Chapter

III Analysis of Trade Liberalization and Revenue Developments

Author(s):
Reint Gropp, Liam Ebrill, and Janet Stotsky
Published Date:
July 1999
Share
  • ShareShare
Show Summary Details

This section considers the interplay between trade liberalization and revenue from international trade taxes in recent decades. The intent is to review what has actually occurred, in order to gain insight into the factors that govern the revenue consequences of trade liberalization and the related issue of how countries have pursued liberalization.

Trade Liberalization Strategies and Revenue Implications: A Comparative Analysis

A detailed comparison of country experiences is a useful way of examining the interaction among the many dimensions of trade liberalization, their revenue implications, and the economic circumstances of individual countries. The six countries whose histories of trade liberalization are described in this section—Argentina, Malawi, Morocco, the Philippines, Poland, and Senegal—all have considerable experience with trade reform, which has sometimes occurred in phases. They were selected to provide a range of country experience, with a focus on Africa, where some of the issues concerning the revenue impact of trade reform have received particular attention. The selection is not, however, intended to be representative of other countries but merely illustrative.

The discussion of Argentina focuses on that country’s reforms in two distinct periods: from 1987 to 1991, and since 1991. The discussion of Malawi focuses on the late 1980s and mid-1990s. In the case of Morocco, a distinction can be made between the macroeconomic context of the reforms implemented in the 1980s and that of more recent reforms. In the Philippines, trade reform was implemented at a time of weak overall revenue performance in the mid-1980s. The discussion of Poland covers the period since 1990, the beginning of its transformation into a market economy, and thus its trade policy reforms exemplify the case of a transition economy. In Senegal, the reform process can again be decomposed into phases: an initial phase in the mid-1980s and a subsequent phase beginning in 1994.

Trade Liberalization Measures

How have these countries handled each of the many dimensions of trade reform? First, in the area of reform of nontariff barriers, reducing quantitative restrictions was a priority in all cases, although the pace and nature of that reform appear to have reflected revenue concerns, among others. Poland in 1990 placed a strong emphasis on eliminating quantitative restrictions: their removal was integral to the country’s “big bang” transformation. The 1991 trade reform package in Argentina included tariffication of the automobile import quota and the elimination of some reference prices. Both Malawi and Senegal pursued strategies in the late 1980s and the 1990s that stressed bolstering the domestic tax system while moving less rapidly with the outright elimination of quantitative restrictions. Nonetheless, both made progress in reducing such restrictions. Malawi’s trade liberalization program in the late 1980s included a focus on eliminating foreign exchange rationing. Similarly. Senegal embarked in 1986 on a phased reduction of quantitative restrictions, concentrating initially on goods not produced locally. The Philippines focused on the tariffication of quantitative restrictions during the initial phases of its trade liberalization efforts. Finally, reflecting a trade liberalization strategy adopted in 1983. Morocco’s reforms included a gradual elimination of quantitative restrictions on imports and the abolition of import deposit requirements.

Again, all cases stressed tariff reductions in their reform programs, although with hesitation at times. Morocco’s reductions in statutory tariffs resulted in a decline in the collected tariff rate 1 from 26.3 percent in 1980 to 14.9 percent in 1995 (Table 10 in Appendix I). This success owes much to improved stabilization policies, which supported an opening of the economy that led to a strengthened performance of domestic tax revenue. In contrast, Senegal’s early liberalization efforts in the mid-1980s faltered in the face of weaknesses in macroeconomic management and stagnant trade; as a consequence, tariff reductions were accompanied by serious revenue shortfalls, which led to a reversal of the tariff cuts. Senegal’s second phase, implemented in conjunction with the 1994 devaluation of the CFA franc, was more successful. In the Philippines, reduced reliance on trade taxes has at times been constrained by the weakness of domestic tax mobilization. The tariffication of quotas there has already been noted. In addition, the Philippines imposed a temporary import surcharge in the early 1990s. As a result, the collected tariff rate declined only slightly, from 16.7 percent in 1985 to 14.4 percent in 1995 (Table 1 in Appendix I). Poland adopted comprehensive trade liberalization at the outset of its transformation program in 1990, with a view to using international prices as a guide to price formation within the domestic economy. However, in the face of sharp output declines and revenue pressures, some reversals occurred, which have been overcome in recent years as reform of the economy more generally has taken hold. Finally, Malawi imposed a temporary export levy to address revenue shortfalls that compromised its trade liberalization efforts.

Table 1.Collected Tariff Rates by World RegionIn percent of value of imports
Region1975198019851990Last

Available

Year 1
All countries12.6211.3612.2711.059.70
OECD25.814.193.512.881.69
Non-OECD15.6414.2715.8914.3712.91
Africa19.3117.3619.0918.1716.01
Asia and Pacific14.0512.0415.6316.5113.13
Middle East16.4714.3314.0710.7011.39
Western Hemisphere12.3712.6713.7711.0910.26
Sources: IMF, Government Finance Statistics, various issues, and World Economic Outlook (October 1997); and OECD. Revenue Statistics, various issues.Note: Data are unweighted averages.

Last year for which data are available is 1995 for most countries and an earlier year or 1996 for some countries.

Excluding the Czech Republic, Hungary, Luxembourg, and Poland.

Sources: IMF, Government Finance Statistics, various issues, and World Economic Outlook (October 1997); and OECD. Revenue Statistics, various issues.Note: Data are unweighted averages.

Last year for which data are available is 1995 for most countries and an earlier year or 1996 for some countries.

Excluding the Czech Republic, Hungary, Luxembourg, and Poland.

Table 2.Taxes on International Trade by World Region(In percent of GDP)
Region1975198019851990Last

Available

Year1
Total trade taxes
All countries4.234.194.283.373.23
OECD21.200.910.770.600.37
Non-OECD5.305.215.364.394.25
Africa6.676.226.505.285.50
Asia and Pacific3.804.835.264.363.73
Middle East5.014.324.163.493.59
Western Hemisphere4.284.524.494.013.70
Of which:
Import duties
All countries3.253.383.503.082.96
OECD21.110.870.750.580.37
Non-OECD4.004.174.354.053.91
Africa4.985.015.304.974.94
Asia and Pacific2.783.143.783.873.28
Middle East4.344.183.953.283.48
Western Hemisphere3.083.673.703.703.51
Export duties
All countries0.860.700.510.220.17
OECD20.070.020.010.000.00
Non-OECD1.140.910.660.300.23
Africa1.611.141.040.310.33
Asia and Pacific0.711.250.710.490.44
Middle East0.560.090.050.040.05
Western Hemisphere1.000.780.400.310.07
Sources: IMF, Government Finance Statistics, various issues, andWorld Economic Outlook (October 1997), and OECD,Revenue Statistics, various issues.

Last year for which data are available is 1995 for most countries and an earlier year or 1996 for some countries.

Excluding the Czech Republic, Hungary, Luxembourg, and Poland.

Sources: IMF, Government Finance Statistics, various issues, andWorld Economic Outlook (October 1997), and OECD,Revenue Statistics, various issues.

Last year for which data are available is 1995 for most countries and an earlier year or 1996 for some countries.

Excluding the Czech Republic, Hungary, Luxembourg, and Poland.

