“Parallel trade” takes place when there is trade in the products of a firm outside (and in parallel with) the distribution network that the firm has established for its products (COM, 2003, p. 6). Parallel Imports (PI) affect a wide range of industries, spreading from traditional luxury and branded consumer products (detergents, cosmetics, wines, cameras, and watches) to industrial (such as automotive parts) and, very importantly, pharmaceutical products. In the latter case, parallel trade often involves goods that are produced under the protection of a copyright, trademark, or patent that are placed into circulation in one country and then imported into a second country without the permission of the owner of the intellectual property rights attached to the product in the second country (Müller-Langer, 2008).

Thus, parallel imported products are very often not counterfeited or pirated but are legitimate products. However, they may not carry the original producer’s warranty and may be packaged differently. Moreover, parallel importing firms ordinarily purchase a product in one country at a price that is lower than the price at which the product is sold in the second country (arbitrage between markets).

PI have been treated very differently in different countries. In EU competition law there is a strong (per se) prohibition of restrictions to parallel trade, which is firmly rooted in the traditional public interest concern with free trade and market integration in Europe¹. While this is often recognized as a specific “public interest” concern² that is incorporated in EU competition law, it is important to enquire whether this can be justified on the basis of standard antitrust arguments: specifically, to ask whether this prohibition would result in the promotion of consumer welfare in one or more countries.

At the opposite extreme to that of the EU policy, we have Russia’s per se prohibitions of PI which can be potentially justified by the country’s industrial policy objectives of protecting its domestic industries. While there is no evidence of a shift in policy by the European Commission in its per se prohibition of restrictions to PI, there is evidence that the policy in Russia shifts away from the per se prohibition of PI and a recognition that “in some cases” PI should be considered legal. Specifically, at the end of 2018, the Federal Antimonopoly Service of the Russian Federation (FAS) presented a draft law to the Ministry of Economic Development aimed at partially legalizing PI of goods in Russia (imports of original goods without the trademark owner’s consent). According to the proposal, one of the reasons for allowing PI will be that there is domestic overcharging of goods (see Galtsova and Dovgan, 2018).

We consider the desire expressed by FAS for a shift in policy towards PI³ as being in the right direction, while we consider unjustified the continuation of EC policy of per se prohibition of restrictions to PI. Here we espouse a middle ground in which whether restrictions of PI must be prohibited or not should be the subject of investigation of the specific economic facts of the case and what these imply in terms of the impact on welfare (and specifically on consumer welfare). That is, whether or not there is law violation should be justified on consumer welfare grounds.

In most of the cases that have been examined by European competition authorities a firm with a dominant position selling in different countries is accused of taking measures that inhibit PI in one or more countries. The firm will be engaging in price discrimination and its price will not be the same in the different countries. PI may flow from the low-price to the high-price country and the inhibition of PI can be considered as a method by the firm to protect its price discrimination strategy⁴. As such, EU competition law does not prohibit this conduct (there is certainly no per se prohibition of price discrimination in EU) as it is recognized that given differences in demand conditions and/or costs in different countries, firms that operate in these, under competitive conditions, will be expected to set different prices for their products. Moreover, this behavior is not expected to necessarily or often lead to negative effects on the welfare of these countries,⁵ or to lower consumers’ welfare.⁶ The core of this argument revolves around the fact that when firms are free to set their optimal price in different markets, depending on demand and cost conditions, then, in general, differential pricing allows them to serve more markets. In other words, if a company is “obliged” to use a uniform price strategy, then even under the assumption of the same cost but different demand conditions, the company may decide not to serve some markets. Price discrimination thus allows producers to make some consumers better off (those of the low price country), without making other consumers worse off (those of the high price country). As Rey (2003) has put it:

“If the firm must adopt a uniform price, it can in fact choose between two strategies: serving both markets at a price p reflecting the average price elasticity (so that p₂ < p < p₁), or withdrawing from the high-elasticity market and thus serving the low-elasticity one at the same price as before: p = p₁. Adopting the latter policy is particularly likely if the elasticity is very high on the second market, since serving both markets would then imply a substantial loss of profitability in the first market. Whenever the firm chooses to withdraw from the second market, price uniformity benefits no customer: in the first market customers are offered the same price as before, while in the second market customers have less choice than before and thus again incur a loss of surplus.”

