7 Why Is Trade Reform So Difficult in Africa?

Zubair Iqbal, and Mohsin Khan
Published Date:
December 1998
  • ShareShare
Show Summary Details
Dani Rodrik

The year 1996 was a rare good one for Africa. The continent registered its highest rate of economic growth in two decades: 5.0 percent (IMF, 1997). Among the larger economies, the star performers were Uganda, Ghana, Cameroon, and Côte d’Ivoire, along with Morocco and Tunisia (see Table 1). While these are good signs, it will take many years of growth at such levels (or better) to undo the damage that more than two decades of stagnation and decline have inflicted on most countries of the region.

Table 1.Recent Economic Performance of African Countries, in Comparative Perspective(Annual percent change)
CountryReal GDPConsumer Prices
Cote d’Ivoire7.
South Africa3.
SAF/ESAF countries24.96.05.522.015.77.4
CFA countries4.65.25.715.36.03.2
All developing countries (average)
All developing countries (median)
Source: IMF (1997), Table 6.


African countries that had arrangements, as of the end of 1996, under the IMF’s SAF or ESAF.

Source: IMF (1997), Table 6.


African countries that had arrangements, as of the end of 1996, under the IMF’s SAF or ESAF.

The turbulence experienced in world markets since the mid-1970s has had severe adverse effects on both Latin America and Africa. The upshot in Latin America has been the wholesale adoption by virtually all governments in the region of orthodox recipes—namely, fiscal retrenchment, deregulation, free trade, and privatization. In sub-Saharan Africa, free market religion has found far fewer converts. Despite tremendous pressure from donor governments and multilateral agencies, African policymakers have generally been more skeptical about the value of opening up their economies and reducing the role of government. Consequently, reforms have progressed rather gradually and have been full of interruptions and reversals. For example, a World Bank review of trade policy reforms in Africa concludes (Dean, Desai, and Riedel, 1994, p. 50):

Reversal of reform has been frequent. In seven of the countries examined, either restrictions which were removed were reinstated, or some existing barriers were strengthened to offset reductions in others. Nigeria, though it eliminated most quantitative restrictions (quotas and licensing) increased dramatically the number of import bans. Ghana, which was the only country to make great strides in cutting formal tariffs, reversed this with the implementation of large special taxes on imports. Cote d’Ivoire raised tariffs significantly, after having reduced QRs. In some cases the motive for reversal appears to be pressure from import-competing industries as they begin to experience competition from abroad (e.g., Cote d’Ivoire, Ghana). In others, resurgence of foreign exchange shortages have slowed the liberalization of tariffs (Madagascar), or reversed the foreign exchange market reform itself (Kenya).

Inadequate implementation of reforms is one of the most common themes running through the literature on African economic policy (see also Metzel and Phillips, 1997, and Gulhati, 1990).

But more than that, it is also the vacillating nature of policymaking that stands out. Collier cites the Nigerian example: “In the past decade, Nigerian trade policy has swung from intense foreign exchange rationing, indicated by a parallel market premium over 300 percent, to a completely free market, back to even more intense rationing and most recently back to a free market” (1995, p. 548). The contrast with Latin America, where governments have stuck with ambitious reforms even under severe macroeconomic difficulties—for example, during the Mexican peso crisis of 1995—is quite striking. As a result, the credibility of African reforms tends to be low, which itself creates a severe problem. The desired supply responses—in investment and exports—are unlikely to materialize when significant uncertainty is attached to the continuation of the reforms.

The very recent evidence notwithstanding, there remains considerable controversy over whether World Bank-IMF–type adjustment programs, of the sort adopted in Latin America, do work in Africa. Compare, for example, the World Bank’s own positive evaluation (World Bank, 1994) with the critique by Lall and Stewart (1996).1 Fortunately, for my purposes I do not need to get into the issues involved in this debate. Extremists aside, there is actually a fair bit of consensus on what constitutes a reasonable trade strategy for the countries of Africa. This consensus can be crudely expressed in terms of a number of do’s and don’t’s: demonopolize trade; streamline the import regime, reduce red tape, and implement transparent customs procedures; replace QRs with tariffs; avoid extreme variation in tariff rates and excessively high rates of effective protection; allow exporters duty-free access to imported inputs; refrain from large doses of anti-export bias; do not tax export crops too highly. Not only is there wide agreement on these policies, there is also less dissent than might appear at first sight on what is to be considered “extreme” or “too high.”2 These desiderata still leave considerable room for policymakers to make their own choices over a wide range of trade and industrial-policy options.

Some aggressive reformers like Ghana and Uganda (and Mauritius before them) have implemented most, but not all, of the above agenda. But even in this group, reversals have not been uncommon. Ghana, as mentioned in the quote above, has implemented a special import tax of up to 40 percent on top of its import tariffs and retains a relatively large tax on cocoa exports (GATT, 1992). Other countries have done much less. Tariffs remain high, trade monopolies continue to exist in many sectors, export crops continue to be taxed, and trade procedures continue to be characterized by red tape and corruption (see the discussion in Metzel and Phillips, 1997). The main question, then, is why so little progress has been made with reforms that are endorsed by economists of diverse persuasions.

As usual, when economists’ prescription meets reality, the answer is: politics. Political scientists who study Africa have long argued that it is distributional issues that prevent the adoption of economically sensible policies. Bates (1981), for example, has provided the classic argument for why African governments tax agricultural exporters so exorbitantly: the motive is to transfer wealth from politically unorganized rural groups to vocal urban groups. Bienen (1991) faults the policymakers more directly (1991, pp. 76–77):

Trade liberalization policies are often extremely hard to formulate and implement in Africa precisely because it is powerful officials (civilian and military) who benefit from the controls that have been established over imports and exports. It is government officials who ration and distribute scarce imports, including foreign exchange. They realize the rents which accrue from the systems they construct and control. Of course, officials have allies—import-substituting manufacturers and urban workers employed by state enterprises who are interested in subsidized urban consumer goods.

