Disciplining exporting state trading enterprises under economies of scale and oligopoly

1. Introduction

The July 2004 Framework Agreement and the December 2005 Hong Kong Ministerial Declaration within the WTO Doha Development Agenda (DDA) round make it clear that, if an agreement is reached, it will contain specific disciplines for agricultural exporting state trading enterprises (STEs). According to the Hong Kong Declaration, direct disciplines will be introduced on export subsidies, government financing and the underwriting of losses with the commitment to eliminate these three forms of subsidies to STEs by 2013. In addition, the use of monopoly power will also be disciplined in the future. Since the December 2005 Hong Kong Ministerial, negotiations on this issue have not made appreciable progress.

This tentative agreement is the outcome of the long-standing political debate on the trade-distorting practices of agricultural STEs. Although STE operations are already restricted by the GATT, further regulation has been proposed in the DDA negotiations on agriculture. The most notable proposals regarding exporting STEs are those by the US and the EU (WTO, 2000a, b). The US has demanded the elimination of exclusive rights for STEs and of the use of government funds or guarantees to support the STEs. The EU, while asking for the abolition of unfair trade practices in exports, has insisted that the elimination of direct export subsidies is conditional on implicit export subsidies (STEs, certain forms of food aid and export credit) being disciplined in the final agreement. Argentina, Brazil and other Latin American countries in their WTO submissions have also expressed concerns about the potential trade-distorting effects of exporting STEs (WTO, 2001a).¹

On the other hand, among countries where exporting STEs operate, Australia has stated that ‘… the extent to which an STE may, or may not, distort trade depends on the particular activities and measures undertaken by the STE concerned’ (WTO, 2001b: 4); therefore, any additional disciplines should be introduced on a case-by-case basis.

The basic argument of countries demanding further disciplines is that the agricultural STEs are potentially trade distorting because they benefit from government privileges that are not available to private firms.² The suspicion is that their special status allows STEs to circumvent WTO commitments: their single-desk status (i.e. their exclusive right to buy domestic production and to export) means that they may exert market power, use price discrimination and practice price pooling not ‘in accordance with commercial considerations’, as stated by GATT Article XVII, but rather with the aim of subsidising domestic agricultural production and exports. Moreover, governments often provide STEs with financial assistance not available to private exporters; if the DDA round ends in success, these forms of support will be eliminated by 2013.

Various effects of exporting STEs have been examined, such as the trade-distorting effects of single-desk STEs (e.g. Alston and Gray, 2000; Fulton et al., 2001; McCorriston and MacLaren, 2005, 2007), and the trade effects of producer payment schemes used by STEs (e.g. Carter et al., 1998; Hamilton and Stiegert, 2000, 2002; Dong et al., 2006).

The aim of this paper is to analyse the impact of eliminating government subsidies to exporting STEs, as tentatively agreed in the WTO DDA negotiations, and to examine the possible impact of this policy change on market structure.³ One of the open issues in the debate over STEs is whether further disciplines for agricultural exporting STEs would improve competition and ensure progress towards free trade in markets dominated by oligopolistic private firms. Although data on market shares of private trading companies and evidence on their market power are rather limited, there is consensus on the significance of their role in a number of international markets where exporting STEs operate. In this respect, the most interesting case is probably the grain trade, where a small number of multinational exporting firms account for a large share of the international market and there are few exits and hardly any new entries (Davies, 1986; Wilson and Dahl, 1999; Hayenga and Wisner, 2000).⁴ The concern is that eliminating STE privileges may reduce their ability to compete in oligopolistic markets and may strengthen the market power of large private trading firms (Fulton et al., 2001; Young, 2005). What the market structure would look like as a result of this, and the impact on trade and welfare, is generally unclear.

A two-stage duopoly model is developed, with market structure being endogenously determined in the first stage of the game. Two key differences between the private firm and the STE are modelled. As is the case in other papers, the distinctive feature of the STE with respect to the private firm is its objective function (e.g. Thursby, 1988; Alston and Gray, 2000; McCorriston and MacLaren, 2005). Although the official objective of exporting STEs is, by and large, the maximisation of the surplus of agricultural producers, what the real objectives pursued by STEs are is quite a controversial issue and some papers have modelled other objective functions (e.g. Carter et al., 1998; Hamilton and Stiegert, 2002; Dong et al., 2006). This paper assumes that the STE, contrary to the private firm which maximises profit, maximises the welfare of agricultural producers. An explanation of how this assumption may influence the results is provided throughout.

