Simon Evenett
Rutgers University, Brookings Institution and CEPR
evenett@fas-econ.rutgers.edu, sevenett@brook.edu

and

Bernard Hoekman
World Bank and CEPR
1818 H Street NW, Washington DC 20433, USA.
Bhoekman@worldbank.org

June 7, 1999

Abstract: This paper analyzes the impact on national welfare and market access of removing discriminatory government procurement policies, focusing in particular on the importance of free entry and exit, and distinguishing between purchases of tradable goods and nontradable services. The nature of the distortions created by a nation’s procurement policy depends importantly on the strength of the nation’s competition policy. In many circumstances, a strong competition regime implies that the effects of procurement discrimination cannot be eliminated by liberalization.

The absence of general rules on public purchasing practices are one of the few major "holes" in the WTO edifice of trade policy disciplines. In the coming round on multilateral negotiations, developing countries can expect to be confronted with substantial pressure to accede to the relevant WTO Government Procurement Agreement (GPA), which to date has been a voluntary instrument. This paper analyzes the impact on national welfare and market access of removing discriminatory government procurement policies, focusing in particular on the importance of distinguishing between cases when competitive firms can and cannot freely enter and exit markets. Since this is in part determined by the competition policies governments pursue, including their foreign direct investment (FDI) regimes, our analysis illustrates that the impact of procurement favoritism is very sensitive to these ancillary policies.

A seminal 1972 paper by Robert Baldwin and J. David Richardson provides our point of departure. We modify their analysis to consider a competitive small open economy where world prices determine initial long run equilibrium prices. In the absence of free entry we duplicate their main findings: (i) if government demand is less than the supply of the domestic industry at free trade prices, a procurement ban has no impact on prices, net imports or national welfare; and (ii) only when government demand exceeds domestic supply at free trade prices does a procurement ban bid up prices paid by the government and reduce imports and national welfare. When free entry is allowed, the second finding requires modification. By raising prices above their initial long run level, the procurement ban creates short run rents for incumbent firms. These induce entry, expanding supply until the price paid by the government falls to the world level and rents are eliminated. Thus, in the long run free entry eliminates the price distortion created by the procurement ban. Therefore, the very existence of the distortions created by a nation’s procurement policy depends on the strength of its competition policy regime.

Whether firms can freely enter at the time the discriminatory procurement regime is imposed has important implications for the effect of subsequently liberalizing the procurement regime. Without free entry, the removal of discriminatory procurement restores the long run equilibrium prices and level of imports that prevailed before the discrimination was imposed. However, in industries with free entry this need not be the case. If the policy raises short run prices and creates rents, entry continues (and industry expands) until the long run equilibrium price returns to the level that prevailed before the policy was imposed. In this case the elimination of the policy provides neither the government nor domestic firms with an incentive to change their plans. With a strong competition policy that fosters free entry, procurement liberalization does not alter domestic industry output or increase imports: the damage done (to foreign competitors) by discriminatory procurement policies cannot be undone by their removal.

GPA members are required to report data on the share of procurement sourced from national firms. We argue that this type of information is not very useful for surveillance purposes. However traditional tools for measuring the incidence of non-tariff barriers are useful, e.g., the difference between the price paid by the domestic government for a good and the world price. The existence of a price gap indicates a welfare-reducing procurement policy, and declining price gaps over time can indicate that the procurement policy is becoming less distortive. Furthermore, when a discriminatory procurement policy is in place, the share of domestic industry sales accounted for by government purchases cannot be used to predict the effects of removing a discriminatory procurement policy.

We conclude that the focus of surveillance efforts in the WTO should center on the level of market access barriers that prevail in sectors where procuring entities have an incentive to source locally, i.e., services and situations where monitoring contract compliance is difficult. More generally, the strength of the competition policy stance that prevails should be assessed, not the endogenous price and quantity responses to procurement liberalization.

With the conclusion of the Uruguay Round negotiations and the creation of the World Trade Organization (WTO), a large number of multilateral disciplines were established that apply to all WTO members. The big exception is government procurement, where signing on to the principles of nondiscrimination remains voluntary. The absence of general trade policy disciplines on public purchasing practices is now one of the major remaining "holes" in the WTO edifice. In the coming years, developing countries can expect to be confronted with substantial pressure to accede to the WTO Agreement on Government Procurement (GPA) or to accept the multilateralization of this instrument. The United States is playing a lead role in this connection. US legislation requires the United States Trade Representative (USTR) to monitor foreign procurement policies that deny access to markets for American goods and services, and procurement policies figure prominently in USTR's annual Foreign Trade Barriers Report. The US has made public procurement a priority trade policy issue, linking this to the broader issue of corruption.¹ At the December 1996 Ministerial meeting of the World Trade Organization (WTO) in Singapore it was agreed to establish a Working Group with the mandate to conduct a study on transparency in government procurement practices. This is to take into account national policies and develop "elements for inclusion in an appropriate agreement" (WTO, 1996, p. 6). [add sentence on where this stands]