Bienen argues that the main constraint is not import-substituting urban producers themselves but self-interested government officials: “A policy that moves away from tariff protection of domestic industries will not face strong private sector capitalists or workers in Africa. Such policy shifts face strong public opposition in Africa …” (1991, p. 82).

To a first degree of approximation, such distributional arguments are indeed powerful, and I will elaborate on them below. But they need to be complemented with other stories. Consider some of the problems.

First, the distributional or interest group perspective is too static and deterministic. The more complete is an explanation of why the prevailing, economically dysfunctional policies are the endogenous outcome of interest group pressures, the harder we make it to understand why governments sometimes do reform (or, for that matter, how they can reform). After all, the relative powers of interest groups rarely change very quickly. Governments, conversely, sometimes do reform, and occasionally do so quite rapidly. In Latin America, where many of the same distributional arguments were made to explain the prevalence of import-substitution policies, we have observed governments undertaking wholesale reforms in a relatively short period of time. Moreover, some of the leading reformers were drawn from the constituencies most associated with the previous set of policies—for example, the Peronist Menem in Argentina.

Second, distributional arguments often have an element of circularity. Every policy configuration generates winners, and it is all too easy to associate the raison d’être of some particular configuration with the political salience of a set of winners. Often unasked is the question: couldn’t an alternative set of policies benefit these winners even more? This is not a hypothetical question. It is a good bet, for example, that the primary beneficiaries of the reforms in the Republic of Korea and Taiwan Province of China around 1960 were the import-substituting industrialists who were resisting these reforms. A corollary is that interest groups can apparently change their minds, as certainly happened in East Asia once exports took off. In Africa, the great majority of civil servants whose salaries greatly eroded during the 1980s as a consequence of fiscal cutbacks would almost certainly have been better off had they followed fiscally more conservative policies and safeguarded the tax base. More broadly, since reforms are expected to enlarge the economic pie, why can’t the losers be compensated through transfers or other policies?

Third, distributional arguments often fall far short of explaining certain typical outcomes. The extreme fiscal distress to which many African countries have succumbed, and the associated macroeconomic instability—inflation, widespread shortages, reduced incomes—do not create any obvious winners. It is hard to identify significant interest groups that benefit from such difficulties. Another manifestation of this same issue is that the extremes of taxation that have often been observed in Africa are actually inconsistent with maximizing the wellbeing of even the recipients of the revenues thereby generated. For example, export crops have frequently been taxed at rates surpassing the revenue-maximizing level. As McMillan (1997) has recently argued, the conventional story of urban bias (for example, Bates 1981) cannot account for this feature of policy. After all, why would governments, and the urban groups they represent, want to kill the goose that lays the golden eggs? Once again, we need to complement the conventional distributional story with other elements (as McMillan does; see below).

I will focus in this chapter on a succession of models that yield partial insights on each of these puzzles. I will start with the standard distributional story, cast in an appropriately general equilibrium framework. Going beyond the identification of losers and gainers, the discussion here will emphasize the likely magnitude of distributional consequences. In particular, I will argue that, in a typical African country, compensating the losers from trade liberalization is impossible for all practical purposes; the amount of redistribution required will more than eat up the efficiency gains generated by the reform. Next, I will turn to a setup where it is the government’s inability to discipline itself—rather than distributional imperatives per se—that lies at the root of the problem. I will illustrate the issues using McMillan’s (1997) argument about the dynamic inconsistency of export taxation. These ideas turn out to be relevant and powerful in the African context, and they also have important institutional implications. Finally, I will turn to incompleteness of information as a source of resistance to reform. Relying on the argument of Fernandez and Rodrik (1991), I will suggest that uncertainty surrounding the identity of likely gainers is a severe obstacle to garnering political support for reform. Paradoxically, however, it can also be a source of unraveling of support over time. This set of ideas puts a heavy burden on the need to identify gainers from reforms early on.

My objective is to show how such models can help us develop a more sophisticated understanding of the political economy of policy choice. In addition, I hope to demonstrate that they can be used to inform policymaking as well.

Distributional Consequences of Trade Reform

It is well understood that trade reform typically entails a redistribution of income among various sectors of the economy. Less well understood is how large the redistributions are relative to the efficiency benefits of the reform. This creates a political problem of major proportions.

To appreciate the issues, it is helpful to lay out a simple model of an archetype African economy. Consider an economy divided into a rural sector and an urban sector. Assume that exports originate from the rural sector, while the urban sector produces import-competing goods. Each sector has one factor of production specific to it: land in the rural sector and capital in the urban sector. In addition, the two sectors employ labor, which is freely mobile between the two. So far, the model is a standard specific-factors (Ricardo-Viner) model. To make it slightly more realistic, I assume that urban capitalists’ profits (rents) are shared with urban workers, so that wages in urban areas exceed rural wages, even though labor is intersectorally mobile. We are interested in the qualitative and quantitative implications of trade liberalization in an economy of this sort.

Formally, let Y and X represent the outputs of the urban and rural sectors, respectively. We assume the two goods are produced by neoclassical production functions of the form:

where K and T stand for the economy’s (fixed) endowments of capital and land, respectively, and the economy’s labor endowment L is divided between the two sectors:

Let the ad valorem equivalent of import restrictions be given by t. The rural or informal sector wage is given by the value marginal product of labor in the two sectors:

Note that I have normalized the relative border price of the importable at unity. The domestic relative price of the importable is consequently 1+t. The urban wage equals this base wage plus a share of urban profits, and is expressed as:

where π is the return to capital and γ is the rent-sharing coefficient. The way we model the urban wage implies that sharing rents (or profits) is efficient; urban employers equate the value marginal product of labor to its opportunity cost, which is the rural wage. The return to capital is given in turn by the residual income in the urban sector (per unit of capital) after labor is paid its marginal product:

On the demand side, we assume preferences are identical and homothetic, so that they can be aggregated. The economywide demand for the importable can be expressed using the Marshallian demand function D((1+t, D, where I is national income at domestic prices:

The quantity of imports, m, is given by the difference between demand and supply of the importable:

The nine equations above determine the nine endogenous variables in this model (X, Y, LX, Ly, wr, wu, π, I, and m) as functions of factor endowments and of trade policy, t.