Unlike the existing literature, we introduce a second distinction between the two firms, that is, their organisational structure. The private firm may choose between two different exporting modes: either to sell the product to other agents performing downstream functions (indirect exports) or to integrate vertically in the shipping, storage and processing industries in the foreign country (direct exports). The latter, although requiring initial fixed costs generating economies of scale, is chosen by the private firm in order to avoid transaction costs arising from having to negotiate with downstream operators. The motivation for direct exports is, thus, based on the transaction costs approach to vertical integration developed in the literature on multinational firms (Caves, 1996; Markusen, 2002) and, more recently, in international trade theory (Spencer, 2005). Despite the limited literature on multinational trading firms operating in agricultural markets, there is a consensus on the existence of vertical integration and on the significance of scale economies.⁵

In the model developed in this paper, the STE can only operate as a ‘pure middleman’; that is, it does not own trading facilities and, therefore, sells the product to other firms that perform most of the international marketing functions. As a result, it does not incur fixed costs generating tangible economies of scale. This means that, if transaction costs in international trade are high, the private firm has a competitive advantage over the STE, since it can decide to skip transaction costs and exploit economies of scale. This assumption is based on the observation that agricultural exporting STEs traditionally do not own trading or processing assets abroad (Carter and Wilson, 1997).⁶ Among the most important agricultural STEs, the Canadian Wheat Board is an example of an STE that makes use of accredited agents to reach final consumers.

The model developed in this paper builds on the game theoretical models of international trade that include multinational firms (e.g. Horstmann and Markusen, 1992; Motta, 1992; Markusen, 2002), even though the setting is substantially different, since it considers specific features of international agricultural trade. In the aforementioned models, private firms invest in the foreign countries to jump the tariffs, whereas in this paper the motive for international integration is the presence of transaction costs. Furthermore, all papers mentioned above assume constant domestic costs and symmetry between firms, whereas this paper introduces increasing marginal costs and asymmetry in firm behaviour and objective functions.

The theoretical model is used to study how the elimination of STE subsidies affects market structure. The trade and welfare effects of possible market structure changes are examined by means of numerical examples. The results show that the elimination of subsidies to the STE may have different impacts on market structure, trade and welfare, depending on the initial values of transaction and fixed costs and of the subsidy.

The paper is organised as follows. Section 2 presents the model. Section 3 determines the feasible equilibria and the impact of the elimination of the subsidies to the STE on market structure. Section 4 discusses the consequence on trade and welfare through a numerical example. Section 5 discusses the implications of the results obtained and offers some concluding remarks.

2. The model

The model considers two exporting countries (i and j) and one importing country (z) of a homogeneous agricultural product. Exports from country i are managed by a private firm, while an STE has the exclusive right to purchase and export from country j. Both firms sell on their own domestic market and export to country z. Production in country z is assumed to be zero. This setting is intended to represent competition between two exporting countries on the same export market, with exports carried out by a large private firm in one country and by an STE in the other.

Let p_i, p_j and p_z be the inverse demand curves. Linear functional forms and identical demand in the two exporting countries are assumed: p_i = α − β XM_i, p_j = α − βXS_j and p_z = λ − δ (XM_z + XS_z) with XM and XS being exports/sales of the private firm and of the STE, respectively.

The private firm may export to z either indirectly or directly. In the first case, the firm operates as a ‘pure middleman’ and sells the product to other agents. With this option, the firm incurs transaction costs, t_z, per unit exported, which are assumed to be constant. In the second case, the firm integrates downstream to eliminate transaction costs, but this option implies fixed costs, denoted by G, including both the cost of acquiring facilities for storage, handling, transportation or processing, and also the information and legal costs necessary to open a subsidiary in a foreign country. The private firm, thus, chooses direct (indirect) exports if transaction costs (fixed costs) are high relative to fixed costs (transaction costs). This hypothesis builds on the literature on the multinational firm, which emphasises that vertical integration is frequently chosen by firms to internalise costly arm's-length transactions (e.g. Caves, 1996; Markusen, 2002). Despite the lack of empirical evidence, there are good arguments to support the hypothesis that transaction costs play a relevant role in the firm's degree of forward integration also in agricultural markets. As for grain, for example, Caves (1977) argues that trading in agricultural products is a time-dependent activity, as the optimal utilisation of trading facilities depends on their available capacity at particular points in time. Opportunism in bargaining processes may lead to high transaction costs and make it advantageous to coordinate such activities within an administrative apparatus rather than at arm's length. For other agricultural commodities, case studies show that one of the reasons for vertical integration by international trading firms is the need to ensure access to supplies without incurring costly transactions with local producers (Chalmin, 1986; Read, 1986).