Table 3.Selected African Countries: Tax Revenue and Taxes on International Trade
1986198719881989199019911992199319941995
(In percent of CDP)
Total tax revenue
Côte d’Ivoire20.3521.5020.4318.0917.5616.8916.9314.7816.4117.84
Ghana12.2012.7012.4012.3010.8012.4010.0012.9016.9015.00
Madagascar9.3210.9610.498.849.436.858.668.167.698.32
Mali11.909.608.509.809.8012.1010.4011.1010.0010.70
Niger9.008.048.138.177.907.006.706.605.406.61
Total taxes on international trade1
Côte d’Ivoire7.307.227.016.845.705.784.964.484.695.09
Ghana2.782.382.433.183.233.223.013.554.354.52
Madagascar2.903.734.563.864.733.073.983.963.504.31
Mali4.473.593.133.663.716.265.556.125.225.77
Niger3.302.702.732.673.012.352.092.231.952.62
Domestic taxes on international trade
Côte d’Ivoire3.133.012.762.662.111.962.001.922.102.25
Ghana0.460.620.911.311.371.361.181.381.751.85
Madagascar1.011.291.641.481.881.271.351.311.582.04
Mali2.011.371.301.581.781.631.431.521.291.69
Niger1.851.401.391.351.491.111.001.040.900.98
Customs duties
Côte d’Ivoire4.174.214.254.183.593.822.952.552.592.84
Ghana2.331.771.521.881.861.871.832.172.602.67
Madagascar1.902.442.922.382.851.792.622.661.932.27
Mali2.462.221.832.081.934.634.114.603.934.09
Niger1.461.301.341.321.521.241.091.191.051.63
(In percent of total taxes on international trade)
Domestic taxes on international trade
Côte d’Ivoire42.9141.6439.3538.8636.9833.9640.4042.9944.7444.21
Ghana16.3525.8837.3741.0342.3742.0539.1138.9440.2840.88
Madagascar34.6934.5135.9338.3639.8041.4733.9932.9845.0347.38
Mali44.9238.2141.6243.1048.0026.0025.8424.8924.7729.25
Niger55.9251.9450.8150.5249.5947.3447.8346.7946.3137.54
Customs duties
Côte d’Ivoire57.0958.3660.6561.1463.0266.0459.6057.0155.2655.79
Ghana83.6574.1262.6358.9757.6357.9560.8961.0659.7259.12
Madagascar65.3165.4964.0761.6460.2058.5366.0167.0254.9752.62
Mali55.0861.7958.3856.9052.0074.0074.1675.1175.2370.75
Niger44.0848.0649.1949.4850.4152.6652.1753.2153.6962.46
Sources: Country authorities; and IMF staff estimates.

Domestic taxes on international trade plus customs duties.

Sources: Country authorities; and IMF staff estimates.

Domestic taxes on international trade plus customs duties.

Table 4.Tariff Revenues and Collected Tariff Rates for the Sample Used to Estimate Model A(In percent)
Measure198019851992
Tariff revenue-GDP ratio3.12.62.7
Collected tariff rate13.715.413.9
Import-GDP ratio23.318.721.9
Source: IMF staff estimates.Note: Data are un weighted averages of the countries in the sample.
Source: IMF staff estimates.Note: Data are un weighted averages of the countries in the sample.
Table 5.Determinants of Import and Trade Tax Revenue, Model A
Dependent Variable1
VariableImport tariff

revenue
Total trade

tax revenue
Constant−0.794**−0.424*
(0.267)(0.254)
Import-GDP ratio0.697**0.249**
(0.091)(0.086)
Export-GDP ratio−0.300**0.151*
(0.095)(0.090)
Per capita GDP0.1200.078
(0.100)(0.096)
VAT dummy0.0710.145**
(0.077)(0.073)
Article VIII dummy−0.086−0.044
(0.115)(0.110)
Real exchange rate index0.382**0.267**
(0.100)(0.096)
Quantitative restrictions0.131**0.128**
reform dummy(0.064)(0.061)
Import tariffs reform dummy−0.0330.062
(0.058)(0.055)
Export restrictions0.021−0.163**
reform dummy(0.075)(0.071)
R20.8500.847
No. of observations344344
Source: IMF staff estimates.Note:** significant at the 5 percent level; * significant at the 10 percent level; standard errors are in parentheses.

The dependent variable is the logarithm of import tariff or total trade tax revenue as a percentage of GDP.

Source: IMF staff estimates.Note:** significant at the 5 percent level; * significant at the 10 percent level; standard errors are in parentheses.

The dependent variable is the logarithm of import tariff or total trade tax revenue as a percentage of GDP.

Table 6.Results of the Instrumental Variable Approach, Model B
First StageSecond Stage
VariableFixed-

effects
Fixed-

effects
Random-

effects
Between

estimator
Collected tariff rate (instrument)0.137**0.178**0.572**
(0.016)(0.015)(0.061)
Square of collected tariff rate(0.016)(0.015)(0.061)
−0.003**−0.004**−0.015**
(instrument)(0.0004)(0.0004)(0.002)
Lagged collected tariff rate0.543**
(0.021)
Average collected0.057**0.038**0.033**0.011
export tax rate(0.021)(0.003)(0.003)(0.010)
VAT dummy−0.779**−0.0008−0.009−0.308*
(0.300)(0.044)(0.043)(0.155)
Article VIII dummy−0.660**−0.0140.0500.554**
(0.357)(0.051)(0.050)(0.170)
Per capita GDP0.588**−0.258**−0.292**−0.233**
(0.289)(0.041)(0.030)(0.057)
Real exchange rate index−0.010**0.001*0.001**−0.003
(0.004)(0.0006)(0.0005)(0.003)
R20.870.470.520.63
No. of observations1,5751,5751,5751,575
Hausman test1460.2**158.0**
Estimated revenue-maximizing23.623.219.6
collected tariff rate2(21.0–26.2)(21.5–24.9)(17.8–21.4)
Source: IMF staff estimates.Note: The dependent variables in the first and second stages are the collected tariff rate and the natural logarithm of trade tax revenue divided by GDP, respectively. The instrument is the lagged collected tariff rate. * * significant at the 5 percent level; * significant at the 10 percent level; standard errors are in parentheses.

Significance indicates a correlation between country-specific effects and the independent variables, implying that a fixed-effects model might be preferable to a random-effects model.

The 95 percent confidence interval is in parentheses. Calculations are based on bootstrapped standard errors.

Source: IMF staff estimates.Note: The dependent variables in the first and second stages are the collected tariff rate and the natural logarithm of trade tax revenue divided by GDP, respectively. The instrument is the lagged collected tariff rate. * * significant at the 5 percent level; * significant at the 10 percent level; standard errors are in parentheses.

Significance indicates a correlation between country-specific effects and the independent variables, implying that a fixed-effects model might be preferable to a random-effects model.

The 95 percent confidence interval is in parentheses. Calculations are based on bootstrapped standard errors.

Table 7.Results of the Regional Instrumental Variable Approach, Model B
RegionVariableSecond Stage Fixed-EffectsEstimated

Revenue-

Maximizing

Collected

Tariff Rate1
OECDCollected tariff rate (instrument)−0.003
x OECD dummy(0.044)
Square of collected tariff rate (instrument)0.011**
x OECD dummy(0.003)
AfricaCollected tariff rate (instrument)0.235**17.8
x Africa dummy(0.023)(16.4–20.2)
Square of collected tariff rate (instrument)−0.007**
x Africa dummy(0.0007)
AsiaCollected tariff rate (instrument)0.158**28.3
x Asia dummy(0.021)(24.3–32.3)
Square of collected tariff rate (instrument)−0.003*
x Asia dummy(0.0005)
Middle EastCollected tariff rate (instrument)0.256**18.3
x Middle East dummy(0033)(16.2–20.4)
Square of collected tariff rate (instrument)−0.007**
x Middle East dummy(0.001)
Non-OECDCollected tariff rate (instrument)0.228**19.0
Western Hemispherex Western Hemisphere dummy(0032)(14.9–23.1)
Square of collected tariff rate (instrument)−0.006**
x Western Hemisphere dummy(0.001)
Collected export tax rate0.038**
(0.003)
VAT dummy0.005
(0.043)
Article VIII dummy0.081
(0.050)
Per capita GDP−0.216**
(0.036)
Real exchange rate index0.002**
(0.0005)
R20.57
No. of observations1.575
Hausman test2227.2**
Source: IMF staff estimates.Note: The dependent variable is the natural logarithm of trade tax revenue scaled by GDP ** Significant at the percent level. Standard errors are in parentheses.

