Measuring the Market Power of the Portuguese Milk Industry

Transcription

1 International Journal of the Economics of Business, Vol. 6, No. 2, 1999,. 209± 222 Measuring the Market Power of the Portuguese Milk Industry MARGARIDA DE MELLO and ANT ÂONIO BRAND ÄAO ABSTRACT This aer is based on the revious work of Aelbaum (Journal of Econometrics, 1979, 9,. 283± 94; 1982, 19,. 287± 99). Iwata (Econometrics 1974, 42,. 947± 66) and Rogers (PhD Dissertation, George Washington University, 1983). We estimate the degree of market ower of an oligoolistic industry, using a linear system and the 2SLS estimation method. Our dearture oint is the work of Aelbaum (1982) where a `market ower index is estimated for each of the samle s 25 years. As the market ower index deends functionally on the conjectural elasticity, the goal is to obtain annual estimates for that elasticity. For this urose, Aelbaum defines a non-linear simultaneous-equation system and obtains, with non-linear methods, the conjectural elasticity estimates for each year of the samle. Considering the conjectural elasticity s functional form that Aelbaum adots, we use a different aroach and obtain a linear system that is easier to estimate. Due to the articular features of the industry analysed, we also derive a much simler form for the equations involved. The model s simlicity is aealing and its generalisation to other industries with homogeneous roducts may be imlemented with ease. Key words: Oligooly; Market ower; Conjectural elasticity. JEL classifications: L13, L Introduction Research on an industry requires knowledge of its features, environment and history. Familiarity with the so-called `basic conditions in which an industry oerates is fundamental since it is desirable that such conditions are accurately reflected in a tractable economic model. This aer involves the estimation of the market ower of the Portuguese milk industry. Because this industry s eculiarities lay an imortant role in the model to be estimated, we find it advisable to first describe both the sector s evolution and its main features. We are indebted for valuable comments from two anonymous referees that led to further significant imrovement on this aer. The usual disclaimer alies. We would also like to exress our thanks to William Collier and Gwyn Lewis for heling us with the editing. AntÂonio BrandÄao, Faculty of Economics, University of Porto Rua Dr Roberto Frias 4200 Porto, Portugal; < abrandao@fe.u.t> /99/ Taylor & Francis Ltd

2 210 M. De Mello and A. BrandÄao The milk industry has no tradition in Portugal. In site of the imortance of milk to human nutrition, its consumtion became relevant only in the 1930s. In 1939, existing technology allowed only for the roduction of butter, Flemish cheese and dehydrated milk. At this time there were about 500 small firms, the majority of which had insufficient technical and hygienic conditions to assure the required minimum roduct quality. Lack of quality and the growing acquisition of cheese and butter from the Azores, did little to enhance the Portuguese mainland roducts, causing milk rices to suliers to dro. This, combined with the insufficient suly of milk to larger cities caused the government to intervene and romote the concentration of the industry, shutting down the firms without adequate conditions, subsidising those which had them, and regulating the rice aid to farmers. Exclusive suly areas to each manufacturing lant were also circumscribed. By 1944, the 54 lants that had survived were able to exand roduction and imrove quality. However, this situation was short-lived. The continuation of World War II had seen the rice of beef rise and the roduction of milk became secondary. Moreover, the lants owners, shielded behind the exclusivity of their suly areas, acted against the farmers, aying them late and below the agreed rice or even not collecting the milk at all. The farmers had no choice but to sell their cattle. Consequently, insufficient milk was sulied to the lants. This lack of raw materials brought anarchy into the industry and the government again had to intervene. In 1948, the government awarded the farmers co-oeratives the legal right to be the exclusive collectors of all the milk roduced. No one else would have direct access to the raw materials. Non-co-oerative firms had to buy the milk they needed from co-oeratives. Soon, the co-oeratives had their own lants and were able to assemble the majority of the farmers into three large cooerative firms that remain dominant today. These three firms artitioned the market, engaging in a territorial imlicit agreement that allowed for their acific coexistence for decades. Excet for certain kinds of cheese, the Portuguese demand for milk and related roducts had been sulied by the home industryð imorts being irrelevant. Within this framework, the Portuguese milk industry flourished until the accession of Portugal to the EC in Due to the lack of cometition, the milk sector was not reared to comete in the oen market. The Portuguese government achieved an agreement to delay the entrance of international cometitors, romising the total liberalisation of the milk suly market (meaning the end of exclusive access to the raw materials for the co-oeratives), until the 1 January Although this might have been a serious blow to the co-oeratives, in the end, their dominance remained ractically intact. The agreement linking the farmers to their co-oeratives is of a symbiotic nature. The farmers romise to sell all their milk to the co-oeratives and the co-oeratives romise to buy it all, whatever the quantities or the quality. Additionally, the co-oeratives suly veterinarian care to the farmers cattle, technical and hygienic advice and financial aid. The rice aid to the farmers for their roduct reflects this understanding and is negotiated every season. When the Portuguese milk market oened u to foreign cometition in 1993, there were 90 co-oerative lants, the majority linked to the `big three, and 50 nonco-oerative lants. In 1992, the `big three reresented 70% of the total milk collected, 80% of the total UHT (Ultra High Temerature) milk roduced, 75% of all butter and 25% of other milk roducts. Uon deregulation, the farmers were offered higher rices for their milk by the new foreign cometitors, but the farmers roved reluctant to change, rejecting the higher rices in favour of their revious bargain. This eretuated the co-oeratives control over raw materials. 1