On the other hand, the STE is assumed to export only indirectly. As its mandate is to operate as a ‘pure middleman’, it cannot purchase trading or processing assets. As already mentioned, this feature is common to exporting agricultural STEs.

In this model, both firms purchase, sell domestically and export domestic agricultural products only. The inverse supply functions for the agricultural product faced by the STE and the private firm are assumed to be linear and are given, respectively, by c_s = θ + ρ (XS_j + XS_z) and c_m = φ + κ(XM_i + XM_z).

The firms also bear firm-specific set-up costs (C_m and C_s), which include the cost of acquiring intangible assets (e.g. information) that are essential for the trading industry (Caves, 1977). Firm-specific fixed costs are assumed to be high relative to demand, and thus markets can support at most two firms. Furthermore, the said costs are considered to be faced by firms on the domestic market. Markets are assumed to be segmented.⁷

Two kinds of subsidies to the STEs are considered: a per unit subsidy s, which is assumed to be paid on both domestic sales and exports, and a fixed subsidy U. The former is aimed at capturing certain government privileges that enable STEs to reduce their variable costs. For example, government guarantees for commercial loans allow STEs to have access to financial resources at a lower interest rate than would otherwise be the case (Carter and Wilson, 1997; Dixit and Josling, 1997). On the other hand, the fixed subsidy U captures other privileges that have the effect of reducing the fixed costs of the STE. For example, the expectation of the government's underwriting of losses may encourage the STE to sustain higher administrative costs (Furtan, 2005); it has also been argued that the government guarantee of STE borrowing and export credit sales creates a ‘financial cushion’ that the STE may use to recover general expenses (Goodloe, 2004).

The firms play a two-stage game. In the first stage, firms choose the entry strategy; moves in the first stage of the game are assumed to be simultaneous. The possible actions considered in the first stage do not include those where the firm chooses to export to z without selling on the domestic market. Therefore, for the private firm there are four potential strategies, whereas the STE has three. Table 1 shows the payoff matrix of the game. The action set of the firms is given by the combinations of their choices on the domestic and foreign markets. More specifically, the private firm in the domestic market chooses either to enter (in Table 1, this is denoted by i) or not to enter (denoted by 0), whereas in the foreign market it chooses between the no-entry option (which is denoted by 0), indirect exports (denoted by z) and direct exports (denoted by z_d); the action set of the STE is equal to that of the private firm, but does not include the option of direct exports.

In the second stage, the firms play a Cournot game. The game is solved backwards, by first considering the second stage decision. As elsewhere (e.g. Thursby, 1988; Alston and Gray, 2000; McCorriston and MacLaren, 2005), the STE and the private firm are assumed to behave differently. The private firm is assumed to exercise monopsony power with respect to agricultural producers and to maximise its profits. The STE is assumed to maximise the welfare of domestic agricultural producers.

If the firm maximises its objective function assuming the rival's exports constant (Cournot competition), first-order conditions are given by:⁸

from which we obtain

and

The STE's objective function is

First-order conditions are

which yield

Using equations (4) and (9), equations (5) and (10) can be solved for XM_z and XS_z to obtain the duopoly equilibrium exports when the private firm exports indirectly. The quantities exported to z under the outcome {2, 2} of Table 1 are

where H = β + κ, D = 2β + ρ, E = κ(α − φ), F = ρ(α − θ), A = D(2δ + ρ) − ρ², B = 2H(δ + κ) − 2κ² and Q = AB − δ²HD.

The payoffs to the firms in equilibrium are then derived by substituting the supply and demand functions and the domestic equilibrium quantities into the objective functions (1) and (6), which yields

and

If the private firm opts for directs exports, then its profits are given by

and maximisation leads to the following duopoly equilibrium exports under the outcome {3, 2} of Table 1:

The payoffs to the firms are then given by

and

The equilibrium exports and the associated payoff in the second stage of the game for all outcomes are reported in Tables A1 and A2 of the Appendix.