In determining how to respond to US pressure, developing countries need to determine the costs and benefits of GPA membership (or acceptance of equivalent disciplines). Conversely, in deciding how much negotiating effort to devote to this issue, "demandeur" countries should ascertain whether discriminatory procurement policies cause significant economic distortions. The aim of this paper is to analyze the impact on national welfare and market access of removing discriminatory government procurement policies, distinguishing in particular between cases when competitive firms can and cannot freely enter and exit markets. Since ease of entry and exit are determined in part by the competition policies that governments pursue, including their foreign direct investment (FDI) regimes, our analysis illustrates how the effect of procurement favoritism depends critically on these ancillary policies.

Whether firms can freely enter at the time the discriminatory procurement regime is imposed has important implications for the effect of liberalizing procurement regimes. Without free entry, the removal of procurement discrimination restores the long run equilibrium prices and level of imports that prevailed before the discrimination was imposed. However, in industries with free entry this will not be the case. If the policy raises short run prices and creates rents, entry occurs until the long run equilibrium price returns to the level that prevailed before the policy was imposed. Elimination of the policy provides neither the government nor domestic firms with an incentive to change their plans, as the government cannot find cheaper suppliers overseas and firms cannot find customers willing to pay more than the price they currently charge the government. Overall then, with a strong competition policy that fosters free entry, procurement liberalization does not eliminate the resource allocation distortion or increase imports: the damage done by discriminatory procurement policies cannot be undone by liberalization. An important asymmetry exists.

Our analysis also sheds light on the accuracy of certain observable price and quantity measures of the distortive impact of procurement regimes. Price gaps have been a traditional tool for measuring the incidence of non-tariff barriers (Baldwin 1989, Feenstra 1995), and in this case equal the difference between the price paid by the domestic government for a good and the world price. Although a price gap indicates a welfare-reducing procurement policy, the absence of a price gap cannot distinguish between two cases when the procurement policy: was never distortive or is no longer distortive. Furthermore, the difference between the quantity supplied by the domestic industry and the quantity demanded by the government, which in the original Baldwin and Richardson (1972) analysis was found to have predictive power before a procurement policy was imposed, is shown to have no predictive power when the procurement policy is actually in place. This calls into question the utility of empirical analyses which employ the Baldwin and Richardson criterion in samples where procurement policies are already in place. More constructively, we show how price gaps can be used to predict the effect of eliminating a procurement ban or a price preference policy.

The remainder of this paper is organized as follows. After a brief survey of some of the relevant literature (Section 1), we present an analysis of the impact of bans of government purchases from foreign suppliers, highlighting the similarities and departures from the original Baldwin and Richardson model (Section 2), and demonstrate that the key findings carry over to price preference polices (Section 3). The analysis then turns to the case of nontradables and goods where purchasers are likely to have a strong preference for being in close geographic proximity to suppliers (Section 4). The implications of these analysis for measuring the impact of procurement policies is taken up in Section 5. Conclusions, caveats and implications for the political economy of procurement liberalization are presented in Section 6.

A major discipline imposed by the GPA is nondiscrimination. Foreign firms are to be treated identically to domestic firms in the procurement process. Although intuitively this rule appears to be unambiguously welfare improving, McAfee and McMillan (1989) have noted that discriminating against foreign bidders may be in the national interest if domestic firms have a cost disadvantage in producing the products to be procured and only a limited number of firms (foreign and domestic) bid for the contract. A policy that gives preferences to domestic firms may then induce foreign firms to lower their bids. In effect preferences act as a profit-shifting device, ensuring that procurement favoritism increases national welfare. Branco (1994) has shown that even if the cost structure of domestic and foreign firms are identical and account is taken of the social cost of distortionary taxation, discrimination may be rational simply because foreign profits do not enter into domestic welfare. In the small numbers context assumed by these models prices will exceed marginal costs, so that shifting demand to domestic firms may also reduce price-cost margins as domestic output expands (Chen, 1995). Laffont and Tirole (1991) demonstrate that favoritism in procurement may be welfare improving from a national perspective because of the greater ease (lower cost) of collecting side payments from domestic firms.