A social planner who does not care about income distribution would maximize the value of national output at world prices, X + Y. This yields the familiar prescription t = 0, as trade restrictions are, by assumption, the only market imperfection in this framework.

Consider now the more realistic setting where distributional issue do carry weight. There are five distinct groups in this model whose interests are at stake: (1) farmers (that is, owners of T), (2) informal-sector workers, (3) urban workers, (4) urban employers, and (5) recipients of revenues that derive from trade restrictions. The last group is likely to be a mixture of rentseekers, government officials, and the national treasury. The incomes of each of these groups (in terms of the exportable) are expressed as follows:

Informal workerswrLx
Urban workerswuLy
Urban employers(1 + t) YwuLy
Trade rents or government revenuestm

It is straightforward to lay out the distributional impact of trade liberalization on each of these groups. Farmers are clear winners, as the domestic relative price of their output is inversely proportional to t. Informal, or nonurban, workers are also winners, provided not too much of their budget is spent on the exportable. Urban employers and workers are clear losers. What happens to trade rents and/or government revenue depends on which side of the Laffer curve we initially were on. When reform is large scale (that is, t is brought close to zero), or when QRs are converted into tariffs, recipients of trade rents are necessarily affected adversely.

Hence, trade reform pits rural groups against urban groups, and ultimately against recipients of trade rents as well. Since urban groups tend to be much better organized, and their ability to bring governments down is considerably greater, reformist politicians face a serious dilemma. This is the conventional account of the political economy of trade reform in the African context, as discussed, for example, in the works cited in the previous section.

To quantify these distributional consequences, I have run some numerical simulations using the model above. For these calculations, I take production functions in both sectors to be Cobb-Douglas, with a labor share of 0.40 in each. Preferences are similarly Cobb-Douglas, with the budget share of the importable set at 0.40. The urban profit-sharing parameter γ is fixed at 0.25. Other parameters (that is, endowments and technical coefficients) are selected so as to produce a baseline solution with a trade and production structure bearing some resemblance to a “typical” African country. The first column of Table 2 shows this baseline equilibrium, with t set to 40 percent initially. In the baseline equilibrium, 27 percent of national income accrues to farmers (owners of land), 61 percent to workers (both rural and urban), 8 percent to urban capitalists, and 3 percent to recipients of trade rents and/or tariff revenue. Urban workers earn a premium of about 17 percent over other workers. Imports are somewhat less than 10 percent of national income (0.153 ÷ 1.574), and 71 percent of the labor force is employed in the rural sector.

Table 2.Distributional Implications of Trade Reform in an Archetypal African Economy
MeasureBaseline Solution, t = 40 percentPercentage of IncomePercentage Change from Baseline Solution
t = 30 percentt = 20 percentt = 10 percentt = 0
Real national income1.5741001.081.972.542.80
Real income by group:
Urban employers0.1338-12.03-23.31-34.59-45.11
Informal-sector workers0.647416.1812.6719.7827.51
Urban workers0.30920-11.65-23.30-34.30-44.98
Quota rents/ government revenue0.05435.56-5.5640.74-100.00
Informal wages1.043-1.92-3.84-5.47-6.90
Urban wages1.217-1.97-3.78-5.51-6.98
Output of rural sector1.2333.086.088.8411.52
Output of urban sector0.361-7.76-15.79-24.10-32.69
Employment in rural sector0.7095.2210.3015.2319.89
Employment in urban sector0.291-12.71-25.09-37.11-48.45
Consumption of urban goods0.5145.8412.0618.6825.88
Volume of imports0.15337.9177.78119.61164.05

The other columns display the consequences of progressively reducing trade restrictions, expressed as percentage changes from the baseline situation. As tariffs are reduced from 40 percent to 30, 20,10, and 0 percent, aggregate real income increases by 1.1, 2.0, 2.5, and 2.8 percent, respectively. These magnitudes are in line with findings of most computable general equilibrium models, at least those that do not allow scale economies. If anything, they are perhaps a bit on the high side. Note also that most of the gains are concentrated in the initial reductions of restrictions: going from t = 0.40 to t = 0.20 yields a real income gain of 2 percent, while going the full way to complete free trade (t = 0) yields an additional gain of only 0.8 percent (2.80 – 1.97). This is not an artifact of the specific assumptions made here. It follows from economic theory, which suggests that the welfare cost of distortions rises with the square of the distortion.

From our standpoint, the results on the distribution front are more interesting. As discussed above, there are clear losers and winners from trade reform; urban groups lose and rural groups gain. Even more to the point, the magnitudes of the distributional impacts are very large. Consider for example the scenario where trade restrictions are reduced from a tariff equivalent of 40 percent (t = 0.4) to a tariff equivalent of 10 percent (t = 0.1). In this scenario, urban employers incur a real income loss of 35 percent, while recipients of quota rents suffer a loss of 41 percent! The gain to farmers is 20 percent. The net gain to the economy of 2.5 percent is an order of magnitude smaller than these distributional impacts. Put differently, the efficiency consequences of trade reform pale in comparison to its redistributive effects. (Once again, this conclusion is a general feature of models of this sort, and does not depend on any particular parameterization.)

This is the sense in which price reforms, and trade reforms in particular, tend to have high political cost-benefit ratios (Rodrik, 1994). It is not only that such reforms entail redistribution, which is well recognized. More significant is that they entail so much redistribution relative to their efficiency benefits—a point that is surely not lost on those groups whose incomes are at stake. This makes it easier to understand why trade reforms are so vigorously resisted, and why it has generally proved difficult to convince African policymakers to embark on ambitious reform efforts.