3 The Portuguese Milk Industry 211 It was only in Setember 1993 that the Italian multinational Parmalat began to sell its milk in Portuguese suermarkets with a rice that was a threat to the establishment. In a few months, Parmalat conquered an imortant share of the market and the `big three began to worry. The so-called `milk war had started and for months the ress and television had something new to add. The `big three accused Parmalat of illegal duming and redatory rices. The Italians accused the Portuguese of collusion and monosonist dominance. Since 1991, merger rumours of the `big three had been sreading but it was only with the entrance of Parmalat and its visible success that the negotiations gained credibility and, in 1995, ended with the integration of the `big three into a `big one. 2 The oinion in some business circles was that the easy entrance of Parmalat was due to the non-cometitive behaviour of the domestic firms. The assumtion was that domestic firms were shielded under informal collusion and governmental rotection, and this exlained the higher rices of their roducts. However, until now, there does not aear to have been any systematic investigation of these issues or any formal roof of collusive behaviour. In trying to clarify this, we decided to obtain an estimate of the Portuguese milk industry s degree of cometitiveness before the full oening of the market. This estimate was obtained by analysing the eriod between 1963 and The Basic Model 3 Aelbaum (1982), defines a non-linear simultaneous-equation system with a market demand function, three factor demand equations and an equilibrium condition written as: (1± u E) = CM(W) (a) Where is the roduct rice, E is the inverse of the demand elasticity, CM is the marginal cost, W is the factor rices vector and u is the conjectural elasticity, a nonobservable variable. If we estimate the system as defined above, u will not vary over time. Since conjectural elasticity cannot be constant, Aelbaum solves the roblem by defining u as: 4 u t = A 0 + A K WK t + A L WL t + A M WM t (b) Here, WK, WL and WM are the rices of caital, labour and materials, resectively. Substituting (b) for u in (a), Aelbaum obtains, with non-linear methods, estimates of the arameters A 0, A K, A L, and A M. These arameters rovide the basis by which the author estimates the annual conjectural elasticity. The linear deendence suggested in (b) (although overcoming the fact that u is a non-observable variable) aears to be somewhat restrictive since conjectural elasticity deends uon the market shares and the conjectural variations, 5 that is, the firms exectations about the effects that changes in their roduction have on the outut levels of their rivals. Given technology and assuming otimising behaviour, it is conceivable that firms choices concerning the level of outut as a strategic variable should include the market conditions. These conditions are associated with multile and laborious asects such as the degree of concentration,