Such economic arguments for preferences hold in small numbers/imperfectly competitive settings where there are profits and rents to be shifted (Mattoo, 1997). Even if there are many potential suppliers, discrimination may be beneficial to the procuring entity if the products to be procured are intangible or there are monitoring or contract compliance problems. In such cases, if procuring entities restrict competition and pay quasi-rents to contractors, the likelihood of performance could be increase by creating opportunity costs--the threat of losing future repeat business (Rotemberg, 1993). Such situations are perhaps most likely to arise in the procurement of services or services-intensive products. Given that services are often the largest category of purchases by governments (Francois, Nelson and Palmeter, 1997)--increasingly so in countries that have been pursuing outsourcing and contracting--this has potentially far-reaching implications for the assessment and design on multilateral disciplines for procurement practices. In particular, they suggest that geographic proximity may be a precondition for contesting procurement markets--i.e., the products involved are effectively nontradable. Problems of asymmetric information and contract compliance may imply that entities have a natural preference to choose suppliers that are located within their jurisdictions because this reduces monitoring costs (Breton and Salmon, 1995). Such proximity incentives will make it more difficult for foreign firms to bid successfully, even if the goods involved are tradable. Indeed, such incentives may make policies which discriminate at the border redundant.

The issues that arise from a WTO perspective in this connection are clearcut: whether there are barriers to entry through FDI; and how entities decide whether suppliers are local "enough"--i.e., the rules of origin for legal persons. More generally, an implication is that the competition policy stance of a government will be an important determinant of the impact of procurement policies, as one role this should play is to ensure that "small numbers" situations are not the result of artificial entry barriers.

While there are a variety of situations under which discriminatory procurement may enhance national welfare and lower procurement costs, simulation studies suggest that welfare gains are likely to be modest at best. Greater profits of domestic firms will tend to be offset by increased average prices paid by public entities. As a result, the potential welfare gains are reduced (Deltas and Evenett, 1997). Given that in most instances the optimal policy will be difficult to determine and generally will vary depending on the specifics of the situation, in practice favoritism can be expected to be more costly than a policy of nondiscrimination. Frequently the information required to judge if diverging from nondiscrimination is beneficial will not be available; nondiscrimination is therefore a good rule of thumb. The case for nondiscrimination is strengthened if account is taken of the rent seeking distortions that may be induced by discriminatory policies and the social cost of corruption/bribery. Nondiscrimination will generally reduce discretion and enhance transparency of the procurement process and thus reduce the scope for rent-seeking. Indeed, transparency arising from the procedural requirements of the GPA may well be the primary benefit of membership for developing countries--even if a strict nondiscrimination is not pursued (Hoekman and Mavroidis, 1997).

The economic analysis presented in this paper abstracts from considerations of transparency and corruption. Instead, the focus is on the long run impact of procurement favoritism. The focus on the long run reflects the fact that many nations’ procurement policies have been in place for decades, and that any rents that have been created may well have been eroded by entry of new domestic, and possibly foreign, firms. In conjunction with empirical studies that suggest price-cost markups tend to fall rapidly once the number of firms in a industry rises to more than [five? (Schmalensee/Bresnahan)?] and given that most developing country governments will be able to source from a large number of firms located anywhere in the world, this provides further justification for adopting a competitive market structure.

In an influential paper, Baldwin and Richardson (1972) analyzed the impact on imports, prices and national welfare of a ban on government purchases of foreign suppliers using a partial equilibrium framework. Their two main findings were that: (i) if government demand is less than domestic supply at free trade prices, a procurement ban has no effect on domestic prices, net imports and national welfare; and (ii) when government demand is greater than domestic supply at free trade prices, a procurement ban raises the price paid by the government and domestic output, reduces total imports and national welfare.

Since their analysis was motivated by the United States "Buy America" Act, they assumed consumers (state and private) faced upward sloping supply curves. This large country assumption is inappropriate for most developing nations, and in what follows we assume the procuring nation faces an infinitely elastic foreign supply curve at a given world price. In contrast to Baldwin and Richardson we also distinguish between short run equilibrium (where the number of firms are fixed) and long run equilibrium, by allowing for free entry into industries which make excess profits in the short run. To pin down the long run equilibrium price we assume that factor prices and the firm’s technology are such that the representative firm has a "U" shaped average cost curve. Our third modification is to assume that the law of demand applies to government purchases (Baldwin and Richardson assume government demand is price inelastic).

As the framework used is quite simple, the analysis that follows uses a standard supply and demand framework, where PW is the world price of the good; PC is the unit price paid by the domestic consumer; PG is the unit price paid by the domestic government; D_G(p) is the demand schedule of the domestic government; D_T(p) is total domestic demand; S_H(p) is the short run supply schedule of the domestic industry; LS_H(p) denotes is the long run domestic supply curve; ATC and MC are the average total and marginal cost schedules of the representative firm in the domestic industry, respectively; and  is the short run profit of the representative domestic firm.

We assume that initially free trade prevails. There are no tariffs or other impediments to trade. Each domestic firm has a U shaped average cost curve that reaches a minimum at level C*. The market is perfectly competitive, with domestic and foreign firms producing perfect substitutes. There is free entry and exit in the long run. In the short run the number of firms in the domestic industry is fixed. In the absence of a ban on government procurement from foreign entities, the long run equilibrium is portrayed in Figure 1.