These numbers also make clear why the economist’s standard trick of assuming (or advocating) compensation is quite unhelpful to the policymaker. Of course, since there are aggregate gains to the economy—the size of the pie is larger—it is in principle possible to compensate all losers and still leave some groups better off. But what is implicit in this recommendation is the idea that the requisite transfers can be accomplished in a relatively efficient manner—in the limit by employing lumpsum transfers. This is counterfactual, especially in sub-Saharan African countries where tax instruments and administrative capacity are extremely weak.

Suppose, for example, that policymakers wished to neutralize the negative impact on urban employers arising from the above trade liberalization scenario. Since this group constitutes 8 percent of national income and its losses are 35 percent, its loss amounts to roughly 2.8 percent of GNP (0.35x0.08). Hence, the government now needs to raise 2.8 percent of GNP in taxes to finance transfers to this set of urban interests. These taxes will naturally engender their own set of distortions. The magnitude of these distortions can be gauged by considering that in the United States most studies estimate the marginal excess burden (MEB) of taxation to be in the range of 0.30–0.40 (see Hall and Rabushka, 1995). This means that raising $1 of revenue in the United States (so as to transfer it to some other group) costs the rest of the economy $1.30–$1.40. In the African context, the MEB of taxation is unlikely to be anything less than twice this figure, that is, 0.60–0.80. Consequently, the proposed compensatory transfer will entail an efficiency loss of at least 1.7–2.2 percent of GNP (0.028 multiplied by either 0.60 or 0.80). Therefore, compensation in this manner would eat up the bulk of the efficiency benefits that arise from the trade reform in the first place (which stood at 2.5 percent). And this before other losers (urban workers, recipients of trade rents) get their turn at compensation, and without taking into account the distortions that are likely to arise in subsidizing the losers.

In practice, there are two ways of getting out of this conundrum. One is to package the trade reform with other reforms that promise to provide substantial all-around gains to significant interest groups in urban and rural groups alike, and thereby dilute the redistributive effects of the former. Such opportunities rarely present themselves, because most reforms do have sharp distributional consequences. An exception is the situation that prevails following a prolonged period of economic decline and macroeconomic instability. There are few identifiable winners in an economy in near-hyperinflation, or where economic institutions and output have completely collapsed. The prospect of stabilization and recovery under such conditions, which would benefit most everyone, can allow trade reforms to be packaged along with the broader macroeconomic reforms. Consider, for example, the situation prevailing in Ghana during the early 1980s (Herbst, 1991):

Rent seekers who can control import licenses are usually a potent source of opposition to devaluation, but the crisis had become so bad in Ghana that the group benefiting from administrative allocation of foreign exchange was extremely limited. Indeed, by the early 1980s, the economy had deteriorated to such an extent that even senior government officials, who normally benefit from access to imported goods even in times of shortage, reported that they were going hungry and were concerned that they could not find food for their families.

From this perspective, it is no surprise that the most ambitious trade reforms in sub-Saharan Africa have been undertaken in countries like Ghana and Uganda where the previous economic decline was sharpest. Extraordinary times provide a window of opportunity for policymakers to undertake reforms that would be politically explosive in normal times.

The second strategy for dealing with redistributive conflict is to undertake partial, or two-track, reforms that preserve the privileges of the existing beneficiaries. This type of reform has been raised to an art form in China, where it has been systematically used to neutralize opposition from groups whose privileges would otherwise be threatened by market-oriented reforms. Hence, two-track pricing and incentive systems have operated in rural and urban areas of China, and in trade and investment regulations, apparently with considerable success.

In Africa, Mauritius provides a nice illustration of this strategy. This country is an African success story, despite its inauspicious beginnings. During the 1960s, Mauritius was a monocrop economy facing a population explosion. A report prepared by James Meade in 1961 was quite pessimistic about the island’s future: “unless resolute measures are taken to solve [the population problem],” the report stated, “Mauritius will be faced with a catastrophic situation” (Meade 1961, 37). To an important extent, the economy’s success was based on the creation of an export processing zone (EPZ) operating under free-trade principles, which allowed an export boom in garments to European markets.3 Yet the island’s economy has combined this EPZ with a domestic sector that was highly protected until the mid-1980s. Gulhati (1990, Table 2.10) reports an average effective rate of protection in 1982 for manufacturing in Mauritius of 89 percent, with a range of –24 to 824 percent (see also Milner and McKay, 1996, pp. 72–73). Hence, Mauritius is an example of an economy that has followed a two-track strategy.

The circumstances under which the Mauritian EPZ was set up (in 1970) are instructive. Here is how one account describes it (Alter, 1990, p. 4):

Given the small size of the domestic market and the negative experience elsewhere, import substitution was not regarded as a viable long-term strategy; therefore, as soon as import-substitution opportunities were exhausted, Mauritius switched to an export-oriented development policy, with the EPZ as the main element of its new industrial policy.

Were things so easy! As in other countries, policymakers in Mauritius had to contend with the import-substituting industrialists who had been propped up by the restrictive commercial policies of the 1960s. Under the development certificates (DC) scheme, local industrialists were provided with tax holidays and protection from imports via tariffs and QRs. A range of industries were set up using these incentives. These industrialists were naturally opposed to relaxing the trade regime.

The EPZ scheme provided a neat way around this difficulty. The point is made nicely in this account by Wellisz and Saw (1993, p. 242):

A completely outward reorientation was politically unfeasible in the 1970s—since protection was the key to the prosperity of the import-substituting industry and DC certificate holders constituted a powerful lobby. But the DC certificate holders were not disturbed by the formation of an export-oriented enclave: on the contrary, they welcomed it as another potential source of profits. Mauritian labor also favored economic segmentation: the high-wage sector—sugar and import-substituting industries—constituted a male enclave. The EPZ industries employed women, whose earnings supplemented family incomes and who did not compete with the men. For the exportoriented industries, too, the enclave solution had obvious advantages in that the quasi-extraterritorial status provided a degree of protection against the government’s dirigiste tendencies.

