as providing a general equilibrium framework in which a country’s tax system can be evaluated. Implicitly, authors seem to assume a closed economy framework with fixed aggregate stocks of capital and labor. This allows one to take many variables, such as the underlying marginal productivity of capital, as constant. One can also ignore the effects of many other taxes. For example, general taxes on gross output may then be consistently presumed to be borne by capital and labor in the aggregate in line with their factor shares in gross domestic product (GDP) and, accordingly, such taxes do not affect the allocation of investment funds between the corporate sector and other sectors of any given economy. This permits one to concentrate on the implications of the interactions between the personal and corporate tax systems of developing countries. However, the assumptions necessary for this general equilibrium interpretation to hold may not be realistic. The economies of many developing countries are quite open. When this is combined with the common adoption of overvalued exchange rates and a relatively heavy reliance on trade taxes, the concomitant disequilibrium in the traded goods market can influence investment flows. Thus, as pointed out in the introduction, additional factors such as foreign exchange availability should be accommodated.

It is clear, then, that cost-of-capital calculations should be interpreted with great care. However, they continue to have the great advantage over alternative techniques of being based on a theoretical rather than on an ad hoc foundation. Even if the absolute values of the calculations are subject to error, they will still be useful if the rankings of tax effects across countries are accurate, thereby affording some insight into the sensitivity of investment to changes in the level of the cost of capital. Further, the technique has the advantage of allowing calculation to be made on the basis of statutory tax rates, where these are generally accessible. Given data availability in developing countries, this advantage is particularly valuable.

Derivation of the Cost-of-Capital Formula

Before discussing the derivation of the cost-of-capital formula (equation (4), below), we introduce the following summary of the notation that will be used throughout.

π = the anticipated rate of inflation;

i = nominal rate of interest;

δ = rate of economic depreciation;

s = real interest rate, net of tax;

n = share of marginal investment financed by new issues;

h = share of marginal investment financed by borrowing;

τ = corporate tax rate;

τ* = corporate tax rate corrected for tax holidays;

MG = effective tax rate on capital gains;

MD = tax rate on distributions;

M = personal income tax rate;

a = exponential rate, historic cost, depreciation allowances;

G = share of investment qualifying for initial investment allowance;

ρ′ = gross real cost of capital;

ρ = net real cost of capital;

ρ* = net real cost of capital of “pioneer” firms;

D(t) = nominal income stream in period t from an additional $1 of investment; and

V(t) = stock market value of a share of the “representative” firm in period t.

As the procedure for deriving the actual formula for the cost of capital is well known, it will only be summarized here. See Agell (1983) for a more detailed discussion of the mechanics since his basic formula, with some modifications, will be used below.

The cost-of-capital calculation is derived from a model of the marginal investment decision facing a corporate firm. The marginal investment, financed in period 0 at least in part by retained earnings, yields a dividend stream D(t) where

Thus, dividends net of corporate taxes are calculated by deducting payments associated with debt and the issue of new shares from the gross yield of a marginal investment and adding the tax value of depreciation allowances. To elaborate, the term in (δ-π)h reflects the fact that that portion of the investment which is debt-financed is amortized at a rate which reflects the rate of economic depreciation,δ, adjusted for the anticipated rate of inflation, π. Following Agell (1983), new share issues can be treated without loss of generality as a special debt instrument. These new shares must yield for their shareholders a competitive net of personal income tax return, where this is gauged by(s +π)n/(1 - MD), and, as with regular debt, their share in the marginal investment is amortized at a rate measured by(δ-π)n..

This dividend stream is reflected in the value of shares, V(0), as follows:

where this expression is a solution to the differential equation resulting from the equilibrium condition that the rate of return required by equity holders equals the sum of net of tax dividend payments and capital gains.

The firm is assumed to be maximizing its value to its existing equity holders. Given that assumption, the capital market will be in equilibrium if, at the margin, the equity holders are indifferent as to whether or not the marginal investment is undertaken; that is, equity holders will be indifferent between having the firm pay out dividends and having it make the marginal investment. This equilibrium condition can be expressed as follows:

that is, the forgone dividend payment, net of personal income tax, must equal the capital gain, net of capital gains tax, associated with the investment. Substituting equations (1) and (2) into (3) yields

where the major differences between this expression and Agell’s corresponding equation (5) are (1) the fact that interest deductibility of debt against corporate tax liabilities is permitted by all the countries selected below is already incorporated in the formula and (2) it is assumed from the outset that no relative price changes take place. This last assumption implies that the anticipated rate of general inflation is equal to the expected rate of change in the price level of capital goods.

Having introduced the cost-of-capital calculation and having discussed the caveats relevant to its use, we now proceed to its estimation and evaluation for a sample of developing countries.

II. Empirical Results

The sample used in this chapter consists of 31 countries chosen with a view to covering a broad range of developing country experience.

Derivation of Cost-of-Capital Estimates

The parameters in the cost-of-capital formula given above are of three different types, namely,

(1) macrovariables (π, i, s)which are exogenous to the representative firm;

(2) microvariables (δ,n,h), which are specific to the firm; and

(3) tax variables (τ, τ*, MG, MD, M, a, G)

Consider the specific parameter values for each of these categories in turn, noting that the values for the first two categories are chosen not only because they are reasonable in their own right but also because they are consistent with Agell’s (1983) results.