The 95 percent confidence interval is in parentheses. Calculations are based on bootstrapped standard errors.

Significance indicates a correlation between country-specific effects and the independent variables, implying that a fixed-effects model might be preferable to a random-effects model.

Source: IMF staff estimates.Note: The dependent variable is the natural logarithm of trade tax revenue scaled by GDP ** Significant at the percent level. Standard errors are in parentheses.

The 95 percent confidence interval is in parentheses. Calculations are based on bootstrapped standard errors.

Significance indicates a correlation between country-specific effects and the independent variables, implying that a fixed-effects model might be preferable to a random-effects model.

Several of the countries accorded high priority to reducing tariff dispersion and consolidating tariff structures. Between October 1988 and June 1991, Argentina reduced the dispersion of its existing tariff rates, in part through the introduction of a 5 percent minimum tariff in 1989; this reduction was partly offset by the introduction of specific duties on a range of goods. In the late 1980s, Malawi reduced the range of its tariffs from 0–220 percent to 0–45 percent; by August 1997 the maximum tariff rate had been reduced to 35 percent. In its early reform efforts, Morocco considerably simplified and rationalized its complex tariff schedules. The maximum tariff rate was reduced from 400 percent in 1982 to 60 percent in 1984 and 35 percent in 1993; in addition, the number of tariff bands was reduced from 47 in 1980 to 6 in 1996. Senegal also made progress in rationalizing its tariff structure. In the Philippines, however, partly as a result of the tariffication of quotas, measured tariff dispersion was not reduced at the outset, although later both tariff rates and tariff dispersion were reduced.

In Argentina, as part of the 1991 reform package, which included steps to broaden the domestic tax base, export taxes on many agricultural products were eliminated, with the exception of unprocessed oilseeds. In Malawi in 1995, the opportunity to tax the windfall gains associated with a sharp depreciation of the currency and a deteriorating revenue performance led to an export levy being imposed on certain agricultural exports, although this was accompanied by a plan to phase it out. In Morocco, the ad valorem taxes levied on exports were abolished in 1995. leaving only raw phosphate exports subject to a specific tax.

Regional developments have also helped promote trade liberalization in this group of countries Malawi’s trade reform efforts in recent years have been guided by its participation in regional trade initiatives, principally the Cross-Border Initiative (CBI).2 Morocco currently faces the fiscal challenge posed by its 1996 Association Agreement with the European Union (AAEU): Morocco must remove all tariffs on EU industrial goods imports over a 12-year period. One estimate (Abed, 1998) suggests that Morocco’s resulting revenue loss might amount to as much as 2 to 2.6 percent of its GDP by the end of the transition period.3

The countries in the sample also afford examples of how domestic taxes can reinforce the protection provided by trade taxes. In Malawi, until the 1992/93 fiscal year, the surtax (a broad-based sales tax) effectively taxed imported goods that compete with domestic goods at a higher rate than domestic goods, to strengthen protection. This differential was eliminated in that fiscal year, in combination with an upward adjustment in import duty rates on imported goods previously carrying the higher effective surtax rate as an offset. In Poland, excise tax rates on some items were 5 percent higher on imports than on domestic production.

Experience in all the countries demonstrates the importance of broadening the domestic tax base as part of an overall liberalization package. In the late 1980s Malawi embarked on a program to reform the tax system by, among other things, lowering marginal tax rates and converting the sales tax into a credit-based VAT. Trade liberalization in Morocco has been accompanied by improvements in the tax structure, as exemplified by the introduction of a VAT in 1986 and, more recently, by a complete overhaul of the tax and customs administrations. In the Philippines, both direct and indirect taxation were reformed, the latter through the introduction of a VAT; however, there was less progress in bolstering tax administration. In Poland, as in many transition economies, the imperative was to implement a modem tax system to accommodate the rapid changes in the underlying economic structure. However, as Poland discovered, even with rapid progress on tax reform the effective implementation of new taxes and associated administration structures is time-consuming. This experience shows that revenue considerations can impede trade liberalization in the short term. Finally, Senegal has been successfully refining its VAT (reducing the number of rates and expanding the base) to bolster revenue mobilization.

The importance of administrative capacity has not always been recognized. Malawi and Morocco both undertook trade liberalization without first improving customs administration. Only recently have both countries made customs administration reform a central component of trade reform.

Impact on Revenue

Clearly, many of the initial conditions and the trade liberalization strategies that would be consistent with minimal revenue consequences, as discussed above, were present to varying degrees and at different times in each of the six countries. It is interesting to consider how these disparate elements were reflected in actual trends in revenue in each country.

During most of the first phase of reform in Argentina (1988–91), overall tax revenue performance was weak, and revenue from taxes on international trade tended to fluctuate around a broadly unchanged trend. Over this period the dispersion of tariffs and the average tariff rate were only slightly reduced. The indications are that the failure to boost other sources of revenue and to achieve macroeconomic stability compromised trade reform in that phase. However, in the post-1991 period, when the base broadening of the VAT (which was extended to most movable goods) took hold, and inflation was reduced, overall revenue performance strengthened. Tax revenue increased from 12.3 percent of GDP in 1990 to 16.0 percent in 1995 (Table 8 in Appendix I), and there was a reduced reliance on international trade taxes, in relation to both total revenue and GDP.

Trade liberalization and economic reform in Malawi also went through phases. During the late 1980s, complementary tax reforms were revenue enhancing. Moreover, the liberalization of foreign exchange restrictions and a 15 percent devaluation of the currency in 1988 resulted in a modest increase in imports as a percentage of GDP. As a result, from fiscal year 1987 to fiscal year 1990. both the collected trade tax rate and the share of trade taxes in tax revenue declined modestly; in contrast, trade taxes as a percentage of GDP increased modestly (Table 9 in Appendix I), This would appear to be an example where the initial focus on reducing quantitative restrictions and tariff dispersion, together with the favorable impact of liberalizing the foreign exchange market and enhancing tax administration, resulted in both strengthened revenue mobilization and a more liberal trade regime. The more recent phase affords an example of how difficult it can be to liberalize in the face of weak domestic tax performance. Despite reductions in tariff rates in the mid-1990s, the collected trade tax rate tended to increase, in large part because of the imposition of a temporary export levy in 1995. The tendency for the collected import tariff rate to increase in recent years could also reflect improvements in tax and customs administration.

As already noted, reform of the domestic tax system look place alongside a program of trade liberalization during the initial reform phase in Morocco in the 1980s. The country was therefore able to maintain a relatively high and stable tax revenue-GDP ratio even as the collected trade tax rate and reliance on taxes from international trade tended to decline (Table 10 in Appendix 1). More recent years have witnessed a further decline both in the collected trade tax rate and in trade taxes as a share of tax revenue, reflecting, among other things, improved tax administration and the pursuit of sound macroeconomic policies more generally.

The interesting phase of reform in the Philippines began in 1986 against a background of weak overall revenue performance: revenue was only about 10 percent of GNP (Table 11 in Appendix I). Budgetary pressures constrained the reform, which consequently focused on the tariffication of quantitative restrictions and reductions in tariff dispersion, while revenue was bolstered by, among other things, a new surcharge on non-oil imports. Reflecting the liberalization, imports increased significantly as a percentage of GDP. This increase in imports, in combination with the reform strategy that has been pursued, has resulted in an increase in trade taxes as a percentage of GDP and of total tax revenue, even as the collected trade tax rate has declined modestly.