4 212 M. De Mello and A. BrandÄao the tye and level of entrance barriers, the legal and institutional framework surrounding the industry, the imlicit or exlicit agreements firms are engaged in, the diversification and substitutability of the roducts and the demand exansion or retraction cycles, amongst others. All of these may exlain the way oligooly members seculate about their rivals moves and each serves to weaken equation (b) as an accurate exlanation of the conjectural elasticity. Although existing theories are not exlicit as to how such seculations are inserted into tractable emirical models, 6 the roblem of omitted variable bias (namely, those related with the downstream market conditions) and the suggested form for the deendence relation, imose limitations on the aroach of Aelbaum. We therefore adot a different aroach, aying attention to both the main downstream and the ustream variables that affect conjectural elasticity. We resent the basic model below, dislaying only the more reresentative equations. The main features of the model and its formal derivation are consigned to the aendix. Suose an oligoolistic industry with n firms roducing a homogeneous roduct using m roduction factors. Taking firm j and solving for its maximisation roblem we obtain the following equation: (1± u je) = CM j (1) where is the market rice, u j is firm j s conjectural elasticity, E is the inverse of the demand elasticity («being that elasticity) and CM j is firm j s marginal cost. We can define the degree of the industry s market ower such that, Lr = ^ j ± CM j s j = ^ j u jes j (2) where s j is firm j s market share. Lack of data at the firm level restricts us to the basic model s aggregate form. This is resented in the next section, where only the key equations are again dislayed. Formal derivation of the model is once more consigned to the aendix The Aggregate Form of the Basic Model Assuming that the marginal costs are the same for all firms and that u j = u ; j, equation (1) can be rewritten as. 7 (1± u E) = CM (3) and the market ower measure of the industry will now be Lr = u E. If the industry is erfectly cometitive, Lr = 0. If the industry is a ure monooly Lr = E. Hence, both u and Lr rovide information about the industry s degree of cometitiveness. Since we want an estimate of the industry s conjectural elasticity, we rewrite equation (3) as: u = ± CM(w) «(4)

5 The Portuguese Milk Industry 213 To reach the conjectural elasticity value in (4), we need to estimate the demand elasticity and the industry s marginal cost. To obtain these estimates we build a simultaneous equation model with a demand equation, three factor demand functions and an equilibrium condition. The equilibrium condition takes into account that, in an oligooly, rice need not necessarily equate to marginal cost. This allows for the ossible existence of a ositive margin. In the next section we will resent the aggregate model. First, we resent the demand equation then the conditional factor demands and finally the equilibrium relation The Demand Equation. Several henomena influenced the Portuguese milk industry during the samle eriod and some of them had a significant imact on roduct demand. The mid-1970s saw the technological innovation of UHT acked milk introduced to the distribution circuits. 8 UHT milk does not require refrigeration and subsequently allows for rolonged storage. This has ermitted changes in the buying eriodicity which favoured consumtion. The free distribution of milk in Portuguese schools, undertaken by the government since 1974 resents another imortant oint regarding roduct demand. The long advertising camaign (1987± 93), financed by the milk roducers confederation, the Portuguese government and the EC authorities which romoted the slogan `milk is youth rovides another. Assuming the effects of each of these to be manifest in the same direction, and that they have been only relevant since 1977, we establish the ossible change of consumers behaviour through a dummy variable DL, included in the demand equation. 9 We additionally osit changes in the rice elasticity of demand to be ossible over the samle eriod. To allow for this ossibility we insert two additional dummy variables in the demand equation, searating the three decades of the samle eriod. 10 The demand equation is thus estimated as: ln Q = ln d 0 + «ln + d 1 ln PIB + d 2 DL + d 3 (D 1 *ln ) + d 4 (D2*ln ) + v 1 (5) Where is the market rice of milk at eriod t in constant (1990) escudos, Q is the milk consumtion at eriod t in millions of litres, and PIB is the Portuguese Gross Domestic Product at eriod t, in constant (1990) 10 9 escudos. The dummy variables are such that, D1 = 1 if t e [1962, 1970] and D1 = 0 otherwise, D2 = 1 if t e [1971, 1980] and D2 = 0 otherwise, DL = 1 if t P[1977, 1992] and DL = 0 otherwise. The disturbance term v 1 is assumed to be well behaved The Factor Demand Equations. The main industry inuts are raw materials (M), caital (C) and labour (L). These three inuts stand for more than 90% of the total cost. Raw materials alone (essentially milk) account for more than 70%. Assuming a Generalised Leontieff Cost Function for this industry, we write the industry cost function in the following form: C(Q,w) = b C w C + b L w L + b M w M + [b CC w C + b LL w L + b MM w M + 2b CL (w C w L ) + 2b CM (w C w M ) + 2b LM (w L w M ) ] Q (6)