Free entry ensures that profits p) are zero in the long run, and that P_W = C*. Free trade ensures that domestic consumers and the domestic governments pay the world price, P_W = P_C = P_G. In the case shown in Figure 1, at price P_W the domestic industry supplies less than total domestic demand. Imports make up the difference. To demonstrate that Baldwin and Richardson’s two findings hold in this model in the short run, consider the impact of a ban on state procurement from foreign suppliers. This has no effect on equilibrium prices or imports and quantity traded: the ban results in every foreign producer who previously supplied the government finding a domestic consumer to supply at world prices; every domestic consumer, abandoned by the domestic firms who now supply the government, finds their needs met by foreign suppliers. Domestic suppliers are unable to raise their prices to the government as other domestic suppliers are willing to step in at the previous world price.

Figure 2 portrays the case when government demand exceeds domestic supply at free trade prices. In this case a procurement ban implies that the supply decisions of the domestic industry can only be reconciled with the government demand schedule at a higher price, P_G. Since P_G > P_W, overall government purchases fall, domestic output rises and the representative domestic firm makes positive profits. Domestic private sector consumers can still source from abroad, and do so, as foreign suppliers are willing to sell the good at a lower price than domestic suppliers. Overall imports fall from AC to BC. The procurement ban has an adverse effect on national welfare, creating consumption and production distortions equal to a deadweight loss of area ADB. Thus our framework replicates Baldwin and Richardson’s findings. Finally, domestic private sector consumer choices are unaffected by the procurement ban, and thus their welfare does not change.

Allowing free entry of firms does not alter the analysis in Figure 1. As the imposition of the procurement ban does not raise domestic firm profits, it does not induce entry. In contrast, when government demand is larger than domestic industry supply at free trade prices (as portrayed in Figure 2), then the procurement ban raises profits in the short run. Domestic firms will enter, expanding total industry supply and forcing prices down until profits are eliminated at P_G = P_W. The long run supply curve LS_H(p) is horizontal, which along with the government’s demand schedule D_G(p), implies that point B, and any point to the right of point B, can be a long run equilibrium. Under free entry, a procurement ban is only distortionary in the short run. Consider a candidate long run equilibrium at point B. Since P_G = P_W there is no consumption distortion and, as production takes place at the lowest possible production cost in the long run, there is no resource allocation distortion either.

Thus, the welfare consequences of a procurement ban depend not only on the relative size of government demand and domestic industry output at the free trade prices, as Baldwin and Richardson originally posited, but also on the strength of a nation’s competition policy as reflected in the ease of entry and exit. The effects of imposing a procurement ban are summarized in Table 1.

The impact of removing a procurement ban are also sensitive to entry. Suppose the market is in a long run equilibrium and the ban is eliminated. In the case of Figure 1 nothing changes: the government need no longer buy from domestic firms, but has no incentive to switch to foreign firms as P_G = P_W. In the case of Figure 2 with no entry, the pre-liberalization long run equilibrium is point D, where P_G > P_W. Removing the procurement ban would shift the equilibrium back to point A, reducing the domestic industry’s output, eliminating the rents and the consumption distortion and increasing imports from BC to AC. Procurement liberalization would increase imports, eliminate the distortions created by the ban, and so improve national welfare. This conclusion does not carry over to the free entry case. If point B is the pre-liberalization long run equilibrium, removal of the ban does not alter the long run equilibrium price: every domestic supplier can find a buyer at this price, and every domestic buyer can find a domestic or foreign supplier. No one has an incentive to alter their plans, and removal of the procurement ban will not change the level of imports or the output of the domestic industry; an asymmetry exists.

In practice governments often employ price preferences to discriminate against foreign suppliers, not outright bans. That is, the government evaluates the supply price of a foreign seller at 100(1+ρ)% of its actual supply price (where ρ>0 is the margin of preference). If the foreign bidder’s inflated price is below that of domestic firms, the government purchases from the foreign seller at its actual (not its inflated) supply price. Otherwise, it buys from the domestic seller. The key finding in this section is that the principal effects of procurement bans carry over to price preferences.

The analogue to Figure 1 is represented in Figure 3: at free trade prices the quantity demanded by the government equals or is less than the quantity supplied by the domestic industry. Irrespective of the size of the parameter ρ, the price preference policy results in the same reallocation observed under the ban. Government purchasers evaluate the price of foreign suppliers as (1+ρ) P_W, domestic suppliers as P_W. Switching their purchases to domestic suppliers has no effect on prices, imports, domestic output and national welfare. Similarly, removal of the preference policy has no effect.