Poland illustrates the difficulties that transition economies experience as they attempt to move to a market-oriented economy. The initial liberalization was dramatic. However, a consequence of economic transformation is that the traditional sources of revenue (e.g., levies on state enterprises) lend to decline sharply even as expenditure pressures remain high. Hence the Polish authorities had to move rapidly to implement the full range of modern taxes with supporting tax administrations. Perhaps inevitably, this process was not always smooth. As a result, the pace of trade liberalization suffered some hesitations, which reflected in part the pressure to protect revenue in the face of a severe economic downturn and in part the fact that it took time for tax reform to contribute significantly to the budget. Between 1991 and 1994 the collected international trade tax rate and the share of trade taxes in total revenue both rose—the latter peaking in 1994 at about 9 percent (Table 12 in Appendix 1). However, the reforms have since taken root, and Poland was able to reduce tariff rates to the point where trade taxes as a percentage of tax revenue and of GDP have fallen, even as the value of trade has soared. By 1996 the share of trade taxes in tax revenue had declined to slightly above 6½ percent.

The early phase of reform in Senegal demonstrates how difficult it can be to pursue trade liberalization in the absence of supporting macroeconomic and structural policies. From fiscal year 1985/86 to fiscal year 1990/91, reliance on trade taxes increased only modestly even though the collected trade tax rate increased by about 48 percent (Table 13 in Appendix I). These trends, combined with the fact that the ratio of trade taxes to GDP also increased only modestly, imply that there was import compression (relative to GDP). This problem was alleviated by the macroeconomic stabilization measures implemented in 1994 in conjunction with the devaluation of the CFA franc.4 Since 1994, import tax receipts have increased, and the revenue implications of significant tariff reductions have been outweighed by higher import levels and temporary tariff surcharges intended to ease the adjustment precipitated by the elimination of import licenses (an action analogous to tariffication of a quantitative restriction).

Summary of Results and Implications for Liberalization Indicators

All of the countries discussed have implemented substantial trade reforms, but their experiences have differed widely. As a prelude to the econometric analysis reviewed below, consider how these experiences are reflected in the behavior of liberalization indicators and in indicators of the countries’ reliance on international trade taxes.

As already noted, a country’s trade policy stance resists easy statistical characterization. In the absence of an ideal measure, this study presents two econometric models with complementary liberalization indicators to gauge the revenue implications of trade liberalization. One of these models uses the collected tariff rate as its indicator of trade liberalization. As is well known, this variable needs to be interpreted carefully for several reasons.

First, because the collected tariff rate is a single summary measure that cannot distinguish between the various forms of trade liberalization, it cannot perfectly track trends in trade liberalization.5 Second, although it is related to the average statutory tariff, the collected tariff rate will be less than the corresponding average statutory rate for several reasons. One reason is that some goods may receive preferential treatment, such as exemptions or regional preferences. Another is that tariffs create an incentive to reclassify goods so as to qualify for lower tariff rates. Yet another is that tariffs create the incentive for outright evasion through smuggling.6 It is worth noting that the correlation coefficient between the collected tariff rate and the simple average statutory tariff rate, an alternative liberalization indicator for which limited observations were available, was about 0.7.7 Third, at some point the collected tariff rate may decline even as the average statutory rate increases, reflecting the impact of increased incentives for evasion and lobbying for exemptions. More generally, under a highly restrictive trading regime, tariff rates might be beyond the revenue-maximizing rates, a point that is considered further in the econometric analysis below.

Against this background, consider how movements in the collected tariff rate have tracked liberalization trends in the six countries. Comparing, for each country, the situation in the year prior to the most important recent trade reform episode with that in the most recent year for which data are available,8 we find that the average collected tariff rate decreased from 17.1 percent to 12.1 percent. This indicates that, for this sample, movements in the collected tariff rate have been tracking underlying liberalization trends (see Tables8 to13 in Appendix I). At the same time, reflecting an increase in the average ratio of imports to GDP from about 16 percent to almost 25 percent, the share of trade taxes in tax revenue rose modestly. Although the increase in imports could reflect independent considerations, the results are consistent with the notion that one of the effects of trade liberalization is an expansion of the importance of trade.

Consider next the behavior of the ratio of international trade taxes to GDP, which was observed to increase in some countries with SAF/ESAF-supported programs. The experience of the country sample demonstrates that interpreting this ratio as an indicator of liberalization can be misleading. Likewise reflecting the import surge following liberalization, the average ratio of import duties to GDP increased from 2.8 percent to 3.6 percent between the pre-reform year and the latest year for which data are available. This is consistent with the notion that the combination of the reform strategy and the positive impact of liberalization (notably, a surge in imports) has worked to bolster revenue for this sample of countries. Indeed, this increase in the ratio of import duties to GDP reinforces the interpretation of the collected rate as a liberalization indicator: had the decline in the collected rate reflected weakened customs administrations, tariffs as a percentage of GDP would also be expected to fall.

The data for the econometric analyses presented below further corroborate the proposition that movements in the collected tariff rate correlate positively with trends in liberalization and are a better indicator of liberalization than trends in the ratio of trade taxes to GDP. Specifically, the data sample of 27 developing countries used in one of the econometric models reported below is reviewed for prominent trends. As is elaborated there, most of the important trade liberalization reforms for the sample were initiated after 1984. For the period 1980–85, the average ratio of tariff revenue to GDP declined, while the average collected tariff rate increased; these trends were reversed for the period 1985–92, when trade liberalization was more pronounced.

Trade Liberalization and Revenue Trends

We now review trends in international trade revenue for a broad range of countries, focusing on import duties. Although movements in the collected tariff’ rate clearly need to he interpreted with care, the previous results suggest that this variable is of some use as an indicator of whether countries are becoming more open to international trade. For the global sample, the average collected tariff rate has declined from 12.6 percent in 1975 to 9.8 percent in the last year for which data are available (Table 1). Average collected tariff rates declined for all regions. The largest proportionate decline (71 percent between 1975 and 1995) was in the grouping of countries that are members of the Organization for Economic Cooperation and Development (OECD). This likely reflects in part the relative ease with which those countries have been able to diversify their tax bases by means of the VAT and other taxes. The smallest proportionate decline was in the Asia grouping (4 percent over the same period).9

The timing of the decline in the collected tariff rate varies markedly between the OECD region and the non-OECD regions. In particular, the OECD region appears to have reduced its reliance on trade taxes gradually over the past 20 years. As a rough approximation, the collected tariff rates for the other regions were broadly unchanged between 1975 and 1985, but all declined in the subsequent decade as the pace of liberalization appears to have picked up momentum.

These outcomes, however, conceal considerable variation within regions (Table 14 in Appendix II). Among countries in sub-Saharan Africa, the tax burden on trade varies widely, as have developments over time. Collected tariff rates declined in only 8 of 17 countries between 1975 and most recently, although some did so dramatically (data on other sub-Saharan African countries are incomplete). The reductions were greatest in the Democratic Republic of Congo, Ethiopia, and South Africa, all of which reduced collected tariff rates by more than 50 percent.10 Many of the larger economies in Asia (Indonesia. Korea, Malaysia, Singapore, and Thailand) experienced lower collected tariff rates. In this region the un weighted average largely reflects the lack of movement in smaller open economies (such as the Solomon Islands), where the implications of revenue diversification toward domestic taxes raise conceptual issues (see Box 3). as well as in India and Myanmar, In the Middle East, the average collected tariff rate fell in only a few countries. These included Egypt (from 43.2 percent in 1975 to 16.7 percent in the last available year), Syrian Arab Republic (from 16.4 percent to 9.9 percent), Morocco (from 25.4 percent to 14.9 percent), and Israel (from 4.4 percent in 1980 to 0.6 percent). In the Western Hemisphere, the apparent lack of progress obscures significant liberalization in the largest economies, notably Brazil and Argentina, and may again reflect the difficulties of the smaller, more open economies (The Bahamas being an extreme example) in diversifying the tax base.

The previous section stressed that trade liberalization need not result, in the short run, in a decline in trade tax revenue. Over time, however, as tariff structures move away from initially restrictive conditions, the shares of trade tax receipts will most likely decline. Table 2 summarizes, by region, trends in the collection of trade tax revenue as a percentage of GDP over the past 20 years. The numbers suggest that trade liberalization over this relatively long period has, on average, resulted in a decline in trade tax revenue. Trade taxes for the full sample have diminished modestly, from 4.2 percent of GDP in 1975 to 3.2 percent most recently.