The nominal rate of interest is presumed to rise pari passu with the rate of inflation (the Fisher assumption) so that

whereir is the real rate of interest.5 The real rate of interest is set at 2 percent a year. Further, the shareholder’s required real rate of return,s, is also equal to the real rate of interest, implying s = 2 percent.

Turning to a consideration of the microvariables, it is assumed that the rate of economic depreciation may be set equal to 0.1225. This estimate is the same as that derived by Hulten and Wykoff (1981) for “general industrial equipment” using U.S. data. Data availability precludes obtaining accurate estimates of n, the share of new financing in the marginal investment, and h, the analogous share for debt. Accordingly, in line with Agell’s (1983) benchmark case, n = 0.10 and h = 0.30 are the chosen values.

Finally, a few comments on the most important tax parameters are in order. The corporate tax rate corrected for tax holidays, τ*, is based on the most generous tax holiday available. There is no guarantee that these terms are available to all start-up firms. The method for correcting the statutory tax rate for tax holidays is presented in Agell (1983). Some of the cases provide for partial tax holidays. The determination of the corrected corporate tax rates for two of these cases is presented in Annex I of this chapter.

The capital gains tax rates are effective tax rates—the statutory tax rates have been corrected for the fact that capital gains are levied on a realization rather than an accrual basis. Account is also taken of the fact that some countries have separate deductions against capital gains liabilities.

As in Agell (1983). the personal income tax rate on dividends is based on the marginal tax rate of the typical investor where the investor is assumed to have a taxable income equal to ten times the per capita 1981 GDP of the country under consideration. Finally, where necessary, the straight-line depreciation allowance formulas have been converted to an equivalent declining-balance formula characterized by the value of the parameter a.6

Consideration of Estimates for Sample

In his study, Agell (1983) concluded that the tax structures of the Association of South East Asian Nations (ASEAN) countries were not a major source of investment disincentives. Does this conclusion hold true for the more extended sample employed here? An answer to this question can be found in Table 1 in which are presented, for the complete sample of 31 countries, estimates of the cost of capital,ρ at a zero rate of inflation; the anticipated rate of inflation,π; and the cost of capital both for ongoing firms,ρ and pioneer firms,ρ*, at these anticipated rates of inflation. The cost-of-capital estimates are derived along the lines laid out in the previous section.7 The expected rate of inflation is estimated by taking an annual average of the actual rate of inflation experienced by the countries over the period 1970-80. This is an admittedly crude measure, particularly so given the possibility that price controls may have been in force in some of the countries.

Table 1.Selected Developing Countries: Cost of Capital at Zero and Actual Rates of Inflation(In percent)
ρ atρ* at
ρ at ZeroAnticipatedAnticipatedAnticipated
Rate ofInflationInflationInflation
InflationRate πRateRate
Argentina0.46130.8-5.857.59
Brazil2.5036.72.802.80
Chile0.80185.6-14.44-14.44
Colombia2.8422.06.936.93
Côte d’Ivoire-3.3013.2-2.130.98
Ecuador2.4714.4-5.91-5.91
Egypt1.2311.52.182.18
Ghana-2.1534.819.9019.29
India3.598.53.342.53
Indonesia2.4020.5-1.802.60
Israel3.1539.78.408.40
Jamaica1.5317.05.394.23
Kenya1.2811.03.363.36
Korea2.9119.87.826.60
Malaysia1.507.5-0.300.85
Mexico0.6919.31.651.65
Morocco2.798.16.293.81
Nigeria1.7218.26.746.74
Pakistan1.9213.54.534.53
Papua New Guinea1.298.84.114.11
Paraguay2.8712.43.433.43
Philippines1.9013.20.880.98
Senegal2.757.62.962.40
Singapore-2.905.1-3.201.00
Sudan6.0015.814.335.97
Syrian Arab Rep.3.9311.48.078.07
Thailand2.009.91.601.11
Trinidad and
Tobago0.8618.51.362.25
Tunisia2.787.75.095.09
Venezuela2.8912.15.075.07
ZaÏre2.7432.24.253.67
Note:π the anticipated inflation rate, is assumed to equal the average annual rale of inflation as measured by the implicit gross domestic prnduct deflator for the period 1470-80.Sources: The data for π were obtained from World Bank.World Development Report. 1982(New York: Oxford University Press. 1982). andWorld Development Report.1983 (New York: Oxford University Press, 1983). Other data sources are described in the sections on “Derivation of Cost-of-Capilal Formula” and “Consideration of Estimates for Sample.”
Note:π the anticipated inflation rate, is assumed to equal the average annual rale of inflation as measured by the implicit gross domestic prnduct deflator for the period 1470-80.Sources: The data for π were obtained from World Bank.World Development Report. 1982(New York: Oxford University Press. 1982). andWorld Development Report.1983 (New York: Oxford University Press, 1983). Other data sources are described in the sections on “Derivation of Cost-of-Capilal Formula” and “Consideration of Estimates for Sample.”