When government demand exceeds domestic supply at free trade prices, two cases must be distinguished. The first is portrayed in Figure 4, where the percentage difference in prices between PW and PG is less than the price preference. Starting from the original long run equilibrium at point A, the preference policy effectively "prices" the foreign supplier out of the market since their supply price to the government, rises to P_W(1+ρ). The government’s purchasing plans can only be met by domestic suppliers at a higher price P_G. In the absence of entry, point D is both the short and long run equilibrium. Similar to a ban, in the absence of free entry a price preference policy creates both a long run consumption and resource allocation distortion. Likewise, removal of the price preference policy would shift the equilibrium back to B, eliminating the distortions and raising welfare.

However, if we allow for free entry, the rents created by the preference policy in the short run attracts firms into the industry. Analogous to Figure 2, any point on the long run supply curve LS_H to the right of point B can be a long run equilibrium. In any of these long run equilibria the price preference policy neither distorts resource allocation distortion or consumer choices. In addition, elimination of the preference policy has no impact on prices, domestic industry output, net imports and welfare. The reduction in imports caused by imposing a price preference policy cannot be reversed by its removal.

The second case is portrayed in Figure 5: here the price preference parameter, ρ (expressed as a percentage) is less than the percentage increase in price that would result if only domestic suppliers were allowed to sell to the domestic government. Thus, ρ is low enough so that the policy does not effectively “price” foreign suppliers out of the market. Any domestic firm that attempts to price above PW (1+ρ) will find itself with no customers. Consequently, in the short run prices rise to PW (1+ρ), and the government still imports some foreign goods (equal to amount DE). Again the price preference policy creates rents for incumbent firms, a consumption distortion and a resource allocation distortion. Without free entry, the long run equilibrium remains at point D and imports continue to be observed. Of course, elimination of this preference policy shifts the long run equilibrium from point D to A, increasing national welfare and imports.

Allowing free entry in Figure 5 has implications not found hitherto. Domestic firms enter until prices fall to P_w and point B, or any point to the right of point B, on the long run supply curve LS_H. However, the additional domestic output reduces long run equilibrium imports to zero. Thus, price preference policies need not eliminate imports in the short run (as a procurement ban would), but with free entry will eliminate imports in the long run. This implies that even a small price preference can create a large long run increase in industry output and corresponding reduction in imports. When the new long run equilibrium has been established after the introduction of a price preference policy, removal of that policy will have no effect on the long run equilibrium price. Once imposed, free entry ensures that any effects of the price preferences cannot be eliminated by removing that policy.

The impact of a price preference policy is summarized in Table 2. Again, determining the long run impact requires some information or priors on the existence of free entry in the industry concerned. Unfortunately in the case where government demand exceeds domestic industry supply at free trade prices, one cannot observe whether the price preference parameter is such that Figure 4 or Figure 5 characterizes the short run and long run equilibria. Consequently, in this case, we cannot predict whether or not imports will collapse to zero in the short run.

As was discussed in Section 1, there are circumstances where it is welfare maximizing for procuring entities to favor sourcing from "local" firms, even if the products concerned are tradable. Of course, in practice much of what is procured by governments is not tradable, as the majority of the products purchased are services. Francois et al. (1996) note that in 1993 purchases by federal, state, and local authorities in the U.S. exceeded US$ 1.4 trillion, equivalent to some 20 percent of GDP. Out of this, federal procurement was $445 billion, of which 68 percent was spent on defense (goods, services, and employee compensation). Of the remaining $141 billion, worker compensation comprised 48 percent, leaving $73 billion. Most of this was used to procure services. Goods accounted for only $14.4 billion, or 3 percent of total federal purchases. At the state and local level the dominant category of expenditure (excluding wages) are structures (construction activities).

Services are therefore to a large extent where the action is in procurement markets. Consequently a government’s FDI policy stance, and more generally competition/regulatory policies, becomes an important determinant of the impact of discriminatory procurement policies for services, as establishment is often a precondition for contesting the market. A key distinction that arises in evaluating policies that affect services procurement is that there will be no world price to pin down the analysis. National markets will be segmented, independent of procurement regimes, as prices will depend to a large extent on (local) factor costs. However, in the long run, zero profits ensures that prices are set at minimum average total costs. Thus, the assumption of U shaped cost curves pins down the long run prices.

In the context of services there are many forms of procurement discrimination possible (Hoekman and Primo Braga, 1997). Here we consider the impact of a ban on government purchases from subsidiaries of foreign firms that have established a local presence. Assume that both home and foreign-owned firms have access to the same technology and confront the same factor prices. Figure 6 portrays the long run equilibrium in the absence of the procurement ban: the prevailing price C* is such that government demand is less than home firms supply, i.e., S_H(C*) > D_G(C*). Again, the imposition of the ban merely reallocates customers, with each foreign affiliate able to find demand from the home private sector to exactly offset the loss of sales to the government. The removal of the ban has no effect on equilibrium prices, home firms output and national welfare.²

The relationship between government demand and the amount supplied by home firms at the long run equilibrium price determines whether the procurement ban is distortive. The nature of any distortion depends on the strength of the nation’s competition policy (Figure 7). Suppose at the initial long run equilibrium price C* (the minimum cost of production) government demand exceeds home supply, so that some foreign affiliates supply the government. A procurement ban then acts to segment the market: home firms supply the government at price P', which is higher than C*. The foreign subsidiaries are left supplying only the private sector at a lower price P'', reflecting the fact that home private consumers are unwilling to buy all of their output at the initial price C*. The short run impact of imposing a procurement ban is therefore to introduce two consumption and two production distortions. Unlike the trade in goods case, the ban affects (actually raises) home consumers’ welfare.