All regions witnessed declines in the ratio of trade tax revenue to GDP. The sharpest reductions have been in export taxes. Note that the Asia grouping actually witnessed some increase in the ratio of import duties to GDP, whereas in the Africa grouping that share was broadly unchanged. Both groupings recorded decreases in their collected tariff rates over the same period. In the 1985–95 period, when trade liberalization in general accelerated, the proportionate decline in collected tariff rates was greater than that in the ratio of import duties to GDP. During that same period, the ratio of imports to GDP rose from 30½ percent to 32¼ percent, likely reflecting the impact of liberalization.

Table 15 in Appendix II presents trends in the ratio of tariffs to GDP for individual countries. Again, there is considerable variation. Perhaps of greatest interest is how movements in collected tariff rates and the ratio of tariff revenue to GDP have differed for individual countries. Specifically, there is some additional support for the proposition that the collected tariff rate may be a better indicator of trade liberalization than the GDP share of tariff revenue. For the Africa grouping, where trade liberalization was typically a late-1980s phenomenon, examples of where the collected tariff rate points to liberalization whereas the ratio of tariffs to GDP increases or declines only modestly include Ghana, Kenya, Malawi, and (in the 1990s) Senegal (Figure 2). Among the Asian countries, in the Philippines in recent years the collected tariff rate has declined even as the ratio of tariff revenue to GDP has increased: for Singapore, on the other hand, both variables fell over the sample period; this, however, is a case where liberalization has proceeded to the point where revenue losses are likely to be inevitable. Singapore shares this pattern with many of the OECD countries. In the case of the non-OECD Western Hemisphere grouping, the pattern in recent years (1985 to the most recent year available) is one of relatively similar movements in both the collected tariff rate and the ratio of tariff revenue to GDP. For this grouping, the typical reliance on tariff revenue (as a percentage of GDP) is intermediate between the OECD and the African experience, suggesting that these countries may on average have reached the stage where further trade liberalization may result in revenue declines.

Box 3.Trade Liberalization in Small, Open Economies

The conflict between trade liberalization and revenue concerns can appear to be greatest for small, open economies, particularly if the initial stages of trade reform have already been completed. These economies typically rely heavily on international trade taxation for revenue; absent revenue concerns, such economies also typically stand to gain the most from free trade. Consider two examples. The Bahamas and the Solomon Islands (see Table).

Trade and Tax Revenue in Two Small, Open Economies, 1995(In percent of GDP)
ItemThe

Bahamas
Solomon

Islands
Imports of goods and services5978
Tax revenue1614
Taxes on international trade1013
Sources: IMF, International Financial Statistics, various issues, and World Economic Outlook, various issues; and staff estimates.
Sources: IMF, International Financial Statistics, various issues, and World Economic Outlook, various issues; and staff estimates.

Both economies are exceptionally open and rely heavily on trade taxes for revenue. However, the implications of shifting the tax system toward greater reliance on domestic taxes such as a sales tax need to be assessed carefully. Analytically, there are many equivalencies between taxes with apparently dramatically different notional bases. In the current context, an across-the-board import tariff levied only on final goods imports can approximate the impact of a domestic sales tax in economies where the bulk of final goods are imported; also, the efficiency consequences of a shift to a sales tax may be quite modest, given that the bulk of sales tax revenue would still be collected at customs on imported goods.1 In cases such as these, revenue needs imply that imported goods will be subject to significant taxation in some form.

Accordingly, in addition to considering a shift to a sales tax or a VAT (including their demands on tax administration) to support trade reform, such countries might consider the intermediate course of ensuring that the trade tax regime is efficient and neutral. They might also consider supplementing such a regime with taxes on domestically produced consumption goods. For example. The Bahamas has taxes that target tourism. It has also been stressing, among other things, improvements in customs administration to bolster revenue, although IMF staff have advised that a sales tax may eventually be necessary. The Solomon Islands has seen a weakening in the performance of trade taxes as a result of exemptions from import duties, compounded by heavy reliance on volatile revenue from export taxes on logging. In this case, one could argue for limiting import exemptions and expanding domestic taxes to reduce reliance on export taxes.

1 The stress here is on economic equivalencies. Different legal standards may apply to different taxes that are economically equivalent in certain circumstances. These different standards could arise from constitutional law or international commitments. In addition, to the extent that there is significant nontraded domestic consumption and/or the potential for increased domestic production, the efficiency gains from the switch to domestic taxes correspondingly increases. See the discussion of this issue in Frenkel and Razin (1992).

As noted earlier, domestic indirect taxes collected on imports are another important source of revenue. It is useful to assess a country’s overall dependence on tax collections from imports when evaluating the impact of trade liberalization on revenue. First, overall collections from imports more accurately reflect the immediate vulnerability of a country’s revenue collections to shifts in foreign trade. Second, the interaction between trade taxes and domestic indirect taxes on imports may be important in the design of trade and tax reforms where eliminating protection is one objective. For sub-Saharan Africa the lack of data allows only a selection of illustrative examples (Table 3). These examples indicate that domestic taxes on imports could be significant. Accordingly, the revenue effect of trade liberalization will depend on both the tariff structure and domestic taxes on traded goods.

Econometric Analysis

As the previous discussion has indicated, movements in international trade revenue will, in general, be determined by many factors. These include the nature and degree of liberalization of tariffs and non-tariff barriers; changes in the foreign exchange regime, the exchange rate, or both; developments in imports and exports; other structural characteristics of the economy, such as the level of development and the effectiveness of tax and customs administrations; the macroeconomic environment; the impact of trade liberalization on growth; and the nature of the domestic tax system. At the same time, the comparative analysis presented above suggests that countries have, in the short run. implemented trade reform in a manner that has not been costly in terms of trade tax revenue, although, in the long run, trade tax revenue has tended to decline in these countries. To supplement these conclusions, we have estimated a series of equations to explore further how trade liberalization has in practice affected trade tax revenue for a larger sample of countries.

A related empirical literature considers the determinants of reliance on trade tax revenue. This literature has tended to focus less on the effect of trade liberalization and more on the impact of other variables (such as trade itself and economic development, the latter measured by per capita income or shares of value added of different sectors) on trade tax revenue (Chowdhury, 1993; Cole, 1992, 1993;Gemmell, 1993; Greenaway, 1984; Greenaway and Sapsford, 1987; Hitiris, 1990; Hitiris and Weekes, 1987; Ram, 1994; and Tanzi, 1987). These studies have tended to find a positive relationship between imports and import duty revenue. In addition, there tends to be a negative relationship between the level of economic development and trade tax revenue. This is thought to reflect a variety of features of developing countries, including difficulties in administering an effective domestic tax system and the relatively small share of the formal sector in the economy. As a country develops, it will gradually move toward a “developed” tax system, relying more heavily on income taxes and broad-based sales taxes.

This paper uses two models, deriving from a common framework, to focus on the effects of different determinants of trade revenue. In each case the model starts with the notion that trade tax revenue is, by definition, the product of the tariff rate and the level of imports, both of which are simultaneously determined by the decisions of governments, consumers, and producers.

Figure 2.Selected African Countries: Trends in Collected Tariff Rates and in the Ratio of Tariff Revenues to GDP

Sources: IMF, Government Financial Statistics, various issues; and World Economic Outlook (October 1997)

In the first model, hereafter called Model A, the tariff rate, and hence trade revenue, are modeled as a function of the import base and other variables. Trade liberalization is captured by dummy variables. This is a virtue of the model in that it allows one to discriminate between the impacts of different types of liberalization on revenue; the drawback is that the use of dummy variables precludes the precise measurement of the relationship between trade revenue and the tariff rate. The model makes use of a panel data set covering 27 countries over the period 1980–92 (described further below). It provides detailed qualitative information on aspects of liberalization in a relatively small group of countries undergoing trade liberalization.