In this paper we propose a balanced effects-based (or rule-of-reason) approach to the antitrust treatment of restrictions to PI. We show that, under many circumstances, PI are unlikely to have a substantial effect on consumer welfare while there may be other negative effects of PI (that we discuss below in Section 4). We demonstrate that under many configurations of the parameters influencing the outcome, PI are unlikely to exert downward pressures on domestic prices and, when firms take measures to inhibit PI, this is unlikely to generate any significant upward pressure on these prices.⁷ This implies that a per se prohibition of restrictions to PI cannot be justified. There are, however, also situations in which a policy of prohibiting restrictions to PI makes good sense on consumer welfare grounds because such restrictions can lead to significantly increased prices relative to the situations without restrictions to PI.

The model that we propose assumes that an oligopolistic firm is selling in different markets (specifically, the “domestic” (d) and “foreign” (f) markets) at different prices. The firm is dominant in the d-market and market conditions⁸ are such that the price in this market without PI is higher, that is, p_d,w > p_f. The firm is facing PI in the d-market from a competitive fridge of parallel importers that take as given the domestic price of the firm. In order for parallel importers to have an incentive to engage in parallel trade the gap between the foreign and domestic prices must be such that the cost (transportation and any other cost) of importing from the foreign market can be covered and a positive profit margin can be made. That is, if the minimum price of PI (that equals the foreign price plus the cost of importing and distributing in the domestic market) is p^PI, this must be less than p_d,w, for there to be an incentive to engage in parallel trade. We examine the firm’s optimal pricing strategy and, given this, the impact of PI on the domestic market — impact on domestic prices and on profits.

We find that there are two potential outcomes (equilibria) that could emerge, that depend on the configuration of a number of (potentially measurable) parameters: the difference between the firm’s domestic price and p^PI (that we denote by δ), the fraction of domestic sales that PI can satisfy (γ) and the extent to which the firm can limit PI, that we denote by m (e.g., through exclusive contracts with some independent domestic distributors).

In one equilibrium, that will tend to emerge when δ, the difference between domestic price and p^PI, is not very large, and m is not very large, the firm’s optimal strategy is to set a price just below p^PI and deter all PI (“deterrence strategy”). This equilibrium can also emerge for larger δ when γ is quite large and m is not large. In this equilibrium, the threat of PI induces a low-price equilibrium that benefits consumers, with prices falling significantly (according to our simulations by even more than 15%) relative to the prices without the threat of PI. So the policy should certainly be one of allowing PI, though in equilibrium no PI takes place and hence no restrictions to PI are necessary.

In the second equilibrium, that will tend to emerge when the difference between domestic price and p^PI is large, as in many cases in practice, e.g., those that have been found to violate EU competition law, the firm’s optimal strategy if it cannot restrict PI, is to set a price above p^PI and allow PI, i.e. accommodate parallel traders (“accommodation strategy”). In this case, the optimal price (p_d^*) will be again lower than the optimal price without PI (p_d,w), but the difference will tend to be very small (close to 3% for many parameter configurations, rising to about 8% in a limited number of cases). So PI should be allowed, though the benefit to consumers will tend to be very small and the main benefit from allowing PI will be to shift profits from the firm to the parallel importers. Further, if the firm can restrict PI in this case (m > 0), it will have an incentive to do so (to minimize the shift in profit to parallel importers). But, very importantly, even if its ability to restrict PI is very significant (being able to reduce PI by even 50%), the effect of this on the price reduction that would be achieved in the absence of any restrictions would be negligible (for reasonable parameter configurations less than 2,5%). This small effect on prices suggests that per se prohibitions of restrictions to PI (as in EU) is not justified — given the existence of other potential negative effects often associated with PI (that we discuss below in Section 4).

There are, however, also situations in which a policy of prohibiting restrictions to PI makes good sense on consumer welfare grounds. These are situations in which the parameter configurations favor a deterrence strategy by the firm when there can be no restrictions to PI but induce a switch to the accommodation strategy if the firm would be allowed and can take measures⁹ that significantly impede PI. Then, not prohibiting restrictions to PI can lead to very significant price increases (relative to the equilibrium in which there is prohibition of restrictions to PI) as we shift from a low price deterrence equilibrium to a high price accommodating equilibrium.