6 214 M. De Mello and A. BrandÄao Since the aggregate factor demand functions for the industry are given by f i = C(Q,w), w i with i = M, C, L, we have: C(Q,w) = f C w C Q C(Q,w) = f L w L Q C(Q,w) = f M w M Q = b 1 C Q + b CC + b CL1 w L w C2 w L2 = b 1 L Q + b LL + b CL1 w C = b 1 M Q + b MM + b CM1 w C w M2 + b CM1 w M w C2 + b LM1 w M w L2 + b LM1 w L w M2 The Portuguese milk industry roduces a multitude of diversified roducts besides lain milk. Since observable data reflect only the industry level, we were not able to gather sufficient information to build an inutð outut table. The available data rovide information about the total quantity of rocessed milk (as an inut), the total amount of acked milk roduced and the total number of tons of all other roducts. Additionally, we have the total cost of all the milk rocessed, the total labour hours used in roduction and the cost of labour er hour, the total amount of caital included and the cost er caital unit, and the milk rice er litre aid to suliers. This information does not ermit calculation of the required amount of milk to roduce one litre of acked milk. We therefore assume that it is necessary to use one litre of lain milk as an inut to roduce one litre of acked milk. If this assumtion is accetable, then the cost of one unit of acked milk is equal to the cost of transforming that unit of milk. As the roduction factors in one milk acket are labour, caital and milk, and since we know the cost of labour and caital er manufactured litre and the cost of milk er unit, the cost of one litre of acked milk is, given this assumtion, exactly the cost of rocessing that litre of milk. 11 Since in equations (7)± (9) we are now dealing with costs of roduced milk and not the cost of milk sulied to the market, we substitute Q with the variable PRM, that is, the total amount of milk rocessed by the industry each year. 12 There are additional consequences from this assumtion. For instance, in equation (9), arameter b MM will now be 1, ; t, and all the other arameters in that equation will be zero. 13 Given these changes, equations (7) ± (9) will now be: f C PRM = b 1 CC + b C CL1 PRM + b w L w C2 f L PRM = b 1 LL + b L CL1 PRM + b w C w L2 (7) (8) (9) (79 ) (89 ) f M PRM = 1 (99 )

7 The Portuguese Milk Industry The Equilibrium Condition. As we are dealing with an oligooly we have to acknowledge the ossible existence of a ositive margin in the model. Suose the following equilibrium equation: 14 (1± m9 ) = where C(PRM,w) PRM (10) C(PRM,w) PRM = b CCw C + b LL w L + b MM w M + 2b CL (w C w L ) + 2b CM (w C w M ) + 2b LM (w L w M ) is the industry s marginal cost, and m9 is an alternative form for the mark-u, being 0 m Since m9 deend on both the market rice and the marginal cost, and since the marginal cost deends, essentially, on the factor rice vector w, it seems ossible to consider m9 as deending on and w, such that m9 = g(, w), is aroximately given by an equation with the following form. 16 m9 = a C w C + a L w L + a w M M or m9 = 1 (a Cw C + a Lw L + a Mw M ) (11) Substituting (11) into (10) and assuming, as above, that b CM = b LM = 0, we have: = (a C + b CC )w C + (a L + b LL )w L + (a M + b MM ) w M + 2b CL (w C w L ) (12) An alternative and more elaborate aroach towards the estimation of the conjectural elasticity is rovided by Aelbaum (1982). The aroach inherent in equation (11) has the advantage of making the emirical estimation of the model easier and is articularly alicable to the industry under consideration. Hence, the comlete system to be estimated is given by equations (13)± (17): ln Q = ln d 0 + «ln + d 1 ln PIB + d 2 DL + d 3 (D1*ln ) + d 4 (D2*ln ) + v1 (13) f C PRM = b 1 CC + b C CL1 PRM + b w L w C2 f L PRM = b 1 LL + b L CL1 PRM + b w C w L2 + v2 (14) + v3 (15) f M PRM = 1 (16) = (a C + b CC ) w C + (a L + b LL )w L + (a M + b MM )w M + 2b CL (w C w L ) + v5 (17)