The conclusion must be that, contrary to Agell, the actual cost of capital facing some countries in this sample is such that the interaction between their respective tax systems and the rate of inflation is the source of investment disincentives—any value greater than 2 percent implies a positive tax wedge. For several countries, the disincentive effects are pronounced.

A comparison of the cost of capital calculated at the anticipated rates of inflation with the corresponding cost of capital at a zero rate of inflation shows that the estimates are sensitive to changes in inflation. The latter set of estimates also demonstrate that, while many countries may have a high nominal rate of corporate income tax, the deductions are such as to leave the cost of capital relatively unaffected at low rates of inflation. Indeed, the large number of values less than two imply that the tax system in many countries subsidizes the cost of investment at low inflation rates. The generous deductions implied by this outcome may have been a response to the higher actual rates of inflation experienced by these countries.

Consider the final column of Table 1 , which presents estimates of the cost of capital for start-up or “pioneer” investments. An interesting feature is that for some countries, tax holidays appear to increase the cost of capital. This is not as strange as it first appears. First, if the tax system is such that, in the absence of tax holidays, the tax wedge is negative, then a tax holiday tends to remove this implicit subsidy. This effect can also be observed in Agell’s work. Second, the effective tax rate on corporate capital is the product of an interaction between the corporate and personal income taxes. In some cases, one observes that taxes at the corporate level tend to be very low, even negative, due to accelerated depreciation allowances, while the treatment of capital income at the personal level involves high tax rates. As a result, a tax holiday at the corporate level implies an increase in the tax burden on corporate capital.

However, while the calculations imply that there are circumstances under which a tax holiday can increase the tax burden, a couple of reservations should be noted. First, as mentioned in an earlier section, the cost-of-capital calculation is set in a world where the representative firm is presumed to know its future tax liabilities with certainty. In the real world, notwithstanding the fact that in present value terms it implies an increase in taxes, a firm might well opt for the certainty of a tax benefit now. Second, and more important, the cost-of-capital formula is symmetric—if the incentives are sufficiently great, it requires the government to pay the firm. In reality, most tax systems do not make such payments. This reduces the potential mentioned above for a tax holiday to result in an increase in the cost of capital.

Even with these reservations, the results in Table 1 are still of interest. They imply that tax holidays do not have a major impact on the cost of capital. This surprising result appears to be due to the fact that most corporate taxes have generous allowances built into them. Another way of expressing this is that the cost of capital is not very responsive to changes in the corporate tax rate. A similar lack of sensitivity occurs when other tax parameters are altered. Where sensitivity does exist it is in the specification of the special provisions. In particular, the interaction between inflation and the historic cost depreciation and investment allowances is of critical importance. Indeed, the only possibility for a negative cost of capital arises when depreciation and investment allowances are so generous as to provide a subsidy. In contrast, reduction of the tax parameters merely reduces the degree to which corporate capital is taxed. It is fortunate that the source of the sensitivity lies in these allowances since their value can be determined with a fair degree of accuracy for any given class of capital equipment.

Regression Results

In this section, regression results based on the cross-section of countries in the sample are presented so as to cast some light on the quantitative impact of the cost of capital on investment flows in the sample of developing countries. Given obvious data inadequacies and difficulties in quantifying nontax determinants of investment, too much weight should not be assigned to these results. Since the results are based on cross-section data drawn from a range of countries rather than on time-series data for individual countries, care must be taken in interpreting the coefficients. It is assumed here that the observations can be interpreted as reflecting the behavior of a representative country; that is, were they all to face the same economic circumstances (e.g., the same cost of capital), every country would exhibit exactly the same response (e.g., the same level of investment). Accordingly, all differences in investment behavior across the sample of countries may be ascribed to differences in their economic circumstances, and the coefficients are, therefore, properly interpreted as gauging long-term effects.

The share of gross domestic investment[GDI) in GDP for 1980, designated as GDI/GDP, was selected as a dependent variable, a choice dictated by data availability. Unfortunately, the variable includes a number of extraneous components. For example, it includes government investment. This component is unlikely to be responsive to market-determined rates of return. It may be possible to assume that it does not vary systematically across the sample. On the basis of the available empirical evidence, this may not be an overly strong assumption. For example, Sundararajan and Thakur (1980) find evidence of some short-term crowding out between public and private investment for the case of India, while for Korea they find these two components to be strong complements. Tun Wai and Wong (1982), testing data drawn from a number of countries, find evidence of financial crowding out for some (e.g., Malaysia and Mexico) but not to the same degree for others.

In an attempt to circumvent this potential problem, estimates of central government fixed capital formation8 were chosen as a proxy for government investment and were netted out ofGDI. The resultant alternative dependent variable is referred to asGDIP/GDP.9

The dependent variable also includes noncorporate private investment and residential housing. It was argued above that the former poses no serious problems as long as the tax bases for the corporate and noncorporate sectors are similarly defined in the sample countries. The latter, however, could weaken the link between the dependent variable and the cost of capital. Thus, to the extent that the increase in the cost of capital is induced by increases in the expected rate of inflation, investors will have an incentive to substitute assets such as owner-occupied housing for corporate assets. Accordingly, if the cost of capital is an insignificant variable in the regressions below, it may be due to this effect. As a final comment on the specification of this variable, it should be noted that the conventional theoretical justification for employing it—rather than a measure of the private capital stock—comes from the presumed existence of adjustment costs in attaining a desired level of the capital stock (Lucas (1967), Gould (1968), Treadway (1969)).