The structure of the paper is as follows. Section 2 sets out the proposed model and the equilibrium conditions describing how PI impacts on a dominant firm’s pricing strategy. Section 3 then derives and discusses our main results. Section 4 discusses whether our results are consistent with the results emerging from recent empirical analyses of the impact of PI and outlines some other recent theoretical arguments that complement our analysis concerning other effects of PI. Section 5 offers concluding remarks.

We assume that the residual demand of one of the firms¹⁰, dominant firm I, in an oligopolistic domestic market is linear and that the firm’s marginal and unit cost is c_d. So:

p_d (Q_d) = a_d – b_d Q_d, a_d, b_d > 0. (1)

Note that given the parameter a, (1/ b) also measures market size since:

(1′)

therefore the smaller the b, the bigger is the market size.

Assume also that in a representative foreign market (f) in which firm I also operates, i.e. the market from where PI originate, the product unit cost is c_f and the residual demand for the same product of I is:

p_f (Q_f) = a_f – b_f Q_f, a_f, b_f > 0. (2)

In the absence of PI, in order for I to maximize its profits in both markets, prices and quantities in the d-market will be, respectively:

(3)

and in the f-market they will be:

(4)

Therefore, in order for Ι’s domestic prices without PI to be higher than the prices in the f-market:

p_d,w > p_f if a_d + c_d > a_f + c_f. (5)

In other words, given the relevant marginal production costs, prices in the dmarket will be higher than prices in the f-market, if the maximum willingness to pay for the good is higher in the dmarket (a_d > a_f). The price difference, which is the result of the different consumer preferences, may become even bigger when marginal cost in the d-market is higher than marginal cost in the fmarket (c_d > c_f).

On the other hand, the quantities sold in both markets depend on the size of the market and thus the quantity that firm I is going to sell to the f-market will be higher than the quantity that it will sell in the d-market when:

(6)

Even if a_d > a_f the quantity that I will sell on the f-market will be higher than the quantity sold in the d-market if b_f is substantially smaller than b_d.

Regarding parallel importers (or distributors of PI), we consider that the most appropriate assumption to make, i.e. the assumption that most closely reflects reality in most instances where competition authorities have to deal with restrictions of PI, is that generally many small firms can potentially enter the market of PI and thus form what is commonly known as a competitive fringe. This implies that PI distributors take the price of I in the d-market (and the f-market) as given: If Ι’s prices in the d-market are higher than those in f-markets (including importation and distribution cost), there exists an incentive for PI. In the opposite case there is no incentive for PI. However, even when the price differences and the transportation cost create an incentive for PI, in practice the actual PI of eachPI distributor are small, in comparison with the total sales of I in the domestic market. The explanation behind this lies in the fact that the imported quantities from other countries come from third parties’ (e.g., foreign wholesalers’) surpluses that are likely to be limited and not directly from the producers of those goods. Another complementary explanation is the existence of sometimes severe legislative barriers on imports creating high administrative/bureaucratic costs.¹¹

Assume that per unit minimum cost and minimum sale price for making available PI in the d-market by a distributor is:

p^PI = p_f + c^PI, (7)

where c^PI is defined as the per unit cost of a PI distributor to transport and distribute Ι’s product from the fmarket to the d-market. The distributor will have the motive to engage in PI, if:

p^PI < p_d,w. (8)

Assume also that the maximum quantity of Ι’s products that PI distributors can import from the f-market equals Q¯^PI.

The Supply curve S(p) of PI distributors can be described as follows:

(9)

The supply curve (9) is incorporated in Fig. 1. It is assumed that p^PI = 55, as in the simulations in the Appendix.

Parallel imports move Ι’s residual demand curve in the d-market down and to the left by an amount equal to the amount of PI, for prices higher or equal to p^PI. For prices lower than that level, Ι’s residual demand remains the same as without PI. The (reverse) demand function (1) is now defined as:

p_d = a_d – b_d (Q_d^r – Q¯^PI) , (10)

where Q_d^r is Ι’s residual demand (where “r = residual”), in particular:

(11)

where ε is a very small number and p (Q¯^PI) is the price corresponding to quantity Q¯^PI .