8 v(3) 216 M. De Mello and A. BrandÄao The variables Q, PRM,, f r, f C and f M, are endogenous, all the others being exogenous. Equations 13± 15 and 17 are stochastic because of otimisation errors so, eriod t disturbance terms in each of these equations is v(i) where i = 1..., 4 and t = 1,...,30. The vector of disturbances is: 3 v(1) t t4 v(2) t v t = N (0; V ) v(4) t where the variance and covariance matrix V is non-singular and such that: E [v(i) t v(j) t ] = s 2 I if i = j = 0 if i ¹ j and i, j = 1,..., 4 E [v(i) t v(j) t± s ] = 0 if s ¹ 0, ; i, j Equation (17) deends exclusively on exogenous variables. If it is assumed that the quantity of milk roduced is unrelated to the quantity of milk demanded, then equation (13) will be indeendent of any other equations in the system. This assumtion seems reasonable given the characteristics of the industry mentioned earlier, namely that the co-oeratives were, until 1989, the exclusive buyers of the farmers milk. Although this legal advantage has been lost in recent years, the cooeratives remain resonsible for the majority of milk collection. Milk collected by other firms is manifestly insufficient since they systematically deend on the surlus of the co-oeratives for their unfulfilled needs. It is also known that the co-oeratives have the obligation of collecting all the milk their suliers roduce, regardless of its quantity or quality. Hence, we must conclude that all the milk roduced by the farmers is collected by the firms. The milk is brought to the lants and transformed into several roducts. The milk surlus of the cooeratives, not sold to other firms, will also be transformed, either by being submitted to a drying rocess or changed into butter (the only ways of storing milk). It seems ossible to resume therefore that all the milk collected is changed either into vendible or into storable roducts. Hence, the milk quantity rocessed by the industry will have more to do with the amount sulied by the farmers than with the amount consumed by the market. Equally, it seems reasonable to assume indeendence between the milk quantity rocessed by the industry and the milk quantity bought by the Portuguese. This being true, equation (13) can be estimated indeendently, with OLS, whilst the others must be estimated simultaneously, with 2SLS. The system of equations dislayed above is, then, a recursive one. 3. The Estimation Statistical Results The results obtained for equation (13) are as follows: ln Q = (0.0600) ± (0.1053) ln lnPIB DL (D1*ln) (0.0804) (0.0402) (0.0151)

9 The Portuguese Milk Industry (D2*ln ) (0.0101) R 2 = 0.98 F = 229 DW = 2.29 where the figures in arentheses are the standard errors. These results confirm demand elasticity variability. In the eriod 1962 ± 70, the elasticity estimate is ± 0.22, in the eriod 1971± 80, it is ± 0.24 and, in the eriod 1981± 91 it is ± The simultaneous estimation results of equations (14), (15) and (17), are the following: Ãf C PRM = ± ± (0.4463) ( ) R 2 = 0.894, F = 114 Ãf L PRM = ( ) ( ) R 2 = 0.899, F = 120 Ã = ± (5.6431) w C ± ( ) R 2 = 0.991, F = PRM ( ) 1 1 PRM ( ) w L w C2, w C w L2, w L w M [2(w C w L ) ], ( ) ( ) Using the estimates of b CC as ± , b LL as and b CL as , we estimated the values of the industry s marginal cost over the thirty years of the samle. With these values and the demand elasticity estimates, we obtained the conjectural elasticity and the market ower index estimates. The market rice and the estimates of the marginal cost, market ower index and conjectural elasticity are shown in Table 1. Table 1 shows that all the estimated values for the conjectural elasticity and the market ower index are ositive, which identifies an imerfectly cometitive market structure and the resence of market ower. However, those estimates must be significantly above zero to allow for more definite conclusions. To test the relevant hyotheses for the resence of erfect cometition and the absence of market ower, we used the samle mean estimates of u and Lr and their standard deviations shown below: Samle Mean Estimates Ù u Ù Lr Estimates Estimate Standard Deviations For a 5% level of significance, the erfect cometition hyotheses (H 0 : u = 0) is rejected with the t statistic samle value The absence of market ower hyothesis (H 0 :Lr = 0) is also rejected with the t statistic samle value of