Turning to a consideration of the independent variables, two alternative estimates of the cost of capital can be used. The first of these, designated ρ gives estimates of the cost of capital faced by ongoing firms and is presented in Table 1, above. The second, for which no explicit estimates are presented below, substitutes the cost of capital faced by “pioneer” firms where such provisions exist.

As mentioned at the beginning of this chapter, other variables could influence the level of investment activity in developing countries. A few proxies for these variables are also included as additional regressors. The annual average real growth rate of exports over the period 1970-80, designated as Δ is included as a proxy for foreign exchange availability—the more rapid the growth rate, the more likely will funds be available for investment. Note that, to the extent that this is the correct interpretation of the role of this variable, it is measuring a supply rather than a demand effect. Accordingly, it must be argued that the single equation regression results presented below represent a reduced form. Note further that this variable can also be interpreted as an accelerator variable. A more accurate proxy for accelerator effects is provided by the annual average growth rate inGDP (ΔGDP).

As mentioned above, the cost-of-capital estimates presume that nominal interest rates rise in line with inflation. As pointed out earlier, this is not always so. The rigidity of nominal interest rates in reality, which can be ignored if inflation rates are not too high, may well be a problem for some of the countries in the sample, notably, Argentina and Chile, both of which experienced annual inflation rates well in excess of 100 percent. The large and negative cost-of-capital estimates recorded for both of these countries—suggesting a positive correlation between inflation and investment—could be very misleading. Indeed, there may be an independent negative correlation between inflation and investment. As McKinnon (1973) observes, the capital market in these countries may be in disequilibrium. As a result of pursuing low interest rate policies, investment funds are rationed with realized investment being less than desired. As a practical matter, this financial repression tends to be more severe the higher the rate of inflation (Galbis (1979)). Accordingly, this suggests that the anticipated rate of inflation,π may act as a useful proxy for the presence or absence of financial disintermediation and, thus, for the availability of in-vestable funds.

On the matter of the available supply of investable funds, there is a frequently expressed concern that the government sector might “crowd out” private investors. To accommodate this possibility, an additional regressor is defined, namely, the current account balance (without grants), designated hereafter asCAB. This specification of a crowding out variable was chosen rather than some measure of an overall surplus or deficit on the presumption that borrowing on the capital account is productively invested, where this investment, as suggested above, is assumed to be neither a substitute nor a complement for private investment.

Since the share of gross domestic investment inGDP can be expected to be influenced by the level of economic development, the 1980 per capita gross national product,Y, of the sample of countries is included as an independent variable. The final independent variable is the share of fuels, minerals, and metals in merchandise exports(Min). This variable is a proxy for the natural resource endowments of the countries under consideration where the relative abundance of these endowments might be expected to influence the investment climate, particularly from the point of view of foreign investors.

We turn now to consider a selection of the more important regression results.



The regression is run in linear form. The t-ratios are presented in parentheses below the relevant coefficient. R¯2

refers to the adjustedR squared.F to theF ratio, andN to the number of observations.

In this regression, the signs of most of the variables are as expected. The accelerator variable, ΔGDP, is positive and significant. Further,Y, per capita gross national product has a positive and significant coefficient implying that one cannot reject the plausible hypothesis that the more developed the country, the greater the share ofGDP devoted toGDI. The sign of the coefficient of the share of natural resources in exports is not significant given the value of the t-ratio for that variable.

The coefficient of the cost-of-capital variable is negative as predicted, and significant. It is not possible to reject the hypothesis that the interaction between the tax systems of these countries and their anticipated rates of inflation can influence the level of investment. The elasticity value, calculated at the means of the variables, is—0.044, Such a value implies, for example, that if the cost of capital were increased from 3.0 to 4.0, a 33 percent increase, the share ofGDI inGDP would fall by 1.5 percent—a substantial effect.10

In view of the potential for the existence of financial repression in some of the countries in the sample, consider the outcome if π, the anticipated rate of inflation, is introduced as a regressor. Specifically, consider:



It appears from this equation that inflation has a significant and strong negative impact on the level of investment (the value of the elasticity is—0.039, which is large given the potential for large changes in inflation).

Note, further, that the cost-of-capital variable is still significant but that its coefficient has increased significantly in magnitude, suggesting that any elasticity estimates should be treated with caution.11

Consider the outcome ifGDIP/GDP, the share of gross investment net of gross governmental fixed capital formation inGDP, is the dependent variable.12 The analogous regression equation to equation (7) above is



It is clear that, at least for this set of regressors, the change in the dependent variable makes little difference.