Ι’s marginal revenue is given in this case by:

(12)

where Q_d^r (p^PI) is the (residual) quantity that corresponds to price p^PI.

Note that Ι’s residual demand increases significantly when its price falls by very little below the minimum sale price of PI, i.e., when price decreases from p^PI to (p^PI – ε) [equation (11)]. For quantities corresponding to prices higher or equal to p^PI, Ι’s optimal quantity (that maximizes Ι’s profit, for the above quantity range) is given by equating the first part of (12) to marginal cost. Assume Q_d^* is defined as this optimal quantity and p_d as the corresponding price given sales of Q¯^PI by parallel importers. There is, however, a quantity range corresponding to the quantities between the quantity at price p^PI and the quantity at price (p^PI – ε), for which Ι’s marginal revenues are not defined. In order to define Ι’s final choice we must therefore compare, its profit with (Q_d^*, p_d^*) , with its profit when price is (p^PI – ε) and sales are Q_d^r (p^PI – ε).

Given the above remarks, let us now examine the relationship between Q¯^PI, the price p^PI and the prices that I will set in the domestic market. Using the first part of (12) the profit maximization condition is:

a_d – b_d Q¯^PI – 2b_d Q_d^r = c_d . (13)

From equation (13) it follows that:

(14)

and thus from equation (10):

(15)

Ι’s profit with these choices equals:

π_d^* = (p_d^* – c_d) Q_d^* . (16)

On the other hand, Ι’s profit with price (p^PI – ε) and sales Q_d^r (p^PI – ε) is as follows:

π (Q_d^r (p^PI – ε)) = (p^PI – ε – c_d) Q_d^r (p^PI – ε). (17)

Thus, the price p~_d and the quantity Q~_d of firm I in the d-market will be:

(18)

(18′)

In the case where the first of the equations for the price in (18) above holds, the parallel importers will make positive profit π^PI > 0, absorbing a part of Ι’s profit that equals

π^PI = (p_d^* – p^PI) Q¯^PI . (19)

If, on the other hand, the price drops to the level (p^PI – ε), then PI (and therefore the profit of parallel importers) drops to zero.

Note that we can also write the condition

π_d^* ≥ π (Q_d^r (p^PI – ε)), (20)

using the equations (14)–(17) as follows:

(p_d^* – c_d) ² ≥ (p^PI – c_d) (a_d – p^PI) , (21)

or:

(21′)

Also, from (3), inequality (21) can alternatively be written as:

(21″)

where

p_d,w > p_d^* > p^PI ≥ c_d (22)

and

p^PI = p_f + c^PI, (23)

The inequality (22) results from comparing (3) with (15). Even though p_d^* was defined above to be in the range of prices that are greater than or equal to p^PI, the inequality (23) is implied by the fact that when the price p_d is close enough to p^PI, firm I would prefer to increase significantly its sales, with a price (p^PI – ε), thus making more profit, since at this price PI would fall to zero (so, p_d^* cannot be equal to p^PI) . Finally, we assume that the third (non-strict) inequality holds (although, in principle, if the marginal cost varies very considerably between countries, this may not hold). If this was not true, (21) and (21′) would always hold and I would always choose to produce quantities that lead to price p_d^*.

We can use (21″) in order to determine the optimal pricing strategy of firm I when faced with PI. Once this is done we can then compare the prices under the optimal strategy with the prices without PI as well as with the prices when PI is restricted by I through contractual clauses with its distributors.

Given the minimum level of the PI prices (p^PI) — the minimum level for which there is incentive to undertake PI, and the marginal costs, we can undertake the comparisons for different levels of three important parameters:

(i) δ: this measures the difference between p^PI and the optimal domestic price without PI, that is:

p_d,w = (1 + δ) p^PI, 0 < δ < 1. (24)

(ii) γ: this measures the percentage of PI in the total domestic sales of I, that is:

Q¯^PI = γ Q_d,w , 0 ≤ γ < 1. (25)

(iii) m: this measures the extent to which restrictive contractual clauses imposed by I on its distributors limit PI, that is, if g is the percentage of PI sales in the total domestic sales of I with the restrictions imposed by I on PI, then:

g = γ (1 – m)p^PI, 0 ≤ m ≤ 1. (26)

Of course, if restrictive contractual clauses are prohibited by competition law then m = 0. But even in the absence of a prohibition by competition law there will be constraints to the extent to which PI can be restricted by I, so in practice, even in such cases m is unlikely to be very large — see also below.