This passage illustrates the political advantages of the two-track strategy. The creation of the EPZ generated new opportunities of trade and of employment (for women), without taking protection away from the import-substituting groups and from privileged male workers. The segmentation of labor markets was particularly crucial, as it prevented the expansion of the EPZ from driving wages up in the rest of the economy, and thereby disadvantaging import-substituting industries.4 New profit opportunities were created at the margin, while leaving old opportunities undisturbed. There were no identifiable losers.

Time Inconsistency: Government Against Itself

Distributional considerations of the type discussed above explain a lot about the pattern of policymaking in Africa. But there are other pervasive features of African policy that cannot be explained without adding an additional layer of complexity on these distributional issues: Many policy distortions are excessive even from the standpoint of the purported beneficiaries of the distortions.

Trade policies, for example, are often too restrictive in that the rents they generate could be increased by relaxing them somewhat. This possibility is illustrated in one of the scenarios presented above in Table 2. We see that trade rents actually increase as t is reduced from 0.40 to 0.30. In other words, when t is set at 0.40, we are on the wrong side of the Laffer curve for trade revenues. This is not an outlandish possibility. With realistic import demand elasticities, and especially with the ever-present reality of smuggling factored in, the revenuemaximizing level of trade restrictions in much of Africa are unlikely to be above 40–50 percent. When we observe restrictions beyond that level, we must question the argument that these restrictions are in place simply because they enrich the government officials who administer them.5

This issue is particularly relevant to the taxation of export crops in many African countries. It is well recognized that export crops have been taxed at extremely high levels in Africa, with the result that export agriculture has often been decimated. According to estimates by Krueger, Schiff, and Valdés (1987), the direct taxation of agriculture—through marketing boards and low producer prices—has averaged 23 percent in Africa, compared with 2.5 percent in Asia and 6.4 percent in Latin America. When the effects of exchange rate overvaluation and import restrictions are added in, total taxation of agriculture more than doubles to 51.6 percent (Krueger, Schiff, and Valdés, 1987). These tax rates are a large part of the explanation for why Africa has consistently lost market share in world trade.

At least since Bates’s (1981) seminal work, the conventional explanation for these pricing policies has been the argument that African governments have used such policies to transfer income to politically powerful urban groups. Since import restrictions are functionally equivalent to export taxes, the logic is similar to that coming out of the model laid out in the previous section.

If this were the whole story, rates of direct taxation of agriculture would never exceed their revenue-maximizing levels. Yet the reality is different. According to calculations by McMillan (1997, pp. 2–3):

Even using the most conservative supply elasticity estimates … governments have taxed cocoa at a rate greater than the revenue maximizing tax rate 63% of the time in the 1970s and 25% of the time in the 1980s. Coffee has been overtaxed 35% of the time in the 1970s and 23% of the time in the 1980s and vanilla [has been] always overtaxed in both periods.

If the principal reason for taxing export agriculture is to raise revenues to be used on behalf of urban groups, why would governments impose taxes that exceed what is required to maximize revenue? More broadly, why would governments discriminate against their cash crops so badly that this sector would be devastated over time, depriving the government of revenues in the longer run? In other words, why would anyone want to kill the goose that lays the golden eggs?

McMillan (1997) provides a convincing explanation, which also provides a neat way of conceptualizing the institutional weakness of African states. She starts by distinguishing between planting and harvesting costs. The former are sunk costs. Therefore opportunistic governments have an incentive to cheat farmers out of these sunk costs by paying them the minimum required for them to bring their crops to market; that is, to pay them only their harvesting costs. Anticipating this eventuality, farmers of course will have no incentive to plant in the first place. Therefore, if the government is unable to precommit to an adequate price, the static, one-shot equilibrium in this game is one in which farmers withdraw from production and the government gets no revenue.

Now consider the more realistic case where the government and the farmers interact repeatedly. As McMillan notes, many more outcomes then become possible. She focusses on trigger-strategy equilibriums, under which farmers “punish” the government by not planting for k periods whenever the government deviates from the “first-best” pricing policy by offering a price below the full (planting plus harvesting) costs. McMillan shows that the condition under which the first-best strategy—government’s price covers the full cost, and farmers’ plant—can be sustained is given by the following expression:

The term on the left-hand side of the inequality represents the one-period gain to the government from deviating from the first-best strategy: the higher is the share of sunk costs in total costs, the lower is the price that the government can offer to the farmers and still get the crop to market, once farmers have already planted. The right-hand side represents the cost that the government bears when it deviates in this fashion. These costs are increasing in the future profit margin that the government will have to forsake when farmers stop planting for k periods. They also decrease with the government’s impatience. Hence, for the first-best strategy to be viable, the ratio of sunk to total costs must not be too high, future world prices must be anticipated to be favorable enough, and the government must not be too impatient.

McMillan confronts these implications with the evidence, using data on costs, prices, and taxes for six crops (cocoa, coffee, cotton, groundnuts, tobacco, and vanilla) drawn from more than 30 sub-Saharan African countries. She classifies her sample of crop-country pairs into subperiods of “high” and “low” taxation, on the basis of whether average (direct) tax rates in each subperiod fall below or above the estimated revenue-maximizing rates. Her results are striking (McMillan, 1997, p. 6):

Crops for which the ratio of sunk costs to total costs is relatively high, e.g. cocoa, coffee, and vanilla, have been taxed more heavily than crops with relatively lower ratios of sunk costs to total costs, e.g. cotton, groundnuts and tobacco. Using the probability of remaining in power as a proxy for the [political leader’s] discount factor… leaders with a relatively high probability of remaining in power tend to tax less heavily. And finally, using a twenty-year average of profits as a proxy for expected profitability of a crop … crops with [high] expected future profitability tend to be less heavily taxed.

These findings confirm the power of the underlying model of dynamic inconsistency in explaining patterns of policymaking across Africa.

This perspective adds a new dimension to the discussion of the politics of reform. It highlights the significance of credible commitments to policies, and the value of creating institutional settings that provide safeguards against opportunistic behavior on the part of state officials. This point is related to Collier’s (1995) discussion of the weakness of the “agencies of restraint” in Africa. As Collier explains, this weakness exhibits itself in all spheres of economic life in Africa: in intra-private transactions, where it shows up as weak contract enforcement and high transactions costs; in intra-public transactions, where it shows up as weakness of central banks and finance ministries relative to spending ministries; and in public-private transactions, as in the export taxation case discussed above.