Both of the above regression equations are potentially affected by simultaneous equation bias. Thus, it could be argued that the greater the share of GDI inGDP, the more rapid the growth rate of GDP. To alleviate the statistical problems posed by this chain,ΔGDP can be replaced with ΔX where the latter variable, as was pointed out above, is a proxy for both accelerator effects and foreign exchange constraints. Also, to test for the possibility that governments may be crowding out private investors,CAB, the governmental current account balance of the countries in the sample, is also introduced into the regression equation. The results corresponding to equations (7) and (8) are, respectively:





It is clear that this measure of crowding out has no effect on the results. Again, the cost of capital and the inflation variables both continue to be significant with the coefficient of the latter having increased substantially.

All of the above indicates that the cost of capital may have a significant impact on investment levels in developing countries. However, it should also be noted that the other variables, and, in particular,ΔX and π, both of which are proxies for a number of important nonneoclassical effects, are also significant. This raises the question of which is the most promising path for policy to take. Further, there is also the possibility that the results are sensitive to the choice of countries in the sample. As a partial test of this possibility, the regressions were rerun with Argentina and Chile excluded. These two countries were chosen on account of their high rates of inflation and the concomitant possibility that the resulting financial disin-termediation has not been adequately accounted for by the inclusion of π as a regressor. The analogous regression equations to (7) and (8) are





It is clear that some of the coefficient estimates are volatile. In particular, the cost-of-capital coefficient is reduced in magnitude and is no longer as significant. Further, the magnitude of the impact of inflation on investment levels has increased—the elasticity of the effect in equation (11) is 0.25. A potential reason for this volatility can be seen in the correlation matrix where the correlation coefficient between π and ρ is now positive at 0.43. This not only reinforces the impression that the observations on Argentina and Chile are extreme outliers but also suggests that there is still some multicollinearity between the regressors. This multicollinearity might be responsible for the reduction in the significance of the cost-of-capital variable, though, as pointed out above, the reason could also lie in the fact that the dependent variable includes investments made in noncorporate assets such as owner-occupied housing. Be that as it may, these results should temper any strong policy conclusions concerning the effectiveness of tax reform proposals. Indeed, they suggest, if anything, that more attention should be paid to the role of inflation since increases in anticipated inflation appear to affect adversely the level of investment via a number of channels; that is, not only do increases in anticipated inflation interact with the tax system to increase the cost of capital but they may also directly affect the level of investment by inducing financial disintermediation. In addition, given the significance of ΔX, the growth rate of exports, the results also suggest that policymakers should concentrate on alleviating distortions which might result in foreign exchange shortages.

Finally, it should be noted that the regressions were rerun, replacing ρ with ρ*. The results, not presented here, suggested a weaker link between the cost of capital and investment levels. This is not surprising given some of the assumptions underlying the specification of ρ* (see p. 126, above).

III. Conclusions

The tentative nature of the results must be emphasized. The regressions are based on cross-sectional data, whereas tax reform proposals for any given country should more properly be based on empirical work (time series) drawn from that country. However, even given these reservations, the results are suggestive. Thus, it could well be the case that to concentrate on taxes as the primary source of economic inefficiency in a developing country’s economic system is too narrow. The regression results indicate that investment levels are influenced by a number of other factors and, in particular, by the rate of inflation and the availability of foreign exchange as gauged by ΔX. This lends support to the approach adopted in World Bank’sWorld Development Report. 1983, where it is argued that if one takes a broad definition of price distortions, then one observes a strong negative correlation between this index and the economic performance of developing countries. Further, if concern should properly be directed at this broader measure, then, as pointed out by Tanzi (1982), given the nature of the political process in many developing countries, radical reform may not be easy to implement; that is, not only are the simplistic tax reform proposals of the pure supply-side approach inadequate but the appropriate reform proposals may not be politically feasible. For example, given the underdeveloped institutional framework in place in many developing countries, the admonition that they reduce their rates of inflation sharply may be extremely difficult to implement.

Even though this all suggests that supply-side based tax reform proposals are only of limited use, this is not to say that they should be ignored. The results do suggest that the provisions of developing country tax structures may well interact with their respective personal tax systems and changes in anticipated inflation to influence investment activity. This implies that the effective tax on capital may be subject to arbitrary changes, which is hardly desirable. Therefore, given the particular sensitivity of the cost-of-capital calculations to the specification of the depreciation allowances, an obvious tax reform proposal would be to recommend the indexation of such allowances. An alternative strategy, which has the merits of being administratively more simple, would be to disallow the deductibility of interest payments on debt and to permit the expensing of all investment. The resulting corporate tax structure would define its tax base in terms of a company’s cash flow and would be neutral as far as changes in inflation are concerned. (Of course, such a cash flow tax would require some modification to assist start-up companies, etc., which are likely to run losses at the outset.) At a minimum, the authorities of those countries in which indexation is not allowed should be aware of the interactions that recur between tax systems and inflation.

ANNEX I
Examples of Calculation of Effective Corporate Taxation Rates for Economies with Partial Tax Holidays

India

An investment in period 0 yields a nominal income stream which depreciates over time at rate

India applies a statutory rate, τ, to the income stream so that 25 percent of profits are exempt for eight years, yielding a present value of expected tax payments equal to

where nominal tax payments are discounted by the required nominal rate of return.