In the spirit of Model A, Winer and Hettich (1991) evaluate developments in the revenue structure of Canada from 1871 to 1913. They incorporate major trade reforms into their estimated equation through the use of dummy variables for years in which there were large, discontinuous changes in tariff rates, tariff coverage, or the U.S. trade regime (which is assumed to have spillovers on Canadian fiscal policy). They find that these trade variables exert a significant influence on Canadian reliance on trade taxes over this period.

The complementary approach (Model B) focuses on the demand for imports. If trade liberalization simply entailed reducing tariffs, the revenue implications would depend on the elasticities of import demand and supply with respect to the tax rate. A sufficiently elastic response would imply that revenue would increase as a result of a decrease in tariffs. Rather than estimate a demand curve, this alternative model subsumes the price elasticity of demand in a relationship between trade revenue and the tariff rate. The restrictiveness of the trade regime is proxied by the collected tariff rate, which for this data set is taken to describe trade policy in all its dimensions. An advantage of this model is that it uses a panel data set that encompasses a broader range of countries than does that for Model A, including countries not undergoing trade liberalization, and a longer time period. Despite the differences, the two approaches and data sets yield complementary conclusions.

Description of Model A

In Model A the effect of trade liberalization is incorporated through a set of dummy variables that describe trade liberalization, and through the ratio of imports to GDP, which captures the effects of liberalization on trade. Accordingly, this model separates two principal theoretical considerations, namely, the effect of liberalization on revenue, holding the import level (and other variables) constant, and the effect of changes in the level of imports on revenue, holding liberalization measures (and other variables) constant. However, because the import-GDP ratio could vary for reasons other than trade liberalization, and because the variable encompasses total imports and not just imports subject to tariffs, this variable plays an independent role in the estimations. This model uses a panel data sample of countries, drawn from a recent World Bank study (Dean, Desai, and Riedel, 1994), which contains information on when each country undertook significant trade reform and the nature of that reform.11

The model uses three dummy variables representing reduction of quantitative restrictions or foreign exchange constraints, reduction of import tariffs, and reduction of export barriers (notably, export taxes). When reform is judged to have occurred for each variable is based on the World Bank study (Table 16 in Appendix II). For each country in the sample, in the year in which trade liberalization was initiated for any of these areas, and in subsequent years, the corresponding dummy variable takes a value of one and otherwise is zero.12 It is not uncommon for countries to undertake liberalization by simultaneously reducing quantitative restrictions and import tariffs, and this might make it difficult to separate the effects of these policy actions with the two dummy variables. The correlation coefficient between the two variables in this data set is only 0.47, however. Also, this specification only tests the empirical relationship between the broad type of trade liberalization and revenue. For instance, countries may accompany liberalization of quantitative restrictions without changing tariffs, or they ma; replace quantitative restrictions for certain goods with tariffs. This analysis controls for the tariff level (through the tariff dummy variable) only insofar as the country also undertook tariff reform as part of its trade liberalization. In addition, the extent to which tariff reform resulted in reductions in tariff rates varied across reforming countries.

The final estimated equation takes the following form:

where TR is import (or trade) tax revenue as a percentage of GDP; M is imports as a percentage of GDP; M represents one or more other continuous variables, such as exports and the exchange rate; andD is the set of trade liberalization and other dummy variables. Specifically, in addition to the liberalization dummies, the independent variables are imports as a percentage of GDP, exports as a share of GDP, per capita income in 1990 U.S. dollars, dummy variables to indicate whether the country has a VAT and whether it has accepted IMF Article VIII status (a possible indicator of liberalization of the trading regime), and a real exchange rate index (defined such that an increase indicates a real depreciation).13 Macroeconomic conditions are captured through the real exchange rate, per capita income, and time dummy variables.

Prior expectations about the revenue effect of trade liberalization—defined as the elimination of quantitative restrictions and reductions in average tariff levels—would suggest that the coefficients on the quantitative restrictions reform and tariff reform dummies might be zero and negative, respectively, given that results are conditioned on the level of imports. The coefficient of the export barriers reform dummy would be expected to be negative when trade tax revenue is the dependent variable. In addition, one would expect the coefficient on the import-GDP ratio to be positive. Although there is some a priori ambiguity, the coefficient on the real exchange rate variable could in general be expected also to be positive, since a real depreciation implies higher nominal values for imports and hence increased tariff revenue. The coefficient on the per capita income variable is expected to be negative, given the observation that the more developed the country, the less likely it is that the country will rely on trade taxes. Priors for the coefficient on the VAT dummy are not strong—the coefficient could be either negative (because the VAT substitutes for trade revenue) or positive (if the presence of a VAT is consistent with strengthened revenue administration). Priors on the remaining variables are also not strong.

As already noted, the data set for this model is based on countries in the 1994 World Bank study, but it excludes centrally planned economies and countries for which significant trade reform episodes could not be identified. The data set consists of 27 countries from Africa. Asia, and the Western Hemisphere, for a total of 344 observations spanning 1980 to 1992, A few observations were dropped because of missing data. The sample is not random in that it only includes countries undertaking liberalization; hence the results are conditional on the countries undertaking liberalization. Before turning to the estimation results, it is useful to review some of the basic characteristics of the data set. Table 4 summarizes developments in the ratio of tariff revenue to GDP and the collected tariff rate. These observations are best interpreted by recognizing that, with two exceptions (South Africa and Tanzania in 1984). trade reforms were initialed for this sample of countries either in 1985 or thereafter (Table 16 in Appendix II).

These trends reinforce the analysis of the case studies.

Specifically, breaking the sample period in two suggests the following interpretation;

  • For the period 1980–85, the ratio of tariff revenue to GDP declined, while the collected tariff rate increased. This was a period during which there was less liberalization on average than after 1985, and in some cases there was significant import compression: the latter is reflected in a decline in the import-GDP ratio.

  • The trends are reversed for the period 1985–92, a period that saw considerable efforts to liberalize. Specifically, there was a modest decline in the collected tariff rate, whereas the ratio of import tariffs to GDP increased. These results likely reflect the positive impact of liberalization on imports.

Estimation Results for Model A

Table 5 reports the estimation results for Model A. The dependent variable is cither import tariff revenue as a percentage of GDP, in column 1, or total trade tax revenue as a percentage of GDP, in column 2.14 The equations were estimated in a linear logarithmic form.15 In addition, the estimations use the least-squares method with a fixed-effects specification for the cross-sectional and time-series units. In the estimations, the Hausman specification test (comparing a cross-section random-effects to a fixed-effects specification) favored the fixed-effects specification.

Contrary to initial priors, the results for the import tariff ratio (column I) indicate that the coefficient for the quantitative restrictions reform dummy is significantly positive, whereas that for the tariff reform dummy is not significantly different from zero. Both results are consistent with this sample of developing countries having, on average, implemented trade liberalization with a view to avoiding adverse revenue consequences. The insignificance of the coefficient on the tariff reform dummy is consistent with there being trade liberalization packages that included outright tariff reductions, some of which reduced very high tariff rates with low import volumes, as well as reductions in the dispersion of tariffs and increased minimum rates. In addition, reforms might have resulted in a shift of the composition of imports in favor of goods subject to tariffs or taxed at higher rates, such as consumer goods, which would buoy revenue and offset the tendency for revenue to decline because of lower tariff schedules.16 The significantly positive coefficient on the quantitative restrictions reform dummy is consistent with reforms of quantitative restrictions that have involved tariffication of constraints, or resulted in a compositional shift within recorded imports toward higher-taxed goods, or both. Finally, the export restrictions reform dummy is not significant.