It is instructive to examine the experience of one country where such problems have apparently not been as severe a problem as in most of Africa. The country in question is Botswana, one of the world’s fastest-growing countries. While Botswana has been blessed with diamonds, the presence of a rich natural-resource base must be seen at best as a permissive condition for its superlative economic performance. Too many countries in Africa with rich natural resources have been economic failures to attach much significance to diamonds per se. A lot of Botswana’s success has to do with its superior governance. The bureaucracy in Botswana is honest and competent, attaches great value to economic expertise, and has consistently produced sensible macroeconomic policies. There has been no large-scale urban bias and no white elephants (Harvey, 1992, p. 348). The government’s philosophy, however, has been far from laissez faire. One indication of this is that government expenditures stood above 50 percent of GDP by the early 1990s, one of the highest levels anywhere in the world. What has distinguished economic interventions in Botswana is its quality, not quantity.

Why this has been so is not altogether clear. The initial conditions, just as in Mauritius’s case, were not favorable. When it became independent in 1966, Botswana was one of the poorest countries in the world (Harvey, 1992, p. 338):

There was not even a capital city before independence; the country was administered from an enclave in Mafeking (now Mafikeng), in the Cape Town Province of South Africa. The education base was negligible. The only tarred roads consisted of a few miles in the towns. There was a railway, built for transit between South Africa and more prosperous colonies to the North, but nevertheless useful to Botswana, an abattoir for the export of beef, and not much else.

Furthermore, Botswana has been virtually surrounded with warfare and violence, as a consequence of wars of independence in Angola, Namibia, and Zimbabwe, and the struggle in South Africa.

One explanation that is often advanced is the rural origin of the political leadership. Harvey (1992, pp. 360–61), for example, emphasizes the

strong influence of rural exporters on economic policy. A large majority of politicians and senior government officials in Botswana own cattle, and an even higher proportion are related to people who own cattle. The income from cattle comes mostly from exporting.

This, it is argued, explains why policies in Botswana have not been anti-export, and why the economy has never been allowed to succumb to the Dutch disease. On closer look, this explanation is somewhat suspect. The urban origin of political leadership in other African countries can perhaps explain why agriculture was taxed; it cannot explain why it was typically taxed excessively, as discussed above. As McMillan puts it, in dismissing a similar explanation for Côte d’Ivoire’s relatively moderate taxation of agriculture (1997, p. 11):

This explanation is unsatisfactory because even elites who do not own farms may derive a large proportion of their income from the production of export crops in the forms of rent and taxation. Moreover, there are few African elites who do not own land or who could not own land if they wanted to.

The situation is analogous to Mancur Olson’s description of the stationary bandit: unlike the roving bandit, the stationary bandit has a stake in maintaining the long-run economic health of his tax base, even if he has no other ties to the society he rules. The social origin of the political elites cannot, in itself, explain why some governments have killed their cash cow while others have nurtured it.

I find an alternative hypothesis more appealing. Along with Lesotho, South Africa, and Swaziland, Botswana has long been a member of the Southern Africa Customs Union (SACU). This means that Botswana has no independent trade policy; goods circulate freely between it and South Africa. In addition, Botswana did not have a national currency and central bank until 1976 (Lewis, 1993, p. 19). The government gets a share of customs revenue collected by South Africa. These customs revenues amount to around 20 percent of the value of Botswana’s imports, which is high. What matters, from our perspective, however, is that government officials have no control over this revenue on a day-to-day basis; nor do they have an ability to interfere with the flow of goods from South Africa. Perhaps more to the point, domestic producers in the urban areas know that this is so, and therefore realize that lobbying policy makers for favors in the trade arena is a futile exercise. To someone used to reading horror stories arising from tariffs and NTBs in any developing country, opening a lengthy volume on Botswana (such as Harvey and Lewis, 1990) and not finding a long chapter on trade policy is an eye-opening experience. Absence of an independent trade policy is an extreme form of an “agency of restraint” (in Collier’s sense of the term).6

Could this externally imposed free trade regime be a key reason for Botswana’s success on the economic front? Obviously, the government’s ability to tax exports, either directly or indirectly, was sharply restricted. But beyond that, the absence of an import-substituting urban lobby—which the free trade regime ensured—could have led to improved governance on other fronts as well. For example, the admirable manner in which the government responded to a large drop in diamond earnings in 1981, by swiftly devaluing the currency and avoiding exchange controls (see Lewis, 1993, p. 19ff), may have been enabled by the absence of entrenched urban interests. Protected behind NTBs, these urban groups would have welcomed such controls and other trade restrictions, and would have made it more difficult for the government to undertake the requisite policy adjustments. That, in any case, was the fate of most countries in Africa (as well as in Latin America), which responded to external shocks by tightening trade restrictions. This is a hypothesis worth examining more closely.

Incomplete Information: The Private Sector against Itself

The frameworks considered up to this point are based on full and complete information. In particular, interest groups are well informed about the precise incidence of gains and losses that would follow from reform. This is a simplification, of course. Relaxing it provides additional insights on the political economy of reform.

The motivation for considering situations of incomplete information is that in reality it is difficult to predict with some certainty who the eventual beneficiaries of reform are going to be. In particular, it is reasonable to expect that while many import-substituting urban groups will be adversely affected, some among them may end up taking advantage of the new opportunities created by the reform. I cited above the examples of Korea and Taiwan Province of China. In both countries, exports were dominated by natural resources before the reforms of the early 1960s. Manufacturers were almost exclusively inward oriented. Yet, many of them discovered during the 1960s and 1970s that they could be successful exporters. Industry as a whole certainly did better under the outward-oriented set of policies than it had under the import-substituting policies of the 1950s. We can expect a similar set of outcomes in many African countries as well. Successful trade reform will spur nontraditional exports. But it is difficult to say ex ante what these exports might be, or who will actually produce them. This unpredictability generates a serious political problem: the politician needs the support of real individuals and groups, people one can actually name and collect money from. When beneficiaries exist only in an abstract, statistical sense, the reformer’s job is made much harder.