As already mentioned, a number of papers have assumed that the STEs do not maximise producer welfare but, rather, that they have other objective functions such as the welfare of the bureaucracy (Carter et al., 1998) or STE profits (Hamilton and Stiegert, 2002; Dong et al., 2006). McCorriston and MacLaren (2007) showed how different objective functions imply different market equilibria. In this model, a different objective function affects not only the equilibrium exports and payoff, but also the market structure, as the payoff to the STE influences its choice in the first stage of the game. A profit-maximising STE always exports less than a welfare-maximising STE: first-order conditions would be similar to those of the private firm exporting indirectly [equations (2) and (3)]. Also the payoff would be similar and, thus, lower than in the case of producer welfare maximisation. The outcomes of the game would differ from those obtained when one assumes an STE maximising producer welfare.

3. The elimination of STE subsidies: impact on market structure

Outcomes of this two-stage game may be a Nash equilibrium for certain combinations of the values of parameters. We are primarily interested in analysing how the Nash equilibrium may be affected by a change in the value of government subsidies to the STE, s and U, and in the value of t_z and G, that is, transaction and fixed costs, which in this model crucially determine the action chosen by the private firm. For this purpose, it is first necessary to derive the constraints that must be satisfied for each of these outcomes to be a Nash equilibrium. The analysis is here limited to some of the potential solutions of the game, and does not consider market structures in which at least one firm operates domestically only. Although these outcomes are, in theory, the possible solutions of the game, they are in fact less relevant in the context of the debate on the trade-distorting effects of subsidies to STEs, and hence, for the specific purpose of this paper.

Consider outcome {2, 2} of the game (Table 1). This is a Nash equilibrium if each firm chooses the best response to the action chosen by the rival, that is, the action that maximises its payoff given the action of the rival. This requires (i) M²² > 0, (ii) M²² > M³² and (iii) S²² > 0 with XM_z²², XM_i²², XS_z²², XS_j²² > 0.

The same procedure has been used to derive a complete set of constraints for all other outcomes.⁹

Each set of constraints defines a region in the parameter space where the corresponding market structure is a Nash equilibrium. Before analysing how the subsidies to the STE influence the outcome of the game, consider first how the Nash equilibrium is affected by different values of costs to export to market z. If t_z is large relative to G, then the optimal choice of the private firm, whatever the choice of the STE, is to export directly to market z. The greater t_z, the more the STE is at a disadvantage. If transaction costs become very high, payoff to the STE becomes negative and its optimal choice is the no-entry option. On the other hand, if G is high enough, the private firm opts for indirect exports and the STE enters because its payoff becomes positive. If both t_z and G are very high, then duopoly profits may become negative.

Figure 1 illustrates the possible outcomes of the game in {t_z, G} space, when the STE does not receive any subsidy from its government and under the assumption of cost symmetry between firms.¹⁰ If the firms face the same costs and do not receive any subsidy, then differences between them depend upon two factors: differences in their organisational structure and/or their different pricing behaviour as a result of the differences in their objective functions.

When both G and t_z are zero, regimes {2, 2} and {3, 2} are equivalent. In this case, the key difference between the two firms is the one in their objective function: the private firm exercises monopsony power with respect to domestic producers and its marginal cost is lower than the competitive price, whereas the STE marginal cost is the competitive supply price. This equilibrium corresponds to that obtained in contributions assuming exogenous market structure and zero fixed costs (e.g. Thursby, 1988; McCorriston and MacLaren, 2005).

If transaction costs are positive, however, market structure may change. Figure 1 shows that there are three critical values of t_z above which a regime shift occurs.¹¹ The lower critical value of t_z (in Figure 1, t_z = 11) is defined by the constraints of regimes {3, 2} and {3, 0}. Assume first that the fixed costs are zero. When transaction costs are below that lower critical value, regime {3, 2} prevails; in this regime, firms also differ because of their degree of integration and this implies that, with respect to the market structure {2, 2}, the private firm's exports and payoff are higher (XM_z³² > XM_z²² and M³² > M²²), whereas those of the STE are lower (XS_z³² < XS_z²² and S³² < S²²). This is because the former skips transaction costs and increases its exports at the expense of the latter. When transaction costs are above the lower critical value, the STE chooses the no-entry option and a monopoly of the private firm, i.e. regime {3, 0}, is the outcome of the game.