This short run equilibrium will not persist as foreign firms are making losses, and some will exit until the price paid by home consumers rises to C*, where foreign subsidiaries break even. In Figure 7 this implies that the home consumers demand returns to point D from point B.

Home firms are making positive profits at the short run equilibrium price P'. If entry is permitted, this will force down the price paid by the government until firms make zero profits (at price C* and government demand at point C). Thus the long rum effect of the ban is to attract additional resources into the industry, increasing the number of domestic and reducing the number of foreign firms. In contrast to the case of tradeable goods and services, both free entry and exit are needed to eliminate the consumption and resource allocation distortion in the long run. Worse still, elimination of the ban has no effect on the equilibrium price and so cannot return the market to the pre-discrimination outcome. Table 3 summarizes our results, reinforcing the central role the strength of a nation’s competition regime plays in determining the long run impact of discriminatory procurement policies that operate within, rather than at, the border.³

The framework laid out above has a number of implications for predicting the impact of removing discriminatory procurement policies; the conduct and interpretation of empirical studies of the effect of procurement favoritism; and the measures used to evaluate compliance with the WTO Agreement on Government Procurement (GPA).

The central point to bear in mind is that the observed price and quantities (market outcomes) may well have been altered by the imposition of a discriminatory procurement regime. For the case of a procurement ban on trade in goods, Table 4 summarizes what can be learnt from observed market outcomes about the distortive effect of the policy and the effect of removing it. Of particular interest is the informational content provided by "price gaps", the traditional tool to measure the impact of a non-tariff barrier (see Deardorff and Stern, 1998). While a price gap (P_G > P_W) is evidence that the ban is at present distortive, the absence of a price gap does not imply the procurement ban was always welfare neutral. More constructively, the existence of a price gap does allow a clear prediction to made concerning the effects of removing the ban.

Once a procurement ban is in place, the difference between domestic industry output and observed government purchases need not the effect of removing the procurement ban or its distortive effect. This is in stark contrast with the Baldwin and Richardson (1972) finding that at free trade prices this difference determines the effect of introducing a procurement ban. In sum, an evaluation of the distortive impact of a procurement ban which is based on observed market outcomes must take into account the fact that the presence of the policy has altered those very outcomes. (Appendix Table 1 presents very similar conclusions for the case of a price preference policy).

Empirical studies analyzing the quantity impact of government procurement practices in OECD countries have tended to use the methodology suggested by Baldwin (1970), the basis of which is to suppose that in the absence of discriminatory policy, government entities propensity to import would equal those of private firms. Thus, under nondiscrimination the government's demand for imports of a good i as a share of its total consumption (m_i) would equal that of the private sector as a whole. If we denote total government demand as G_i, hypothetical public sector imports would be m_iG_i, and if the government does not collect tariffs on its own imports, the impact of a discriminatory procurement policy can be estimated as:

where the assumed non-discriminatory level of imports is using the private sector import share is m_i^ps G_i, M_i^G are actual government imports, j is the price elasticity of demand for imports, and P_i is the implied preference margin for good i.

Using 1958 data for the U.S. and an elasticity of -2, Baldwin (1970) estimated that the preference margin was some 20 percent. After adjusting for the fact that certain large import items such as oil were not subject to discriminatory policies, the margin for the residual set of covered goods increased to some 40 percent. More recent estimates of preference margins in 1992 for a number of countries using the Baldwin method give an estimated preference margin for U.S. purchases of 16.3 percent (Francois, Palmeter and Nelson, 1997). This suggests there has been a small decline in preference margins. If preference margin calculations are done on a sectoral level, positive margins in OECD countries may be as high as 50 percent, but in many cases are negative (Table 5). Margins are invariably the highest for procurement of services.

Note: -- denotes a preference margin that is less than or equal to zero.
Source: Francois, Nelson and Palmeter, 1997.

The flaw in this approach is the assumed counterfactual level of imports that prevails in the absence of the procurement policy. Only under certain conditions will the level of imports change when the procurement policy is removed, restricting the applicability of the technique. Note that those conditions would also generate a "price gap", which is somewhat ironic, for if a price gap were observed directly then there is no need to calculate the implied preference margin. Similar criticisms can be made of those empirical studies that employ gravity equations to forecast what government imports "should be" in the absence of procurement discrimination (e.g., Evenett, 1998; Trionfetti, 1997). For a wide range of circumstances (see Tables 1-3) the long run impact of procurement liberalization on imports is zero.