The effective tax rate,τ*, is that tax rate which, if applied to the same income stream at a uniform rate from period 0, would yield the same present value of expected tax payments. Thus, set

and solve for τ* It follows that

where s + δ = 0.225 and τ is taken to equal 55 percent. Hence,τ* = 46.5 percent.

Korea

Korea grants a complete tax holiday for four years followed by a halving of corporate tax liabilities for a subsequent two years. Following a procedure analogous to that above yields

which implies, for δ + s = 0.225 and τ = 0.396, that τ* = 0.196.

Cost-of-Capital Formula with Indexed Depreciation Allowances

Indexed depreciation allowances imply that the term measuring the tax value of depreciation allowances in equation (1), namely,τae-atreplaced by τ-(a-π)t If the additional assumption of δ = a is made, then substituting the resulting modified equation (1) together with equation (2) into equation (3) results in

This is the expression which is relevant for the case of countries such as Argentina. Note that some countries, notably Israel and Mexico, allow indexation of depreciation allowances but the enabling legislation for this was so recent that it does not enter the cost-of-capital calculations in this chapter.

ANNEX II
Table 2.Selected Developing Countries: Selected Developing Countries: Data for Regression Analysis
GDIGDPΔXMinΔGDPY
Argentina269.322.22.390
Brazil227.5118.42,050
Chile1810.9592.42,150
Colombia251.945.91,180
Côte d’lvoire284.656.51,150
Ecuador257.5468.81,270
Egypt31-0.7478.1580
Ghana5-8.416-0.1420
India233.783.6240
Indonesia228.7697.6430
Israel229.624.14,500
Jamaica16-6.831-1.11,040
Kenya22-1.0216.5420
Korea3123.019.51,520
Malaysia297.4297.81,620
Mexico2813.4395.22,090
Morocco212.1445.6900
Nigeria242.6916.51,010
Pakistan181.274.7300
Papua New Guinea272.0462.3780
Paraguay297.118.61,300
Philippines307.0186.3690
Senegal151.2292.5450
Singapore4312.0278.54,430
Sudan12-5.744.4410
Syrian Arab Rep.256.87410.01,340
Thailand2711.8127.2670
Trinidad and Tobago28-2.8915.14.370
Tunisia284.8527.51,310
Venezuela25-6.7985.03,630
Zaïre112.2560.1220
Note: The variables are defined as follows:GDI/GDP = share of gross domestic investment in gross domestic prirtluet for 1980;ΔX = average annual growth rate of exports over ihe period 1970-80;Min = share of fuels, minerals, and metals in merchandise exports in 1979;ΔGDP = average annual growth rate of gross domestic product over the period 1970-80; andY = per capita gross national product in 1080.In regression equations (8), (10), and (12), the estimates of gross fixed capital formation of central government (the most recent year available), and in regression equations (9) and (10).CAB. Ihe current account balance of consolidated central government as a share ofGDP (data for 1980 or the most recent year available), were obtained from International Monetary Fund,Government Finance Statistics Yearbook.1982 (Washington. 1982) and Fund staff estimates.Sources:GDI/GDP. ΔX. Min, ΔGDP. and Y were obtained from World Bank.World Development Report. 1982 (New York: Oxford University Press, 1982). andWorld Development Report. 1983 (New York: Oxford University Press, 1983).
Note: The variables are defined as follows:GDI/GDP = share of gross domestic investment in gross domestic prirtluet for 1980;ΔX = average annual growth rate of exports over ihe period 1970-80;Min = share of fuels, minerals, and metals in merchandise exports in 1979;ΔGDP = average annual growth rate of gross domestic product over the period 1970-80; andY = per capita gross national product in 1080.In regression equations (8), (10), and (12), the estimates of gross fixed capital formation of central government (the most recent year available), and in regression equations (9) and (10).CAB. Ihe current account balance of consolidated central government as a share ofGDP (data for 1980 or the most recent year available), were obtained from International Monetary Fund,Government Finance Statistics Yearbook.1982 (Washington. 1982) and Fund staff estimates.Sources:GDI/GDP. ΔX. Min, ΔGDP. and Y were obtained from World Bank.World Development Report. 1982 (New York: Oxford University Press, 1982). andWorld Development Report. 1983 (New York: Oxford University Press, 1983).
Table 3.Correlation Matrix
N=31
ρΔXYCABπ
ρ1.0
ΔX-0.451.0
Y-0.230.201.0
CAB-0.350.030.221.0
π-0.510.220.200.021.0
N=29
ρΔXYCABπ
π1.0
ΔX-0.401.0
Y-0.170.171.0
CAB-0.360.010.211.0
π0.43-0.200.17-0.341.0
Note: N = 29 refers to the correlation matrix when Argentina and Chile were excluded.
Note: N = 29 refers to the correlation matrix when Argentina and Chile were excluded.
References

    AgellJonas“Subsidy to Capital Through Tax Incentives in the ASEAN Countries: A n Application of the Cost of Capital Approach Under Inflationary Situations,”Singapore Economic Review (Singapore) Vol. 28 (October1983) pp. 98128.