Consider now the effect of an increase in fixed costs. If transaction costs are below the lower critical value, an increase in fixed costs results in a change in the private firm's export strategy from direct exports (region {3, 2}) to indirect exports (region {2, 2}). This regime shift implies an increase in the STE's market share at the expense of the private firm. But when transaction costs are between the lower and the higher critical value, then the initial regime is {3, 0}. An increase in fixed costs, and the resulting change in the entry strategy of the private firm, makes the STE payoff positive. As a result, when transaction costs are smaller than the intermediate critical value, the regime shifts from a monopoly, i.e. {3, 0}, towards a duopoly, that is, {2, 2} (in Figure 1, this occurs for 11 < t_z < 17). However, if transaction costs are higher than the intermediate critical level, then an increase in G makes direct exports infeasible, while duopoly profits become negative because of the high transaction costs. In this case, the feasible equilibrium is a monopoly of the STE {0, 2} (in Figure 1, this occurs for 17 < t_z < 28). The intuition behind this result is that when fixed costs are high, the private firm cannot exploit its structural advantage; if transaction costs are also high and duopoly profits are negative, the payoff to the private firm becomes negative, while that of the STE can still be positive. This is due to the different behaviour of the two firms. Although the payoff to the private firm is the result of monopoly and monopsony profits, the STE payoff includes monopoly profits. It is possible that the STE payoff is positive, whereas it is negative for the private firm. Thus, when both transaction and fixed costs are high, the particular objective function of the STE may provide it with a strategic advantage with respect to the private firm.

When transaction costs become greater than the higher critical value (in Figure 1, t_z > 28), and fixed costs are high enough, then both firms face negative payoffs and opt for the no-entry option.

The feasible equilibria illustrated in Figure 1 would be somewhat different if the STE were assumed to maximise the profit instead of its producers' welfare. First, assuming cost symmetry between firms and a profit-maximising STE, market regime {0, 2} would never be a feasible equilibrium, as in this case the payoff to the STE would never be greater than that to the private firm. The second difference is in the critical values of transaction costs above which the market regime of the private firm shifts to a monopoly; this would be lower than in Figure 1, because with the profit maximising STE the payoff becomes negative for lower values of transaction costs.

Figure 2 illustrates the Nash equilibrium as a function of t_z and the subsidy s, assuming again that U = 0.¹² As already illustrated in Figure 1, when the subsidy is zero, and transaction costs are smaller than the lower critical level (t_z < 11), there is a duopoly (regime {2, 2}). A relatively small subsidy does not change market structure, even though it changes market shares to the advantage of the STE. If t_z > 11, then the private firm opts for direct exports and this implies a monopoly of the multinational firm (regime {3, 0}). In this case, an increase in subsidy s allows the STE to enter the market despite its structural disadvantage. In Figure 2, even with high transaction costs, a per unit subsidy equal to 5 is high enough to make the STE payoff positive. Clearly, the higher the transaction costs, and thus the disadvantage of the STE, the greater the subsidy needed to make the STE payoff positive. Figure 2, therefore, shows that a subsidy to the STE may re-establish a duopoly and ‘compensate’ the STE for the high transaction costs it faces with respect to its rival.

If subsidy s is high, an STE monopoly may be the outcome (regime {0, 2}). The critical value of s above which regime {0, 2} prevails depends on transaction costs. More specifically, if t_z is low enough (t_z < 15), then the higher the transaction costs, the lower the critical value of the subsidy.

Conversely, if t_z > 15, the higher the transaction costs, the higher the subsidy needed to shift the regime to a monopoly of the STE. This is because when transaction costs are high and the private firm enters through direct exports, an increase in transaction costs reduces the payoff only to the STE. Consequently, the level of subsidy necessary to shift to an STE monopoly increases with transaction costs. On the contrary, when the private firm also faces transaction costs, then the critical value of the subsidy decreases with transaction costs.

What is the impact, then, of eliminating the subsidy to the STE? Clearly, it depends on the initial level of the subsidy and on the value of transaction costs.