Turning to the other variables, the coefficient on the import-GDP ratio is positive and significant, as expected, and likely reflects the fact that countries that import heavily are more likely to have more import tax revenue. This implies that when trade liberalization leads to higher imports, revenue may rise, and that trade liberalization accordingly need not be accompanied by a decline in revenue.17 The positive coefficient on the real exchange rate is also in line with expectations, and because imports are held constant, it suggests that as the real exchange rate index rises (i.e., as the currency depreciates in real terms), the value of imports rises in local currency terms, and hence so does revenue. This result is consistent with the observation that supporting macroeconomic policies can assist successful trade liberalization. The coefficient on the export share of GDP is negative and significant. At least two interpretations of this result are possible. First, a higher export share may be associated with an import pattern weighted more toward the use of imported inputs, which typically have lower tariff rates. Second, a higher export share, all else equal, is consistent with a more open trading regime and hence a higher probability of associated pressures by exporters to limit tariffs. The coefficient on the per capita income variable is insignificant, possibly reflecting the fact that this sample is composed entirely of low-income countries.

The results for the trade tax ratio (column 2) are similar to those for the import tariff ratio equation, with the differences being mainly intuitive. In this equation the import-GDP ratio, the real exchange rate index, and the quantitative restrictions reform dummy variable are, again, positive and significant. However, the export share of GDP is now positive and marginally significant, rather than negative and highly significant. This presumably reflects two conflicting influences: a higher share of exports in GDP might tend to exert a downward influence on import revenue, but it would exert an upward influence on export tax revenue, and this effect appears to dominate. The export restrictions reform dummy in this equation is negative and significant, which is consistent with expectations, since export restrictions often take the form of export taxes.

The interpretation of the positive coefficient on the VAT dummy across the two equations (insignificant in the import tariff ratio equation, significant in the trade tax ratio equation) is not obvious. Although a positive sign is consistent with the interpretation of the VAT as representing a better overall quality of tax administration, and hence leading to higher collections, it is not clear why this link would be stronger for trade tax revenue than for import tariff revenue.18

To sum up, Model A’s results for this sample of developing countries are consistent with several conclusions. The first is that tariff reforms for a given level of imports have not been significant in reducing trade tax revenue, whereas reforms of export restraints (taxes) have had a significantly negative impact. Second, reforms of quantitative restrictions may have had a positive impact on revenue. Third, trade liberalization that stimulates imports may serve to strengthen reform efforts by buoying international trade tax revenue. Fourth, supporting macroeconomic policies, especially exchange rate policies, may be useful in strengthening revenue.

Description of Model B

Against the background presented earlier, an alternative model is specified that takes the following form:

where TR is trade tax revenue as a percentage of GDP; CT is the collected tariff rate; w refers, as before, to one or more additional continuous variables, and D represents the set of dummy variables. The independent variables are the same as in Model A (and have the same priors), with the following exceptions. First, the liberalization dummies and the import and export ratios of that model have been replaced by the collected tariff rate, the principal measure of trade liberalization in this case. As already noted, the collected tariff rate may decline when tariffs are sufficiently high. More generally, the relation between the collected tariff rate and the dependent variable is nonlinear. A quadratic form is therefore used as a local approximation of observed behavior: the overall marginal effect of the collected tariff rate on trade revenue is then (b1+2b2CT) and hence depends onCT. Second, in some countries export taxes are an important component of trade tax revenue. Hence, parallel to the collected tariff rate, a collected export tax rate is also used as an independent variable. This variable would be expected to be positive when the dependent variable is defined as trade taxes as a percentage of GDP—it would not be expected to be significant when the dependent variable is import duties as a percentage of GDP.

Countries typically face various constraints on their choice of tariffs, including scheduled reductions in tariffs as part of international or regional trade agreements. However, to the extent that the collected tariff rate is viewed as a choice variable of a government subject to an overall revenue constraint, it is likely to be endogenous in equation (2).19 Moreover, as already noted, the collected tariff rate is likely to measure the underlying average statutory rate with an error. We treat these issues using an instrumental variable approach. This requires finding an instrument that is related to trade tax revenue, but unrelated to the error component of the collected tariff rate. Two alternatives are adopted here. The first is to use the lagged value of the collected tariff rate. The alternative instrument selected is acceptance of Article VIII, which is interpreted here as an indicator of liberalization of the trade regime unrelated to revenue. The instrumental variable models are estimated defining all variables as before and substituting predicted values, estimated by the first stage, for the average collected tariff rate and its square in equation (2).

The data set for Model B encompasses 105 countries, spanning 1980 to 1995.20 All countries in the world were included, except those in transition from central planning, those that did not exist throughout the sample period, and those that had observations for fewer than four years for any of the variables. The resulting data set consists of 1,575 observations (1,680 observations when Article VIII status is used as an instrument). Missing observations were imputed using country-specific means.21 Finally, because the data set employed in Model B is similar in scope to that presented in the discussion of trends above, no further summary statistics are presented.

Estimation Results for Model B

The model was estimated by the least-squares method with both fixed and random effects. Table 6 reports the estimation results for the preferred version of the overall model for the case where the dependent variable is the ratio of trade taxes to GDP and the instrument is the lagged value of the collected tariff rate.22 Column 1 presents results for the first stage and columns 2 through 4 for the second stage. Table 7 reports the disaggregated results for that model for each major region in the sample (OECD, Africa, non-OECD Asia, Middle East, and non-OECD Western Hemisphere), This disaggregation is obtained by interacting the instrumental values for both the collected tariff rate and its squared term with regional dummies. As with Model A, the Hausman specification test statistic implies that the fixed-effects specification is preferable to the random-effects specification for the overall model, but the lest statistic is more ambiguous for the model with the regional disaggregation. As expected, in both specifications the collected tariff rate and revenue from trade are significantly positively related, which implies that reductions in the collected tariff rate are likely to entail a revenue loss. The results also confirm the hypothesis that there is a diminishing revenue return to increases in the collected tariff rate, as the squared term is significantly negatively related to trade revenue for both specifications.

Specifically, the coefficients suggest that there is a revenue-maximizing collected tariff rate. For this regression, that rate is about 231/2 percent. This finding supports a result reported above, namely, that trade liberalization may not be very costly in terms of revenue. Whereas earlier it was argued that revenue losses could be mitigated through the appropriate prioritization of reforms, these results suggest that even a significant reduction in tariffs, as measured by the collected tariff rate, might not entail a revenue loss for countries with high initial levels of tariffs.

The empirical result indicating the existence of a revenue-maximizing collected tariff rate is of greater significance than the specific value of that revenue-maximizing rate. That value will clearly depend on a country’s circumstances, including the structure of the economy, the stale of customs administration, and the nature of supporting policies. Indeed, some indication of how much the actual rate can vary can be seen from the 95 percent confidence interval around the 231/2 percent value, which implies a range from 21 percent to more than 26 percent.23

When acceptance of Article VIII is used as an instrument, the revenue-maximizing collected tariff rate is about 15 percent.24

Turning to other variables in the specification, the positive coefficient on the collected export tax rate conforms to expectations.25 The coefficient for per capita GDP is significantly negative, as expected, supporting the hypothesis that, as countries develop, trade taxes become less important as a source of revenue. A possible explanation for this variable being significant here, but insignificant in the estimating equations for Model A, is the greater range of countries covered in the Model B sample. In addition, the real exchange rate has a positive and significant coefficient, conforming with priors. The VAT dummy is not significant.

To provide further evidence on the model, in Table 6, column 4, estimates are presented for a “between estimator,” which uses country-specific means of all variables and, hence, measures only cross-country variation. As expected, the magnitudes of the coefficients are quite different, as in this specification the information from the time-series variation is neglected. Nevertheless, the estimated revenue-maximizing collected tariff rate is broadly unchanged.