Assume also for simplicity that:

c_d = c_f = c. (27)

Then, given (3), (25) and (27):

(28)

Given (28), taking into account (24) and (27), (21″) becomes:

(29)

Substituting, from (3), for:

a_d = 2p_d,w – c = 2(1 + δ) p^PI – c, (30)

we have, taking into account (24), that the condition (20) becomes:

(31)

(31) determines the optimal pricing strategy of firm I for any given value of the parameters γ and δ, in the absence of any contractual restrictions on its distributors and given the minimum price level required for PI to take place (p^PI) and c.

When firm I restricts PI by imposing contractual restrictions (such as exclusivity agreements) on its distributors, then condition (31) becomes:

(31′)

We can also write the optimal price (p_d^*) , from (15) and (28), as follows:

(32)

and so, given (30):

(33)

.Given p^PI, (33) determines the optimal price of firm I (if it chooses the first pricing option mentioned above) for any given value of the parameters γ and δ (and in the absence of contractual restrictions on its distributors).

Finally, the price when firm I restricts PI, by imposing contractual restrictions on its distributors that prohibit PI sales by them, is given by:

(34)

where g ≤ γ is given by (26) and, of course, (p_d^*) ^restr. > p_d^* .

Thus, given the presence of contractual restrictions that limit PI in the domestic market, the optimal price p~_d set by firm I in the presence of PI will be given by:

(18″)

where, of course, (π_d^*) ^restr. is profit at price (p_d^*) ^restr..

We can now establish a number of results using the model presented in the previous section. The main results are presented below in the form of propositions and a number of corollaries. The results are supported by the simulations presented in the Appendix, that are based on the above model calibrated by using reasonable values for the parameters γ, δ and m (and hence g) that can be found in real world cases. Specifically: 0.1 ≤ γ ≤ 0.2; 0.1 ≤ δ ≤ 0.5; 0.1 ≤ m ≤ 0.5.¹²

Proposition 1 (accommodating equilibrium as a result of smallγor large m):

(i) Even when firm I cannot restrict PI,¹³ i.e. m = 0, if PI are a sufficiently small fraction of Ι’s domestic sales (γ is sufficiently small), it will find it optimal to accommodate all PI (rather than to deter PI by lowering its price), setting its optimal price at p_d^*. Since p_d^* < p_d,w , there is a decrease in domestic price relative to the situation with no PI. The exact effect on domestic price depends on the value of γ, but is likely to be neglibible when, as we assume here, γ is small.

(ii) When Ι’s ability to restrict PI is significant (m is large), firm I will have incentives to restrict PI (even if PI is a significant fraction, γ, of domestic sales) and set a price at (p_d^*) ^restr. > p_d^*, accommodating all (non-restricted) PI. The effect of this on domestic price will depend on whether, if I were unable to restrict PI¹⁴, the optimal stategy would be also accommodating (i.e., as in (i) above) or it would be deterring (as described below in Proposition 2). In the first case, the loss in price reduction relative to non-restriction of PI (m = 0) would be negligible. In the second case, the loss in price reduction relative to non-restriction can be very significant (as described in Proposition 4).

Proof: Condition (31′) can also be expressed as follows:

(p^PI – c)(λ – 1) [(1 + δ) p^PI – c + δp^PI] + λ (δp^PI) ² ≥ 0. (31″)

where

(35)

Thus, for part (i) of the Proposition 1, given that g = γ when m = 0, when γ is small, λ will be close to 1 and the expression on the LHS in (31″) will be positive, so I will prefer to accommodate PI and set price equal to p_d^*.

For part (ii) of the Proposition 1, even if γ is not small, if m > 0 and sufficiently large, g will be close to zero, λ will be close to one and (31″) will again hold, so Ι’s optimal strategy will be to restrict PI and set price (p_d^*) ^restr..