Consider the framework laid out in the paper by Fernandez and Rodrik (1991), and schematically shown in Figure 1. All individuals in the economy are aligned on the horizontal axis between 0 and 1. Look first at the top panel. Before reform, individuals in the interval (0, B) are assumed to be in the import-substituting sector, while individuals in the interval (B, 1) are in the exportable sector. After reform, some of the individuals who are in the import-substituting sector—those in the interval (A, B)—will move to exportables; they will be winners as well. Note that subsequent to reform, there will be a majority of gainers: those in the interval (A, 1). Therefore, if the precise distributional incidence of the reform was known in advance, a majority of the individuals would support it.7

Figure 1.Reform and Uncertain Benefits

But consider now the outcome if the aggregate consequences of reform are known, but the individual-specific outcomes are not. More precisely, suppose that individuals in the import-competing sector—those in the interval (0, B)—are ex ante all alike, and they do not know who among them will be able to move to exportables. Suppose further that the prospective gains or losses per individual are represented by the height of the boxes in the figure, and that individuals are risk neutral. The way the figure is drawn, all individuals in the import-competing sector are losers in expected-value terms. And since a majority of the population is in the import-competing sector initially, there will be only minority support for the reform. Hence the reform is rejected by a majority, even though a majority would benefit from the reform and it is common knowledge that it would do so. This is a case where incomplete information blocks a reform from being adopted even though the reform would be politically popular if imposed dictatorially.

This is not the only perverse possibility. The other case is illustrated in the bottom panel of Figure 1. Here, a majority—individuals in the interval (0, C)—would end up losing from reform. But in expected-value terms, the reform is a winner for all individuals in the import-competing sector initially—those in the interval (0, D). Consequently, there is unanimous support for reform ex ante, but there will be a significant backlash against reform once the distributional effects become clear for specific individuals.

One general implication of this line of reasoning is the following: over time, there will be a bias toward the status quo and against reforms that would prove ex post popular. This is illustrated in Figure 2. Consider two stages of voting. In the first, a reform is either adopted (Yes) or not adopted (No). Subsequently, there is a second vote that decides whether to continue with reform, to abort it, or, if the reform had not been adopted in the first vote, to do so now. There are four possible sequences of votes: Yes-Yes, Yes-No, No-Yes, and No-No. Of these, No-Yes is not an equilibrium outcome, since no new information is revealed when reform is not undertaken. If reform is rejected once, it will continue to be rejected in the future. Hence, reforms that are “mistakenly” adopted—mistakenly from a political perspective—will soon be reversed, while reforms that are “mistakenly” rejected will not get a second chance. There will be a systematic bias against reforms.

Figure 2.Status-Quo Bias: Voting on Reforms Over Time

A second implication is that there are circumstances when the promise of compensating the losers, even when economically feasible, will not be sufficient to eliminate resistance to reform. To see this, return to the case illustrated in the top panel of Figure 1. Remember that in this case, there will be majority support for reform if the reform is ever adapted, even absent compensation. What this implies is that a strategy of compensating losers is not time consistent for the reformers; why compensate the losers if the reform will stick even without their support? Of course, the losers can anticipate this result ex ante, and they will therefore not buy the promise in the first place.

This framework shows the importance of incomplete information and the dynamics of information revelation in shaping the configuration of interests for and against reform. It highlights the importance of identifying winners and reaching out to them. The task is especially difficult in sub-Saharan Africa, where informational incompleteness of the type discussed here is likely to be particularly severe. It is all the more urgent, for the same reason.

Concluding Remarks

The value of thinking analytically and coherently about the political economy of trade reform is not only that this helps us understand the sources of resistance to reform, but that it also helps us think creatively about ways of circumventing the resistance. Hence the difficulties identified in this chapter—the strength of the redistributive effects, the low credibility and commitment inherent in government policies, the uncertainty surrounding the identity of the gainers—are suggestive of broad strategies that can be employed to evade them. I have cited instances of such strategies above, using African and other examples. However, I have only scratched the surface. There is no substitute for creative political leadership in identifying and exploiting the opportunities that difficult times present.

The larger methodological point underlying this chapter is that advocacy of reform has to be complemented with a more sophisticated understanding of what blocks it. Stock phrases like “ownership of reform” or “commitment of the leadership” that one encounters in discussions of failed reform attempts do not begin to do justice to the real political challenges that reformers face. Strengthening the reformers’ hand requires not only technical economic advice, but also help with political strategies.


    AlterRolf G.1990“Export Processing Zones for Growth and Development: The Mauritian Example,” IMF Working Paper 90/122 (Washington: International Monetary FundDecember).

    BatesRobert1981Markets and States in Topical Africa (Berkeley: University of California Press).

    BienenHenry1991“The Politics of Trade Liberalization,” in Trade and Development in Sub-Saharan Africaed. byJ.H.Frimpong-AnsahS.M.R.Kanbur and P.Svedberg (Manchester: Manchester University Press).

    CollierPaul1995“The Marginalization of Africa,”International Labour ReviewVol. 134No. 4–5 pp 54157.

    DeanJudith M.SeemaDesai and JamesRiedel1994“Trade Policy Reform in Developing Countries since 1985: A Review of the Evidence,”Discussion Paper 267 (Washington: World Bank).

    FernandezRaquel and DaniRodrik1991“Resistance to Reform: Status Quo Bias in the Presence of Individual-Specific Uncertainty,”American Economic ReviewVol. 81No. 5 pp. 114655.

    GATT (General Agreement on Tariffs and Trade)1992Trade Policy Review: Ghana (Geneva, General Agreement on Tariffs and Trade).