At present, members of the GPA, which entered into force in 1981, report information on the share of procurement that goes to domestic firms (which in practice includes sales of foreign affiliates); the share of procurement that is below a minimum threshold value; and the share of contracts that are awarded directly without competitive bidding. Our analysis suggests that not much can be learnt from these data. While increases in the share of government procurement that are tendered selectively or fall below a certain threshold may be indicative of efforts to avoid GPA disciplines, in the short run only if government demand is larger than domestic industry output will such efforts create distortions. Furthermore, in the case of goods under many circumstances the share of government purchases supplied by domestic firms will not change after liberalization. It is therefore inappropriate to conclude that a country has failed to implement the GPA just because this share has not fallen. Likewise, if a nation eliminates discrimination against foreign affiliates, this need not result in any new sales to these firms. Thus, existing GPA surveillance attempts are unlikely to shed much light on compliance.

After joining the GPA, in the EU, Japan and the U.S. the very high share of domestic firms in total procurement by covered entities remained virtually unchanged, while for small countries the share of procurement from national sources declined over time (Hoekman, 1998). This cannot be interpreted as the large countries having been able to "avoid" implementing the GPA, as should be clear from the forgoing analysis. In some small countries such as Singapore the share of nationally sourced procurement grew steadily during the 1980s. It is likely that this reflects the great increase in inward foreign direct investment that has occurred, complemented by a shift towards the procurement of services as per capita incomes increased.

More appropriate from a compliance perspective would be to identify those sectors where free entry is not possible and where, before liberalization, government purchases completely absorb domestic industry output, as it is only then that elimination of procurement favoritism will affect prices, imports and domestic output. In such situations, failure to observe lower prices paid by the government, greater imports and reduced domestic output would be evidence of noncompliance. Therefore, surveillance efforts need to be sector-specific, using information on the strength of competition policy enforcement and the size of public purchases to identify the sectors which could reveal whether procurement discrimination has in fact been removed. Once identified, then examining whether the share of government purchases sourced domestically has fallen (and the price paid by the procuring entities) is useful.

Another concern is that a nation, having abolished procurement discrimination, not reintroduce them at a later date. If so, net imports will be unaffected in those sectors where government purchases are less than or equal to domestic industry output. Only in sectors where this does not apply should concerns be raised and additional information sought to determine if the price paid by the government exceeds that paid by other consumers. Taken together, these findings suggest that GPA surveillance needs to focus more on disaggregated data, as opposed to the aggregate data that are currently reported.

Surveillance efforts in the WTO should also center on the level of market access barriers that prevail in sectors where procuring entities have a "natural" incentive to source locally, i.e., services and situations where monitoring contract compliance is difficult. Clearly if there are significant barriers to entry for foreign firms this can cause major welfare losses for the economy as a whole, making any procurement policies either redundant or of second order importance (Low, Mattoo, and Subramanian, 1997). More generally, the strength of the competition policy stance that prevails should be assessed, not the endogenous price and quantity responses to procurement liberalization.

The foregoing analysis suggests that the impact of discriminatory procurement regimes is likely to be limited. Often government demand will be too small to affect outcomes. If the demand of government entities is large relative to domestic supply--a situation that is likely to be relatively more common in developing countries--account must be taken of the incentives that are created for entry, which in the long run will eliminate the distortions created by the procurement policy. Procurement favoritism can result in decreases in imports that are not reversible by liberalization, and even elimination, of these policies.

Many of the products purchased by government entities are effectively nontradable: they are either services or there may be economic forces that favor procuring from local suppliers. In such cases, the size of the government becomes irrelevant, in the sense that it will by definition be less than or equal to domestic supply. Procurement preferences will then only be binding if foreign firms cannot contest the market through FDI, or if government entities are able to differentiate across firms on the basis of their "nationality". Outright market access restrictions that take the form of a ban on FDI are clearly going to be very costly to the economy as a whole, and policy efforts should focus on elimination of such barriers. If FDI is allowed, discrimination by government entities against foreign affiliates has effects that are similar to those that arise if the products are tradable. The main difference is that in the short run, imposition of discrimination in procurement will increase the welfare of private sector consumers as they can source at a lower price.

The analysis suggests developing countries will not have uniform preferences with respect to pressures to join the GPA. If countries have a vigorous competition policy and there is free entry in most sectors, benefits of membership will be limited at best. If free entry does not prevail (competition policy is weak), joining the GPA is likely to generate benefits.