To see the effect on domestic price described in the Proposition 1, from (33), taking into account of (24), the difference between optimal domestic price with and without PI for this case (expressed in percentage terms) is:

(36)

of 50% at the optimal price without PI, then from (36), the reduction in price from PI at the accommodating equilibrium will be just 2,5% (i.e. very small, as mentioned in part (i) of the Proposition 1). Note that the main impact of PI here is to reduce significantly the profit of firm I that loses sales volume of 10% and sells at a price 2,5% lower which is redistributed to the parallel importers.

When firm I can restrict PI (so m > 0), as in part (ii) of the Proposition 1, its optimal price would be (p_d^*) ^restr. > p_d^* and this will limit the reduction in price from PI. Specifically, now the reduction in price will be:

(37)

Even if m is as high as m = 50%, with γ = 10%, the value of g would be 5% and the reduction in price would be 1,25%. What is important is that, relative to the reduction in price without restrictions to PI (m = 0) there is a loss in price reduction from restrictions of PI of just 1,25% (i.e. from 2,5% to 1,25%). If γ were 20%, the reduction in price without restrictions (m = 0) would be 5% while with 50% restrictions (m = 50%), and so g = 10%, the price reduction would be 2,5%, so there would be a loss in price reduction from restrictions of PI of just 2,5%. It is worth remembering here that since firm I will be able to reduce PI by imposing restrictions only on those large distributors (e.g., supermarket chains) with which it has direct collaboration, it is unable to control a potentially large part of PI. Therefore, m = 50% is likely to be an overestimate¹⁵.

Proposition 2 (deterrence equilibrium): Given that there is a potential for PI to take place (γ > 0), and m < 1 (so it is impossible to restrict all PI), if the difference between the optimal price in the absence of PI and the minimum price required for PI to take place (p^PI) is sufficiently small (that is, if δ is small), firm I will prefer to set price (p^PI – ε) and exclude all PI (deterrence strategy). The deterrence strategy will also emerge for higher δ for as long as γ is quite large and m is not large.

Proof: in (31″) the first term of the LHS is negative (since λ < 1, with m < 1). If δ is sufficiently small, the positive second term of the LHS will be close to zero and the expression in (31″) will be certainly negative, implying that firm I will prefer to choose the pricing strategy of setting price (p^PI – ε) (the deterrence strategy) thus excluding PI from the domestic market. From (31″) we also see that if m is not large and γ is quite large (so g is quite large and λ quite small), the first term on the LHS will be more negative and this will make the LHS negative even for higher δ. This is confirmed by our simulations results in the Appendix — see Table A.8, where the deterrence strategy is, for example, chosen with δ = 0,3 and γ = 0,2 (with m = 0 and m = 0,25). Table A.9 shows that the deterrence strategy will be chosen even with δ = 0,4 (and γ = 0,2), if m = 0.

WIth this deterrence strategy, that may be maintained — as illustrated in the Appendix simulations — even for very significant levels of m (m = 0,5), as shown in Table A.7, the threat of PI induces a low-price equilibrium that benefits consumers, with prices falling significantly (16,7% in Table A.7; 23,1% in Table A.8), relative to the prices without the threat of PI — so the policy should certainly be one of allowing PI, though no PI (and, of course, no restrictions to PI) is observed.

Proposition 3 (accommodation equilibrium due to high value of δ) : Given a maximum level of PI that is not too large,¹⁶ if the difference between optimal price in the absence of PI and p^PI is sufficiently large (δ is large), firm I will, even if it cannot restrict PI (m = 0), prefer to set price equal to p_d^* and accommodate all PI (i.e. it will again choose the accommodation strategy). For larger values of γ, the same result will hold for larger δ.

Proof: In condition (31″) if the maximum level of PI is not too large (γ is not large), λ will not be much smaller than one (even with m = 0), the negative first term on the LHS of (31″) will be small and thus, for large δ, the second positive term in (31″) will dominate the first (negative) term and the expression on the LHS in (31″) will be positive. Our extensive numerical simulations imply that for γ not larger than 10%, differences between domestic prices in the absence of PI and p^PI that exceed 25% will lead I to set price equal to p_d^* and accommodate all PI.¹⁷ For larger γ (e.g., γ = 20%), larger values of δ (δ > 40%) will lead to the same result. As in the case of Proposition 1 (i), in this case, too, the reduction in domestic price will be given by (36) and is likely to be very small: the consumers do not get significant benefit and the main effect of PI is to shift profits from firm I to the parallel importers.