If transaction costs are below the lower critical value (in Figure 2, t_z < 11) and the subsidy is such that the initial regime falls in region {2, 2}, then market structure does not change. Even after the elimination of the subsidy, both firms continue to export to the third market, even though the private firm's market share increases at the expense of that of the STE. Conversely, if the value of s is such that the initial regime is {0, 2}, then eliminating the subsidy may allow the private firm to enter the market and, thus, a duopoly prevails. Finally, if transaction costs are larger than the lower critical value (t_z > 11), then the elimination of the subsidy, whatever the initial regime and the value of the subsidy, will result in the exit of the STE ({3, 0}).

These findings differ only slightly if one assumes the STE maximises profits. The feasible equilibria would be the same, even though the critical values of subsidies above which market structure changes would be different. More specifically, with a profit-maximising STE, a monopoly of the private firm exporting directly would prevail even with relatively high values of the subsidy to the STE. Regimes {0, 2} and {3, 2} would be feasible equilibria only for higher values of s to the STE. Therefore, the direction of the impact of eliminating the STE subsidy would be the same as that depicted above, even though the ‘point’ at which market structure changes would be different.

The impact on market structure of subsidies that enable the STE to reduce fixed costs is rather different. Figure 3 shows the Nash equilibrium regimes in the {t_z, U} space, under the assumption that s = 0.

It is worth noting that, if transaction costs are not too high (in Figure 3, t_z < 11), unlike s, a fixed subsidy does not affect market structure: even high values of U do not shift market structure away from regime {2, 2}. This is because the entry decision of the private firm is affected by the STE's variable costs, but not by its fixed costs. Therefore, for low values of t_z, the private firm opts for indirect exports in any case, and U, although increasing the STE's payoff, does not have any effect: market structure, exports and market shares do not change.¹³

However, if transaction costs are high and the private firm opts for direct exports, then the level of U may affect the STE's decision regarding entry: high values of U allow the STE to enter the market and to compete with the private firm, while with low values of U a multinational monopoly prevails.

Therefore, the impact of eliminating the subsidy that reduces the STE's fixed costs may be negligible if transaction costs are low and the private firm exports indirectly. However, if transaction costs are high relative to fixed costs and the multinational already exports directly, then eliminating the subsidy may shift the regime from a duopoly to a monopoly.¹⁴

Overall, the findings of the model suggest that the effects of eliminating STE subsidies may or may not change the market structure depending on the initial regime, which is a function of the values of t_z, G, s and U. If before the policy change there is a duopoly, but the private firm has a strong competitive advantage because of high transaction costs relative to fixed costs, then eliminating subsidy s may cause a change in market structure, with a shift to a monopoly of the private firm. Conversely, if transaction costs are low enough, eliminating s may or may not induce a change in market structure, depending on the initial value of s in relation to t_z: the higher s with respect to t_z, the lower the possibility of a regime shift.

One implication of these findings is that, when assessing the impact of eliminating STE subsidies, one should first consider the key features characterising each export market—in particular, whether or not the private traders are vertically integrated in that specific export market, and whether or not the subsidy granted to the STE is high enough to give the STE a substantial strategic advantage. When looking at the competition between private traders and STEs in a variety of markets, several situations may arise. For example, one key difference is between export markets where an importing STE operates and countries where imports are managed by private traders. In the former countries, as importing STEs have often the exclusive right to import, private exporters cannot be vertically integrated and, therefore, do not exploit economies of scale and their potential for a more efficient marketing structure, whereas the exporting STE exploits the advantages of its peculiar mandate and nature, even strengthened by the fact that importing STEs prefer to deal with exporting STEs, rather than with private traders.¹⁵ In this case, subsidy elimination may keep the STE's payoff positive. The impact would be very different in countries where private exporters are able to integrate vertically. In that case, subsidy elimination could threaten the STE's ability to compete with the private firm, as is the case when the subsidies counterbalance the STE's structural disadvantage.¹⁶

The results obtained also show that eliminating a per unit subsidy (s) may have quite a different impact from the removal of a fixed subsidy (U). The main difference is that, while the value of s influences the decisions of both firms, subsidies affecting the fixed costs of the STE determine the entry choice of the STE only. This means that removing these subsidies may affect market structure as long as it causes the STE payoff to become negative; that is, only if before the policy change the STE's disadvantage with respect to the private firm is so large that without subsidies the STE cannot enter the market. In this model, this is likely to happen when transaction costs are high with respect to fixed costs.