    BallentineJ. Gregory and Charles E .McLure Jr.“Taxation and Corporate Financial Policy,”Quarterly Journal of Economics (Cambridge, Massachusetts) Vol. 94 (March1980) pp. 35172.

    BhagwatiJagdishForeign Trade Regimes and Economic Development: Anatomy and Consequences of Exchange Control Regimes (Cambridge, Massachusetts: Ballinger Publishing Co .1978).

    BillsborrowRichard E.“The Determinants of Fixed Investment by Manufacturing Firms in a Developing Country,”International Economic Review (Osaka, Japan) Vol. 18 (October1977) pp. 697717.

    BradfordDavid F . and DonFullerton“Pitfalls in the Construction and Use of Effective Tax Rates,”inDepreciation Inflation and the Taxation of Income from Capitaled. by CharlesR. Hulten (Washington: Urban Institute Press1981).

    EbrillLiam P. and Uri M .Possen(1982a) “Inflation and the Taxation of Equity in Corporations and Owner-Occupied Housing,”Journal of Money Credit and Banking (Columbus, Ohio) Vol. 14 (February1982) pp. 3347.

    EbrillLiam P. and Uri M .Possen(1982b) “The Interaction of Taxes and Inflation in a Macroeconomic Model,”Quarterly Journal of Economics (Cambridge, Massachusetts) Vol. 97 (May1982) pp. 23150.

    FeldsteinMartin(1980a) “Inflation, Tax Rules and the Stock Market,”Journal of Monetary Economics (Amsterdam) Vol. 6 (July1980) pp. 30932.

    FeldsteinMartin(1980b) “Inflation, Tax Rules, and the Prices of Land and Gold,”Journal of Public Economics (Amsterdam) Vol. 14 (December1980) pp. 30917.

    FeldsteinMartin(1980c) “Inflation and the Stock Market,”American Economic Review (Nashville, Tennessee) Vol. 70 (December1980) pp. 83947.

    FeldsteinMartin and MichaelRothschild“Towards an Economic Theory of Replacement Investment,”Econometrica (Evanston, Illinois) Vol. 42 (May1974) pp. 393424.

    FeldsteinMartin and LawrenceSummers“Inflation, Tax Rules and the Long-Term Interest Rate,”Brookings Papers on Economic Activity: 1 (1978) Brookings Institution (Washington) pp. 61110.

    FlemmingJ.S.“A Reappraisal of the Corporation Income Tax,”Journal of Public Economics (Amsterdam) Vol. 6 (July-August1976) pp. 16369.

    FullertonDon and Roger H .Gordon“A Reexamination of Tax Distortions in General Equilibrium Models,”inBehavioral Simulation Methods in Tax Policy Analysised. by MartinFeldstein (Chicago: University of Chicago Press1983).

    FullertonDonA.Thomas KingJohn B.Shoven and JohnWhalley“Corporate Tax Integration in the United States: A General Equilibrium Approach,”American Economic Review (Nashville, Tennessee) Vol. 71 (September1981) pp. 67791.

    GalbisVicente“Money, Investment, and Growth in Latin America, 1961-1973,”Economic Development and Cultural Change (Chicago) Vol. 27 (April1979) pp. 42343.

    GouldJ.P.“Adjustment Costs in the Theory of Investment of the Firm,”Review of Economic Studies (Edinburgh) Vol. 35 (January1968) pp. 4755.

    HallRobert E . “Tax Treatment of Depreciation, Capital Gains, and Interest in an Inflationary Economy,”inDepreciation Inflation and the Taxation of Income from Capitaled. byCharlesR.Hulten (Washington: Urban Institute Press1981).

    HallRobert E . and D. W.Jorgenson“Application of the Theory of Optimum Capital Accumulation,”inTax Incentives and Capital Spending,ed. byGary Fromm (Washington: Brookings Institution1971).

    HarbergerArnold C . “The Incidence of the Corporation Income Tax,”Journal of Political Economy (Chicago) Vol. 70 (June1962) pp. 21540.

    HarbergerArnold C . “Efficiency Effects of Taxes on Income from Capital,”inEffects of Corporation Income Taxed. byMarianKrzyzaniak(Detroit: Wayne State University Press1966).

    HultenCharles R . and Frank C .Wykoff“The Measurement of Economic Depreciation,”inDepreciation Inflation and the Taxation of Income from Capitaled. byCharles R.Hulten (Washington: Urban Institute Press, 1981).

    JorgensonDale W. and Martin A .Sullivan“Inflation and Corporate Capital Recovery,”inDepreciation Inflation and the Taxation of Income from Capitaled. byCharles R.Hulten (Washington: Urban Institute Press1981).

    KingMervyn A. “Taxation, Corporate Financial Policy, and the Cost of Capital: A Comment,”Journal of Public Economics (Amsterdam) Vol. 4 (August1975) pp. 27179.

    KingMervyn A. and DonFullertoneds.,The Taxation of Income from Capital: A Comparative Study in the United States the United Kingdom Sweden and West Germany (Chicago: University of Chicago Press1984).