10 218 M. De Mello and A. BrandÄao Table 1. The market rice and the estimated values for the marginal cost, market ower index, and conjectural elasticity of the Portuguese milk industry Market rice ÃCM t ÃL r à u t Therefore, both tests are consistent with the resence of market ower and an imerfectly cometitive market structure. With the figures for estimates of MC, u and Lr rovided in Table 1, we can ask if they resent a coherent icture of the industry and whether we should just look at the overall average or ay attention to sub-eriod trends or even individual values. Although the overall average trend can give some exlanation about the behaviour of the industry along the 30 years of the samle there are other interesting features we should ay attention to. For instance, we should oint out two distinguishable eriods in Table 1 where imortant events lead to some uheaval in the milk sector: first; the eriod between 1975 and 1979, following the revolution of Aril 1974 when democracy was attained nationally; second, the eriod between 1986 and 1991 following the accession of Portugal to the EC. The first eriod is characterised by left-wing government legislation rotecting the co-oeratives and reinforcing their role in the industry. In this eriod, milk roduction rose 13.5%, on average, er year as consumtion grew sharly. The number of co-oeratives linked to the `big three increased whilst the imlicit agreement of market artition took lace. The market rice of milk fell due to

11 The Portuguese Milk Industry 219 government subsidies designed to suort growing demand and imlant a durable habit of milk consumtion among the Portuguese oulation. The co-oeratives were faced with agreeing to higher rices and further technical aid to farmers in resonse to the increase in demand. The average wage of the manufacturing sector also rose sharly between 1974 and 1978 with the introduction of the minimum wage and higher wage claims from the recently formed unions. Higher inut rices combined with the use of obsolete roduction techniques justify the marginal cost growth we observe in Table 1. The second eriod relates to the threat of new cometitors entering the domestic market following the accession of Portugal to the EC. The immediate results of Portuguese accession were that milk rices aid to the farmers should equalise EC rices 17 and that liberalisation of the milk suly market should take lace. While the rices aid to the farmers raidly attained stagnation, the liberalisation of the milk suly market was only fully comleted by 1 January The co-oeratives had time to invest strongly, and did so, not only in new roduction techniques but also in the imrovement of asturage, cattle artificial insemination, modern milking facilities and more rational and cheaer ways of collecting the milk from their suliers. In 1989, uon deregulation of the suly market, the co-oeratives could verify that the loss of their revious legal rivilege had little effect on their dominant osition. Anticiating the strong cometition they would have to endure from 1990 onwards, and relying uon their dominant osition in the suly market, the co-oeratives increased the rice to consumers, in an attemt to obtain as much rofit as ossible before the entry of new firms. The year of 1990 went by and no new firms were installed. The co-oeratives knew that new entrants would find it difficult to obtain raw materials and, relying uon this advantage, increased the market rice of milk further. Only in 1993 did Parmalat start selling its roducts in the Portuguese domestic market, triggering the so-called `milk war. Between 1993 and 1994 the co-oeratives lost a market share of around 12% to Parmalat while the market rice fell by aroximately 7%. The resonse of the co-oeratives was to merge. This was fully accomlished in January Conclusion Taking Aelbaum (1982) as a dearture oint, we estimated the degree of cometitiveness of the Portuguese milk industry in the eriod before the market was fully oen to foreign cometition. Due to both the eculiarities of this industry and the different aroach we used to define conjectural elasticity, we were able to obtain a simultaneous linear equation system that was easier to estimate. Since our hyothesis was that a constant demand elasticity over the samle s 30 years was imrobable, we allowed for its variability by adding two dummy variables to the demand function. 18 However, if a linear demand equation had been used instead of a Cobb± Douglas, a changing demand elasticity could be derived from the demand equation without additional variables. The motive for this work aeared when a huge debate, known as `the milk war, arose in Portugal during The central issue of this debate was the easy way in which foreign firms entered the Portuguese milk market. Some critics sustained that the entrants success was due to a long ractice of market ower exloitation by domestic firms. Others believed that the ractice of redatory rices and illegal