A further implication is that the trade and welfare effects of a policy shock are different from those predicted by models that consider market structure as exogenous and ignore economies of scale. These models, in fact, assume that market structure cannot change and that firms are identical with respect to their degree of forward integration and that the market structure is always the duopoly regime {2, 2}. If in Figure 2 one assumes exogenously regime {2, 2} as market structure, then eliminating subsidy s would not make the STE payoff negative; on the contrary, when transaction costs are high, the private firm is vertically integrated and eliminating s makes market structure shift to a monopoly of the multinational firm. Likewise, eliminating U may have no trade effect if the private firm is assumed to export only indirectly, whereas in our model it may induce a shift to a monopoly of the multinational firm if the latter is vertically integrated. As shown in the following section, the trade and welfare effects of STE subsidy elimination may be very different when market structure changes.

4. Trade and welfare effects of eliminating STE subsidies: numerical examples

The trade and welfare effects of the changes in market structure discussed in the previous section are illustrated here by means of numerical examples.¹⁷ Table 2 presents the changes in total exports, prices and market shares in country z when transaction and fixed costs change and the STE subsidies are zero, under the various possible market structures illustrated in Figure 1. Table 3 provides the corresponding changes in welfare. For country i, welfare is obtained by aggregating the profits of the firm and domestic producer and consumer surpluses, while for country j, welfare is the sum of the payoff to the STE and consumer surplus, minus government expenditure. In country z, welfare is given by consumer surplus only. The percentage changes in the tables are calculated with respect to the benchmark t_z = G = U = s = 0.

Table 2 shows that in regime {2, 2} the STE's market share is always higher than that of the private firm. In this regime, STE exports are well above the level of a profit-maximising firm exerting monopsony power with respect to domestic producers. In addition, within area {2, 2}, an increase in transaction costs further increases the STE's share at the expense of the private firm; welfare in country i diminishes more than in country j, essentially because of the steep decline in the private firm's profits. Overall welfare declines.

Things change if market structure changes. The trade effects when a shift towards regime {3, 2} occurs are rather different: the private firm gains substantial market share, as it skips transaction costs and increases exports at the expense of the STE; the increase in the private firm's exports may be sufficiently high to ensure an increase in total exports. This happens in the example when G = 100 and t_z increases from 5 to 10. Price in market z falls and welfare of countries i and z improves, while that of country j drastically worsens. Producer welfare increases in country i and declines in the STE's home country. Overall welfare may improve; this means that an increase in transaction costs inducing the private firm to change its export strategy may be less harmful for world welfare than one that does not change market structure. The costs of this change in market structure are borne by producers in country j, whose welfare declines significantly.

A sharper increase in transaction costs implies a shift from a duopoly to a monopoly of the private firm, i.e. regime {3, 0}. The numerical example confirms expectations: as a consequence of this shift in regime, a sharp decrease in overall exports to z and a price rise imply a worsening of total welfare. Only the private firm and producers in country i benefit from this change in market structure.

When transaction and fixed costs are high enough for regime {0, 2} to be a Nash equilibrium, the trade and welfare effects are also straightforward. With an STE monopoly, exports are lower and price in market z higher than under the duopoly {2, 2}; the winners are producers in country j whereas the other two countries are worse off. An interesting insight from Table 3 is that the overall welfare in regime {0, 2}, everything else being equal, is lower than in the case of a monopoly of a vertically integrated firm, regime {3, 0}. This is because exports are higher¹⁸ and, thus, country z, is better off. The gains for producers in country i, for the private firm and for consumers in country z, are thus larger than the losses of producers in country j.

The trade and welfare effects of the regime shifts due to a change in subsidy s, illustrated in Figure 2, are reported in Tables 4 and 5. Figure 2 shows that eliminating s does not change the market structure when initially regime {2, 2} is a Nash equilibrium, and when t_z is below the critical value. As expected, eliminating s implies an increase in the private firm's market share at the expenses of the STE. As a consequence, however, total exports decline and price in market z increases because, under regime {2, 2}, exports by the profit-maximising firm exerting monopsony power are lower than those of the STE maximising producers' welfare.¹⁹ In this case, the only winner is country i, which benefits from the higher profits of the private firm and higher producer surplus, whereas both countries j and z are worse off. In country j, the sharp decline in producer surplus is not compensated for by the decrease in government expenditure, if s is initially low enough (in our example, s ≤ 8). The elimination of subsidies to the STE without any change in market structure makes overall welfare decline.