    KrzyzaniakMarian and R. A.MusgraveThe Shifting of the Corporation Income Tax: An Empirical Study of Its Short-Run Effect upon the Rate of Return (Baltimore: Johns Hopkins University Press1963).

    LucasRobert E. Jr.“Adjustment Costs and the Theory of Supply,”Journal of Political Economy (Chicago) Vol. 75 (August1967) Part 1 pp. 32134.

    McKinnonRonald I.Money and Capital in Economic Development (Washington: Brookings Institution1973).

    McLureC. E. Jr.“General Equilibrium Incidence Analysis: The Harberger Model After Ten Years,”Journal of Public Economics (Amsterdam) Vol. 4 (February1975) pp. 12561.

    RaoPotluri and Roger LeRoyMillerApplied Econometrics (Belmont, California: Wadsworth Publishing Co.1971).

    ShovenJohnB.“The Incidence and Efficiency Effects of Taxes on Income from Capital,”Journal of Political Economy (Chicago) Vol. 84 (December1976) pp. 126183.

    StiglitzJoseph E.“Taxation, Corporate Financial Policy, and the Cost of Capital,”Journal of Public Economics (Amsterdam) Vol. 2 (February1973) pp. 134.

    StiglitzJoseph E.“The Corporation Tax,”Journal of Public Economics (Amsterdam) Vol. 5 (April-May1976) pp. 30311.

    SundararajanV. and SubhashThakur“Public Investment, Crowding Out, and Growth: A Dynamic Model Applied to India and Korea,”Staff PapersInternational Monetary Fund (Washington) Vol. 27 (December1980) pp. 81455.

    TanziVito“Is There a Quest for Public Sector Efficiency in Developing Countries?”paper presented at the Thirty-Eighth Congress of the International Institute of Public Finance (Copenhagen,1982mimeographed).

    TreadwayA. B. “On Rational Entrepreneurial Behaviour and the Demand for Investment,”Review of Economic Studies (Edinburgh) Vol. 36 (April1969) pp. 22739.

    Tun WaiU and Chorng-hueyWong“Determinants of Private Investment in Developing Countries,”Journal of Development Studies (London) Vol. 19 (October1982) pp. 1936.

    World BankWorld Development Report 1982 (New York: Oxford University Press1982).

    World BankWorld Development Report 1983 (New York: Oxford University Press1983).

For an exception to this, consider Krzyzaniak and Musgrave(1963).

Some important examples arc Stiglitz (1973) and (1976), King (1975), and Ftemming (1976).

Note that the cost-of-capital calculation can similarly be viewed as a proxy for the cost of capital facing foreign investors. Although no account is taken of the home (developed) countries’ costs of capital, the relative attractiveness of any developing country as a location for foreign investment can be derived from its relative cost-of-capital ranking. A caveat to this is, of course, the fact that there may be variations in the magnitude of nonfinancial restrictions to foreign investment flows across developing countries.

Although this may appear to be a partial equilibrium assumption, it is consistent with the empirical results reported for the United States by Feldstein and Summers (1978). The assumption is suspect when it is applied to those developing countries where interest rales are regulated. Indeed, as shall be seen below, the empirical results appear to be influenced by this effect.

Taken at face value, this assumption would appear to be violated in those countries where interest rates are institutionally fixed at an artificially low rate. However, to the extent that the resultant financial disintcrmcdiation results in the creation of secondary financial markets, such as the curb markets of Korea, this problem is ameliorated—the interest rates in such markets are likely to be responsive to changes in inflation. Further, as shall be seen below, the regression results will be altered in an attempt to accommodate those cases where the assumption is most likely to be seriously in error.

Since some countries index their depreciation allowances, the cost-of-capital formula is somewhat altered. The manner of the alteration is described in Annex I of this chapter.

An important reservation to these estimates should be noted, namely, that they do not incorporate many specific tax incentives such as regional and industry-specific grants. For these and other reasons mentioned above, too much stock should not be placed in the absolute values of the estimates. Rather, it is hoped that the ranking implied by the calculations is accurate. The data sources for the tax variables were the relevant country issues of Price Waterhouse and Company’s Doing Business series and the relevant supplements of the International Bureau of Fiscal Documentation. The relevant tax parameters were those in force in the early 1980s.

Note that this docs not include capital formation by public enterprises which may respond

to changes in the cost of capital and should, therefore, be part of the dependent variable. The data used in the regressions are presented in Annex II of this chapter.

empty

The volatility of the regression coefficients indicates that the independent variables may be multicollincar. Indeed, as can be seen from the correlation matrix presented in Table 3 , in Annex II, the simple correlation between ρ and π is—0.51. which is consistent with the change in the coefficient value of ρ when π is included in the regression (Rao and Miller (1971)), As an aside, note that the negative correlation between ρ and π, a cross-section result, is not inconsistent with the fact that, within any given country, the cost of capital tends to increase with increases in anticipated inflation. As can be seen from Table 3 , the positive correlation is due to the inclusion of Argentina and Chile whose cases, as pointed out on page 129, are extreme.

Depending on the criteria used by public enterprises in their investment decisions, it might have been preferable to net out the gross investment by public companies. Lack of data precluded this possibility.

    Other Resources Citing This Publication