12 220 M. De Mello and A. BrandÄao duming by foreign firms was to blame. The degree of market ower underlies both such claims. Thus, measuring the market ower of the industry aeared imortant, if one wanted to hold a valid oinion. Our results corroborate the resence of an imerfectly cometitive market structure and the exloitation of market ower by domestic firms. These results were exected since they rove comatible with detectable signs of collusive behaviour. These signs include the governmental rotection awarded to cooeratives for decades; their imlicit territorial agreement; a low milk demand elasticity (confirmed through the estimates obtained); the concerted actions of cooeratives concerning the advertising of their main roduct; control over raw materials and, finally, merger negotiations initiated as the market was oened. There is not, at resent, sufficient information to examine, in detail, the events following the merger of the `big-three but we can confirm from data currently available that Parmalat lost roughly 5% of its reviously conquered market share. Further research on this issue can be undertaken as soon as more data become available. Notes 1. `Oligosony is not an accurate concet to describe this situation. The link between the farmers and the co-oeratives has more to do with mutual trust than with merely commercial interests. Although farmers and co-oeratives are different legal entities, their historical and contractual bonds suort a comlex structure that could classify the industry as vertically integrated. 2. The `big three merger negotiations contemlated two stes. The first, comleted in January 1995, concerned only the distribution and the milk collecting circuits. The second, finished in January 1997, concerned the roduction rocess. 3. This basic model follows that of Aelbaum (1982), although the way in which it is derived could be found with other authors, namely Iwata (1974) and Rogers (1983). 4. See Aelbaum (1982: 293): `In general, u will not be constant but a function of various relevant variables. 5. If we consider the aggregate model, the conjectural elasticity will deend on the rice± cost margin and on the demand elasticity. Either way, the conjectural elasticity deends on the market conditions: ustream, by the factor rices and, downstream, by the roduct rice, market shares and demand elasticity. 6. Even if they were exlicit, the obstacles that inevitably would aear, related to the data collection on the multilicity of variables involved, would be difficult to overcome. 7. We follow Aelbaum s argument `... it is clear that, if marginal costs are the same for all firms then, in equilibrium, the conjectural elasticities must be the same as well Aelbaum (1982: 292). 8. The technological innovation of UHT milk had more imact on the demand side than on the suly side as the costs of change in roduction rocess seemed negligible. UHT milk only requires a new kind of ackaging and to be heated to higher temeratures and for longer than other tyes of milk (for examle, asteurised milk). 9. Although initiated in 1974, the school milk rogramme only covered a significant number of uils in the school year 1976/77. We also had to consider that the UHT milk weight in the total milk consumtion was already 15% in 1978, reaching 70% in Consequently, concerning the dummy variable DL, we divided the samle eriod in two: before 1977 and thereafter. 10. We tried alternative sub-eriods but those we selected guaranteed the best results. 11. Suose, for examle, that we have 1000 L of milk and that 5000esc is the cost of transforming that quantity of milk into 10 kg of cheese, 5 kg of butter and 300 L of acked milk. Given this assumtion, we know that, to roduce 300 L of acked milk, we need 300 L of milk, and the total transformation cost associated is 1500esc. We ignore the quantity of milk necessary to roduce 1 kg of cheese or 1 kg of butter but we can say that 700 l of milk are necessary to roduce 10 kg of cheese and 5 kg of butter, and we can add that the total roduction cost of those goods is 3500esc. 12. This new variable PRM will now substitute the variable Q in every equation that concerns the suly side of the model.