Of course, in these circumstances firm I will have incentives, if this is feasible, to restrict PI (so make m > 0) to limit the loss of its profit. As in case (ii) of Proposition 1, the reduction in domestic price will be given by (37) and the loss in price reduction relative to non restriction of PI (m = 0) would be negligible.

Proposition 3 will hold when the difference between domestic and foreign prices is large and there are relatively small transport or other costs that the parallel importers have to incur.

Finally, we derive the following important result that describes the case in which firm I switches from a deterrence to an accommodating equilibrium as m (ability to restrict PI) increases:

Proposition 4 (switch in strategy): Restrictions in PI can have a very significant effect in the extent to which domestic price is reduced (i.e. in limiting the extent of price reduction induced by PI) when, without restrictions in PI, firm I chooses the deterrence strategy and switches to the accommodating strategy under sufficiently high levels of PI restrictions. That is, when large increases in m induce I to shift from a low-price (deterrence) strategy (when m = 0 or small) to a high-price (accommodation) strategy (when m is large).

Proof: A switch in strategy will occur when the LHS of equation (31″) is negative when m = 0 or m is not large (so the deterrence strategy is optimal) and becomes positive as m increases. Certainly, as m increases, g approaches 0, so λ approaches 1, the first term on the LHS of (31″) approaches zero and so the LHS of (31″) is certainly positive. Thus with sufficiently large m, certainly an accommodating strategy will be chosen.¹⁸ On the other hand, as we have seen in Proposition 2, there will be a large range of parameter configurations, with m not large for which the LHS of (31″) will be negative and the deterrence strategy will be chosen.

Proposition 4 is illustrated in Appendix Table A.2 of the simulation results in the Appendix below. With γ = 0,1 and δ = 0,2 (moderate differences between domestic and foreign prices), if m can increase to 0,5 (so firm I can take actions that restrict PI to half of their potential maximum level) then firm will switch price-setting strategies and by doing so a price reduction of 16,71% will be limited to just 1,17%. A similar situation under different parameter configurations is described in Appendix Tables A.8 and A.9. In Table A.8 a price reduction of 23,1% is limited to just 3,6% by the switch in strategy when m = 0.5. As we have noted already, we consider a value of m = 0.5 extremely high: firm I will only be able to reduce PI by imposing restrictions on those large distributors (e.g., supermarket chains) with which it has direct collaboration (and who are likely to avoid using PI even in the absence of any contractual restriction imposed by I) — therefore it is unable to control the potentially very large part of PI which is distributed through smaller distributors and informal channels.

Results from calibrated simulations. In the Appendix we present a number of simulation results. In the first set of five Tables we assume that γ = 10%, which will often be close to the maximum percentage of domestic sales that PI can capture.¹⁹ We allow δ, the difference between domestic price and the minimum price of PI, to vary between 10% and 50%. And we allow m, measuring the extent of contractual restrictions that can be imposed by firm I on PI, to vary from 0% to 50%.

In the second set of five Tables we repeat the simulations for a value of γ = 20%. In all the Tables the equilibrium domestic price is p~ and this is obtained by the profit comparison in (18″).

In each Table we explicitly indicate the strategy (of deterrence or accommodation) followed by firm I and whether there is a switch in strategy when such a switch occurs). The simulation results support the conclusions reached in the four Propositions given above.

In recent decades, the European Court of Justice has repeatedly upheld European Commission decisions against firms that had sought to limit parallel trade within the EU. This strong (per se) prohibition of restrictions to parallel trade is firmly anchored in EU competition law. However, against this background, economic literature’s (theoretical and empirical) support of this approach has not been demonstrated (see also Petit, 2011).²⁰

Proponents of PI, usually intuitively, argue that PI leads to downward price equalization and increased intra-brand competition (competition between perfect substitutes, i.e. products/services of a same brand) to the benefit of consumers. However, this argument has not been documented in the theoretical or empirical economic literature. The model of PI presented in this paper suggests that even quite large levels of PI in a domestic market are not expected, in many realistic cases, to have significant effects on domestic prices, which also explains why economists have often doubted that PI can induce positive effects on consumer welfare (albeit also on the basis of rather informal theorizing).²¹