Elimination of preferences will imply that domestic industry will contract, so there is likely to be opposition to signing the agreement. Note, however, that the fact that much of procurement concerns services will have an impact, as many of services will have to be provided by firms that have a local presence. Thus, labor will be less affected by foreign entry into services markets than it would be by a shift in government sourcing away from a domestic industry that produces tradable goods. The political economy of liberalization decisions will also be affected by the prospect of obtaining better access to export markets. On the OECD side, procurement markets are already relatively open, as most are members of the GPA and many do not discriminate against nonmembers (e.g., the EU pursues a full MFN policy in this area). Moreover, as these countries have strong competition policies there is unlikely to be much scope for entry by developing country firms (except in sectors where there are barriers to market access more generally, as in many service industries). Pursuit of other developing country markets may be more profitable, as these are countries that will tend to have weaker competition regimes and pursue greater discrimination in procurement.

The analysis also helps to rationalize the US trade policy stance that is being pursued with respect to procurement and antitrust in the WTO. Multilateralizing the GPA would provide better access to developing country procurement markets, and this has value to US firms given that competition policy tends to be relatively weak in many developing countries. Strong WTO-based disciplines on competition law will provide much less of a payoff to American firms as it will erode whatever rents were created by the absence of free entry. Elimination of procurement preference is particularly valuable to US service firms, as discrimination in the procurement of services has the greatest detrimental impact on foreign firms.

Baldwin, Robert. 1970. Nontariff Distortions of International Trade. Washington, DC: Brookings Institution.

Baldwin, Robert and J. David Richardson. 1972. "Government Purchasing Policies, Other NTBs, and the International Monetary Crisis," in. H. English and K. Hay, eds. Obstacles to Trade in the Pacific Area. Ottawa: Carleton School of International Affairs.

Branco, Fernando. 1994. "Favoring Domestic Firms in Procurement Contracts," Journal of International Economics, 37:65-80.

Breton, Albert and Pierre Salmon. 1995. "Are Discriminatory Procurement Policies Motivated By Protectionism?," Kyklos, 49:47-68.

Chen, Xiangqun. 1995. "Directing Government Procurement as an Incentive of Production," Journal of Economic Integration, 10:130-40.

Deardorff, Alan and Robert Stern. 1998. Measurement of Nontariff Barriers. Ann Arbor: University of Michigan Press.

Deltas, George and Simon Evenett. 1997. "Quantitative Estimates of the Effects of Preference Policies," in Hoekman and Mavroidis (eds.).

Evenett, Simon. 1998. "Liberalizing Government Procurement in the APEC Nations," Rutgers University, mimeo.

Francois, Joe, Douglas Nelson and David Palmeter. 1997. "Government Procurement in the U.S.: A Post-Uruguay Round Analysis," in Hoekman and Mavroidis (eds.)

Hoekman, Bernard. 1998. "Using International Institutions to Improve Public Procurement," World Bank Research Observer, 13: 249-69.

Hoekman, Bernard and Carlos Primo Braga. 1997. "Protection and Trade in Services: A Survey," Open Economies Review, 8:285-308.

Hoekman, Bernard and Petros C. Mavroidis. Editors. 1997. Law and Policy in Public Purchasing: The WTO Agreement on Government Procurement. Ann Arbor: University of Michigan Press.

Laffont, J.J. and Jean Tirole. 1991. "Auction Design and Favoritism," International Journal of Industrial Organization, 9:9-42.

Low, Patrick, A. Mattoo and A. Subramanian. 1997. "Government Procurement in Services" in Hoekman and Mavroidis (eds.).

McAfee, R. Preston and John McMillan. 1989. "Government Procurement and International Trade," Journal of International Economics, 26:291-308.

Mattoo, Aaditya. 1997. "Economic Theory and the Procurement Agreement," in Hoekman and Mavroidis (eds.).

Miyagiwa, Kaz. 1991. "Oligopoly and Discriminatory Government Procurement Policy," American Economic Review, 81:1320-28.

Rotemberg, Julio. 1993. "Comment," in Jim Leitzel and Jean Tirole (eds.), Incentives in Procurement Contracting. Boulder: Westview Press.

Trionfetti, Frederico. 1997. "The Government Procurement Agreement and International Trade: Theory and Empirical Evidence," London School of Economics, mimeo.

World Trade Organization. 1996. "Singapore Ministerial Declaration." WT/MIN(96)/DEC.

* The views expressed in this paper are personal and should not be attributed to the World Bank.

1 "This Administration is determined to ... push initiatives to clean up government procurement practices around the world" (Financial Times, May 1, 1995, p. 5). In April 1996, largely at the insistence of the US, OECD members agreed not to allow firms to write off bribes against tax obligations (Oxford Analytica, April 18, 1996).

2 In nontradable services markets the appropriate comparison is between government purchases and home firms supply at free trade prices, not between government purchases and total supply.

3 Note that the analysis can be applied to policies that discriminate across different types of domestic firms, e.g., legislation that grants preferences to small and medium sized or minority-owned businesses.