13 P The Portuguese Milk Industry This is lausible since the quantity of milk used to roduce one litre of acked milk does not seem to deend on the rice of milk when related to those of labour or caital. Whatever these rices are, the quantity of milk used to obtain 1 L of acked milk will deend, exclusively, on the technical roduction rocess. So, it seems imossible to substitute the milk factor for any other factor, even if their rices were relatively lower. The ossible variability of inut± outut coefficients should not affect acked milk since it is our belief that, to roduce 1 L of acked milk, 1 L of inuted milk will always be needed. We also assume efficient roduction and, in this case, the tendency to waste should be zero. 14. A similar assumtion, but using the mark-u definition in a slightly different way, can be found in Rogers (1983). 15. The mark-u is often considered to be some quantity above the unitary cost that is usually exressed by the equation = (1 + m)cm. This mark-u, m, is based on cost and ranges from [0, ` ]. So, considering costs as strictly ositive, if CM =, then, m = 0 and if CM 0, then, m `. In our view, we look at rice instead of cost to define `mark-u m9. We could call it `mark-down, for we make use of the fact that the unitary cost is, generally, a fraction of the rice. Consequently, m9 ranges from [0, 1] since, if CM > 0 Þ(1 ± m9 ) > 0 Þ (1± m9 ) > 0 Þm9 < 1 and if CM Þ (1 ± m9 ) Þ 1 1 ± m9 Þ m As mentioned in note 7, Aelbaum (1982: 293) claims that the conjectural elasticity deends on various relevant variables and states also that `... its equilibrium value, as can be seen from equation (13), will deend on the exogenous variables. Aelbaum s equation (13) is (1± u E) = C(W) and, based on it, Aelbaum assumes that u could be aroximated by the linear function u = A 0 + A K W K + A L W L + A M W M. As we see it, even taking into consideration Aelbaum s equation (13) alone, u deends not only on the exogenous variables but also on an endogenous one,. 17. In Aelbaum s model, changes of the demand elasticity over time are not taken into consideration. We assume variability of the milk demand elasticity and our estimates confirm that is not advisable to overlook this issue. 18. Portuguese milk rices to suliers were greater than the EC rices in References Aelbaum, E., `Testing Price-taking Behaviour, Journal of Econometrics, 1979, 9,. 283± 94. Aelbaum, E., `The Estimation of the Degree of Oligooly Power, Journal of Econometrics, 1982, 19,. 287± 99. Iwata, G., `Measurement of Conjectural Variations in Oligooly, Econometrica, 1974, 42(5),. 947± 66. Rogers, R., `The Behaviour of Firms in a Oligoolistic Industry. A Study of Conjectural Variations, Unublished Ph.D. Dissertation, George Washington University, May Aendix The Basic Model Suose an oligoolistic industry with n firms roducing a homogeneous roduct using m roduction factors. Let the cost function of firm j be: C j = C j (q j, w) (A1) where q j is firm j s outut and w is the factor rices vector. Let also the market demand function be: Q = D(, V) (A2) of which derivative with resect to the market rice is suosed to be negative for all > 0. In equation (A2), Q is the total industry outut, such that Q = S q j is the industry suly, and V is a vector of exogenous variables that affect demand. Assuming maximising behaviour and defining firm j s rofit as j = R j ± C j where R j and C j are, resectively, firm j s total revenue and total cost, its rofit maximising roblem is: Max [q j ± C(q j, w), subject to Q = D(, V)] (A3)

14 222 M. De Mello and A. BrandÄao The first order condition for this roblem s solution requires that firm j s marginal revenue (RM j ) equals its marginalcost (CM j ). Consequently, we have the equation: (1± u je) = CM j (A4) = (1) Where u j is the firm j conjectural elasticity concerning the industry outut such that: u j = dq q j dq j Q 1 = dq i dq i dq n q i dq j dq j dq j2 Q 1 = dq k q j 1 + ^ k ¹ j dq j2 Q = (1 + g j) q j Q (A5) dq k where g j = ^ is firm j s conjectural variation and q j k ¹ j dq j Q = S j is firm j s market share. In equation (A5), the conjectural elasticity definition for firm j (which includes both the market share and the conjectural variation), stands for the exected ercentage change in firm j rivals oututs as a reaction to a 1% change in firm j outut. In the Cournot behaviour case, dq/dq j = 1 (since dq k /dq j = 0, ; k, j; k ¹ j), and u j will simly be the market share (u j = s j ). In the erfect cometition case, u j = 0 (since g j = ± 1) and rice equals the marginal cost. In the ure monooly case, u j = 1, since there is only one firm imlying q j = Q and dq/dq j = 1. Hence, the monooly rice± cost margin equals the inverse of the demand elasticity. Rewriting equation (A4) as ± CM j = u je, we can define the market ower degree of firm j as the Lerner index l j such that: l j = ± CM j = u je (A6) Multilying both sides of equation (A6) by the market share s j and summing u over j, we will have the industry market ower degree Lr, such that: ^ j l j s j = ^ j ± CM j s j = ^ j u jes j = Lr (A7) = (2) The Aggregate Form of the Basic Model Assuming the usual aggregation condition that the marginal costs are the same for all firms and also assuming that u j = u for all j, equation (A4) can be rewritten as: (1 ± u E) = CM (A8) = (3) Since u j = u and S s j = 1, the market ower measure will now be Lr = u E. The estimate for the industry s conjectural elasticity, will be obtained from the equation (A8) rewritten as: u = ± CM(w) «(A10) = (4)