Additional information is obtained when the results are disaggregated by region. First, the OECD subsample does not have a revenue-maximizing collected tariff rate, suggesting that the revenue-maximizing collected tariff for the full sample needs to be interpreted with caution. The other regional subsamples, however, have revenue-maximizing collected tariff rates that range from 17.8 percent (for the Africa region) to 28.3 percent (for the Asia region). On further examination, the result for the Asia region is heavily influenced by the inclusion of India and Myamnar. Excluding these two outliers from the sample results in a revenue-maximizing tariff rate for the Asia region of 20 percent, which is within the range reported for the other regions.

The data indicate that many countries with very high levels of tariffs have not reduced their collected tariff rates. The number of countries with a collected tariff rate in excess of 25 percent increased from 7 in 1980 to 9 during the sample period (1980–95); the number of countries with a collected tariff rate in excess of 20 percent also remained significant, falling only from 19 in 1980 to 16 in 1995 (out of a sample of 105 countries). These data suggest that a significant number of countries could achieve further reductions in tariffs without major revenue loss.

Finally, in interpreting this result, note that both the weighted and the unweighted average statutory tariff rates associated with any specific revenue-maximizing collected tariff rate would be significantly higher in the presence of evasion or of tariff exemptions for some imports. An additional reason why the actual tax burden may be significantly higher than implied by any collected tariff rate is that many goods subject to tariffs are also subject to domestic taxes.

The main conclusions that emerge from the econometric analysis are the following:

  • The results, particularly the dummy variable results, are consistent with the proposition that trade reforms in many developing countries have been fashioned so as to minimize adverse revenue consequences.

  • The results indicate that trade liberalization that leads to increased imports has a positive impact on trade revenue. This indicates more generally that, whatever the initial impact of reform on revenue, economic growth and increasing imports are likely to buoy trade tax revenue over time.

  • The evidence of a revenue-maximizing collected trade tax rate implies that some countries with collected tariff rates above revenue-maximizing levels may be able to liberalize without adverse revenue consequences.26 Both the weighted-average and the unweighted-average statutory rates associated with this result would be significantly higher than the collected rate, given evasion and tariff exemptions. In that connection it is important to remember that many goods subject to tariffs are also subject to domestic taxes, increasing the incentive to substitute lower-taxed goods for higher-taxed ones and to evade taxes.

  • In interpreting the calculated revenue-maximizing tariff for any given country, it is also important to take account of the circumstances of that country (e.g., the degree of its openness to trade, the efficiency of customs administration, the presence or absence of supporting policies). In addition, as the empirical results in Model A show, the revenue impact will be influenced by the pattern of trade liberalization.

  • The other side of the behavior of the collected tariff rate variable is that trade tax revenue will lend to decline as trade liberalization results in ever deeper cuts in tariff levels.

  • There is also support for the proposition that the level of trade revenue depends on supporting macroeconomic policies. In particular, the results indicate that a depreciation in terms of the real exchange rate commonly enhances trade tax revenue, at least until the full effects of the depreciation are felt throughout all sectors of the economy.

As already mentioned, the collected tariff rate is the ratio of import tax collections to import values. Reference is also made in this paper to the collected trade tax rate, which is the ratio of trade taxes—including export taxes—to the value of trade (imports plus exports).

The CBI, endorsed at a ministerial-level meeting in 1993, calls for countries in eastern and southern Africa and the Indian Ocean to pursue comprehensive trade reform by, among other things, dismantling quantitative restrictions on a reciprocal basis, and to establish a common external tariff.

The range reflects alternative assumptions about how much Morocco will substitute LU imports for non-EU imports.

IMF staff reports from that time suggest that the high tariff and tax rates were encouraging the growth of the informal sector, which might help explain the reported drop in imports as a percentage of GDP.

For example, although the outright removal of quotas will tend to result in a decline in the collected tariff rate as imports increase, when quota reform takes the form of tariffication the collected tariff rate and trade liberalization may move in opposite directions.

In the case of smuggling neither the tariff due nor the actual import is captured by customs-based trade data. This paper, however, uses balance of payments import data, which often attempt to incorporate estimates of imports not captured by customs data. More generally, these points suggest that interpreting the absolute level of the collected tariff rate as a gauge of trade liberalization would be problematic.

The data for the simple average tariff rate, which were limited to a few observations—often from alternative sources—for Argentina, Morocco, the Philippines, Poland, and Senegal, are available from the authors upon request.

Given that changes to trade regimes are typically ongoing rather than isolated events, deciding when the most significant recent reform phase began is inevitably somewhat arbitrary. One of the criteria used was whether trade reform occurred along with supporting reforms, such as a strengthened domestic tax system, that might indicate a break in policy formulation. For the six countries, the pre reform years were identified as follows: Argentina. 1990; Malawi, fiscal year 1986/87; Morocco, 1986; the Philippines, 1985 (i.e., the whole period for which data were available); and Senegal, 1993, Poland is excluded from The calculations given the short time period in that case.

The data for the European Union represent trade taxes collected by individual EU members and transferred to the central EU budget. Since goods are freely transshipped within the union, collections for individual countries can overstate or understate trade taxes on goods for domestic use. For example, the data for the Netherlands include taxes levied on goods transshipped to Germany.

Absent detailed information on these cases, the possibility that the fall in the collected tariff rate in individual cases reflected a collapse in administrative abilities cannot be ruled out.

The initial level of openness in these countries is quite different, and a few, such as Korea and Sri Lanka, had already undertaken reforms prior to the period of study.

Trade liberalization was rarely reversed, and when it was, it was not reversed completely. Hence, it seems more accurate to assign all subsequent years in the sample a value of 1 (with the exception or Senegal, where the trade reform of the mid-1980s was clearly reversed after a certain point).

For both models, all financial data are obtained from the Government Finance Statistics and World Economic Outlook databases of the IMF. Per capita GDP data are from the World Bank database, and information on whether the country had a VAT and had achieved Article VIII status is from IMF sources.

Modeling this revenue us a share of total tax revenue, as in some past studies, had little effect on the results.

Data for the import or trade tax share, import and export shares, and per capita income are in logarithmic form. The real exchange rate index is center in linear form.

The insignificance of the tariff reform dummy could also reflect the fact that, as a result of liberalization, the share of goods subject to tariffs in total imports might rise as evasion declines, in response to improved customs administration, or both.

The coefficient in the regression equation indicates that a 1 percent increase in the import-GDP ratio results in a 0.7 percent increase in the ratio of import revenue to GDP.

One possible explanation is that exporters under a VAT have an interest in getting a credit on any VAT they may have paid; as a consequence, more exports may be captured by export taxes.

The Hausman exogeneity test was highly significant for this equation and supports the need for an instrumental variable approach.

A list of the countries in the sample for Model B is available from the authors on request.

This method will generally increase the standard errors of the estimation relative to other methods of imputation. An alternative is simply to drop observations. Sensitivity analysis suggested that this did not have a substantial effect on the results.

In addition to tests of alternative instruments, other specifications were also run, including linear forms and a specification using import tariffs alone as the dependent variable. The results remained basically unaltered and are available from the authors upon request.

In an internal World Bank study, estimates for Bangladesh, for example, using import elasticities of different categories of goods (consumer, raw materials, etc.) are broadly in line Will the figures estimated here. In the case of Bangladesh, a reduction in the trade-weighted average tariff from 17 percent to 14 percent was revenue enhancing. In the related literature that estimates import demand functions and import demand elasticities, similar results are obtained; see most recently, for example, Clarida (1996) and Goldstein and Khan (1985).

Finally, when the lagged value of the dependent variable is used as an instrument and the ratio of import duties to GDP as the dependent variable, the revenue-maximizing rate is about 20 percent.

As expected, this variable is insignificant when import tariff alone is the dependent variable.

To stress the obvious, the existence of a revenue-maximizing rate should not be interpreted as providing countries with lower tariffs a rationale for increasing their revenue by raising tariff rates.

    Other Resources Citing This Publication