    GulhatiRavi1990The Making of Economic Policy in Africa (Washington: World Bank).

    HallRobert E. and AlvinRabushka1995The Flat Tax2d ed. (Stanford, California: Hoover Institution Press, Stanford University).

    HarveyCharles1992“Botswana: Is the Miracle Over?”Journal of African EconomiesVol. 1No. 3 pp. 33568.

    HarveyCharles and Stephen R.LewisJr.1990Policy Choice and Development Performance in Botswana (London: Macmillan).

    HerbstJeffrey1991Exchange Rate Reform in Ghana: Strategy and Tactics (Washington: World Bank).

    IMF (International Monetary Fund)1997World Economic Outlook (Washington: International Monetary Fund).

    KruegerAnne O.MauriceSchiff and AlbertoValdés1988“Agricultural Incentives in Developing Countries: Measuring the Effect of Sectoral and Economywide Policies,”World Bank Economic ReviewVol. 2No. 3 pp. 25571.

    LallSanjaya and FrancesStewart1996“Trade and Industrial Policy in Africa,” in Agenda for Africa’s Economic Renewaled. byBennoNdulu and NicolasvandeWalle (New Brunswick, New Jersey: Transaction Publishers).

    LewisStephen R.Jr.1993“Policymaking and Economic Performance: Botswana in Comparative Perspective,” in Botswana: The Political Economy of Democratic Developmented. byStephenJohn Stedman (Boulder, Colorado: Lynne Rienner Publishers).

    McMillanMargaret1997“A Dynamic Theory of Primary Export Taxation: Evidence from Sub-Saharan Africa” (unpublished; New York: Columbia University).

    MeadeJ.E.1961The Economics and Social Structure of Mauritius—Report to the Government of Mauritius (London: Methuen).

    MetzelJeffrey with LucieColvin Phillips1997“Bringing Down Barriers to Trade: The Experience of Trade Policy Reform” (Washington: U.S. Agency for International Development).

    MilnerChris and AndrewMcKay1996“Real Exchange Rate Measures of Trade Liberalization: Some Evidence for Mauritius,”Journal of African EconomiesVol. 5No. 1 pp. 6991.

    RodrikDani1994“The Rush to Free Trade in the Developing World: Why So Late? Why Now? Will It Last?” in Voting for Reform: Democracy Political Liberalization and Economic Adjustmented. byS.Haggard and S.Webb (New York: Oxford University Press).

    SachsJeffrey and AndrewWarner1995“Economic Reform and the Process of Global Integration,”Brookings Papers on Economic Activity: 1 (Washington: Brookings Institution).

    SvedbergPeter1991“The Export Performance of Sub-Saharan Africa,” in Trade and Development in Sub-Saharan Africaed. byJ.H.Frimpong-AnsahS.M.R.Kanbur and P.Svedberg (Manchester: Manchester University Press).

    WelliszStanislaw and Philippe Lam ShinSaw1993“Mauritius,” in The Political Economy of Poverty Equity and Growth: Five Open Economiesed. byRonaldFindlay and StanislawWellisz (New York: Oxford University Press).

    World Bank1994Adjustment in Africa: Reforms Results and the Road Ahead (New York: Oxford University Press).

    YeatsAlexander J.1997“Did Domestic Policies Marginalize Africa in International Trade?” (unpublished; Washington: World Bank).

There is a related debate about the sources of the relative economic decline of Africa during the past three decades, especially as regards its trade performance. Most Western analyses lay the blame on African countries’ policies, rather than on the policies of Western nations or on underlying trends in world markets. See Yeats (1997) and Svedberg (1991).

Consider, for example, the well-known paper by Sachs and Warner (1995). In this paper, the authors showed that countries classified as “open” grew at significantly higher rates than countries classified as “closed.” Five indicators were used for purposes of classification (tariffs, NTBs, black market premiums, socialist regime, and export marketing boards), with failure to pass any of the thresholds associated with these five indicators resulting in classification as “closed.” The threshold for tariffs was set at 40 percent, while the threshold for NTBs was set at 40 percent coverage. What is noteworthy is how permissive these levels are, even by African standards. In fact, only five sub-Saharan African countries fail to pass either one of these two tests (according to the Sachs-Warner data): Burkina Faso, Malawi, Nigeria, Rwanda, and Zimbabwe. (The bulk of African countries were classified as “closed” on the basis of the other three criteria.)

The full story is of course more complicated than that. There were highly profitable sugar exports, thanks to a generous quota in the European market. The EPZ appears to have been spurred, in its initial stages at least, by local capital and domestic investments. Profits from the sugar trade appear to have been the source of the savings that financed early growth in the EPZ.

This segmentation lasted until the mid-1980s. According to Wellisz and Saw, “As of 1985, the minimum wage for male workers ceased to apply to EPZ enterprises” (1993, p. 248), after which the EPZ began to compete for male workers with the sheltered parts of the economy, and the share of male workers in the EPZ rose rapidly.

One caveat is that such “inefficient” outcomes can arise when policymakers do not coordinate in setting the level of trade restrictions. With each official having decisionmaking authority over part of the trade apparatus, the Nash equilibrium of the game will typically entail too restrictive a trade regime overall and too little revenues for the officialdom in the aggregate.

As a small country in SACU, Botswana was essentially forced to inherit South Africa’s relative price structure. Its gains from trade derived from the difference between this relative-price structure and that which would have obtained under autarky in Botswana. The fact that the external tariffs in SACU were fairly high—and that South Africa’s relative structure was distorted relative to the rest of the world—is largely irrelevant to the existence of gains from trade for Botswana. To the extent, however, that the external tariffs in SACU pushed South Africa’s relative prices in the direction of Botswana’s autarky price ratio, these external tariffs reduced Botswana’s gains from trade (relative to free trade outside the SACU structure).

Although I will use the language of majority voting for the purposes of this discussion, the ideas are more general. One could think of decisions being made by leading interest groups instead, with the support of a plurality or majority of groups needed for the reform to be undertaken.

    Other Resources Citing This Publication