The above results of our model are perfectly aligned and confirm the existing studies which indicate that the effects of PI are at best ambiguous. Indeed, many empirical studies have found no effect (or significant influence) of PI on domestic prices and the effect of PI seems to be mainly to redistribute to parallel importers a part of the profit of the firms whose products are imported — see for example the review of N. Petit (2011) that refers to the empirical studies of NERA (1998), the LSE (2004), Linnosmaa et al. (2003), the Swedish Competition Commission (1999), and Person et al. (2001) that find that there are no effects on domestic prices or if there are, they are limited to specific products and are at best very small or negligible.²²

There are also other reasons for these empirical results, reasons that cannot be dealt with within the context of our model. One of these reasons is that often clients of parallel importers (for example the supermarket chains) have sufficient market power that allows them not to pass on to final consumers the lower prices at which they buy products via PI — so the PI simply lead to an increase in supermarket profit and the profit of parallel importers and a reduction of the profit of firms whose products are imported.

Apart from the empirical studies, as we have already mentioned (see Rey, 2003), the economic literature has examined a number of other reasons that PI should not be encouraged, indeed should be discouraged as harmful to consumers. One of these reasons is that they prevent firms from engaging in discriminatory pricing when consumers in different countries/markets have different willingness to pay for the same product — which may be due to different preferences, differences in culture, investment in “branding” etc (see NERA, 1998). The non-discriminatory pricing constraint is likely to lead firms to not sell their products in countries/markets where consumer willingness to pay is low. By following a differential pricing policy in these cases the total consumption can be increased as well as the total welfare (see also Valetti and Szymanski, 2005; Grossman and Lai, 2008; Kenny and McNutt, 1999).²³

Parallel imports can lead to significant negative long-term effects on consumers when, because of the considerable reduction of profit, they have a negative effect on firms’ incentives to invest in research and development (R&D) and innovation. This effect has been noted particularly in the huge literature (see also Müller-Langer, 2008) on PI in the pharmaceutical industry; this analysis can also be applied to any market where companies are investing in R&D with the aim of improving the quality of their products. Given that in many cases PI would not lead to price reductions, but simply to the transfer of part of firms profits, that invest significant amounts in R&D, to wholesalers who do not invest in R&D or in some productive activity and whose activity has no additional social value (in essence they are simply “speculators”), the application of competition policy in Europe to the issue of PI in the past has basically encouraged a redistribution of profits (from producers to wholesalers) with highly controversial implications.

Finally, PI may have several, direct or complementary, negative effects on consumer welfare (see also Petit, 2011), such as:

An important conclusion that emerges from the Propositions above and the simulations in the Appendix, is that prohibiting restrictions to PI may not affect (reduce) significantly dominant firm I’s domestic price (and hence may not increase consumers’ welfare) in many circumstances, something that is particularly important given that PI may have other detrimental consequences. Specifically, when firm I chooses an accommodating strategy, as described in Propositions 1 and 3, in equilibrium, domestic price reductions will be small and the effect of even very significant restrictions to PI will affect negligibly these price reductions. The intuition here is that firm I would prefer in this case to reduce its own domestic sales thus accommodating parallel importers, enabling them to make all the imports that they can and so maintaining domestic prices at relatively high levels. Although this results in a loss of profit, as sales pass to parallel importers, the profit with this strategy is still greater than the profit that would be made by a deterrence strategy of increasing sales to a level which would lead the domestic price to the minimum price required for parallel importers to be active in the domestic market.

More generally, a policy of prohibiting restrictions of PI will not have a significant effect on price unless (a) in the absence of any restrictions to PI firm I will have chosen a (low-price) deterrence strategy, and (b) its ability to restrict PI is significant, and will switch to a (high-price) accommodating strategy having restricted, if allowed to do so, sufficiently PI. In case (a) and (b) hold, on the other hand, restrictions in PI can have a very significant effect in limiting the extent of price reduction induced by PI and hence a very significant adverse effect on consumer welfare, that would be unlikely to be outweighed by other potential benefits of restricting PI.

Due to the fact that our analysis points to a wide range of potential outcomes that would emerge under different conditions characterizing different antitrust cases in the area of PI in the real world, we consider that their appraisal should rely on case-by-case investigations of the specific economic facts of each case and what these imply in terms of the impact of the conduct on consumer welfare.