Are Cartel Laws Bad for Business? Evidence from the UK

Transcription

1 Are Cartel Laws Bad for Business? Evidence from the UK George Symeonidis University of Essex Abstract. This paper examines the impact of cartel policy on firms' profits using a panel data set of UK manufacturing industries over The introduction of cartel laws in the UK in the late 1950s caused an intensification of price competition in previously cartelised manufacturing industries, but it did not affect those industries which were not cartelised. The econometric results from a comparison of the two groups of industries suggest that the UK cartel legislation had no significant impact on firms' profits, although it had a strong effect on market structure. These results are in line with theoretical models that endogenise market structure by means of a free-entry condition, such as Selten (1984) and Sutton (1991). Keywords: Cartels, price competition, market structure, profits, UK manufacturing. JEL classification: L1. I would like to thank Amanda Gosling, Sajal Lahiri, and seminar participants at the Universities of Essex and Maastricht and the 1999 EARIE conference for helpful comments. Adress for correspondence: Department of Economics, University of Essex, Wivenhoe Park, Colchester CO4 3SQ, U.K. [email protected]

2 1. Introduction. Recent advances in oligopoly theory have shed new light on the much debated issue of the links between firms' conduct, market structure and market performance. In particular, a number of models have examined how changes in conduct may affect market structure and performance. In a seminal paper, Selten (1984) predicted that a switch from collusive to noncollusive behaviour caused by a toughening of competition policy would have no adverse effect on firms' profits, although it would cause a decrease in the number of firms in a homogeneous good industry. The reason for this result is that the number of firms in an industry is determined by a free-entry condition which requires that net profit is driven to (almost) zero by entry irrespective of firm conduct. This framework was recently enriched in Sutton's (1991, 1997, 1998) comprehensive theory of market structure, which distinguishes between exogenous sunk cost industries and endogenous sunk cost industries and examines the links between market size, sunk costs and concentration in the two types of industries. This paper provides an econometric analysis of the impact of the intensification of price competition following the introduction of cartel laws on firms' profits. A unique opportunity to study this issue is given by a "natural experiment" that occured in the UK in the 1960s. As a result of the 1956 Restrictive Trade Practices Act, restrictive agreements between firms, covering a wide range of industries, were cancelled. This caused an intensification of price competition in many industries during the 1960s. These can be compared to a "control" group of industries which had not been subject to agreements significantly restricting competition and were therefore not affected by the Act. The econometric results from the analysis of a panel data set of manufacturing industries suggest that the intensification of price competition following the 1956 Act had no significant effect on profits, while it had a strong negative effect on the number of firms. These results are in line with the predictions derived from the Selten-Sutton theoretical framework. They suggest that, in long-run equilibrium and in the absence of any institutional barriers to 1

3 entry, cartels do not result in higher profits, but rather they allow for excessive entry (and/or insufficient exit). The long-run effect of cartel policy is then to reduce the number of firms, rather than their profits. The paper is organised as follows. The next section presents a theoretical framework based on Selten (1984) and Sutton (1991). Sections 3 and 4 describe the evolution of competition in UK manufacturing and the construction of the data set. The econometric model and results are presented in section 5, and the final section concludes. 2. Theoretical framework. Consider an industry in which the only significant sunk costs are the exogenously determined costs of setting up a plant of minimum efficient scale. A two-stage game can be used to model competition in such an industry. At stage 1 firms decide whether or not to enter at a given sunk cost f. At stage 2 those firms that have entered set prices. The equilibrium outcome of the second-stage subgame can be represented by a vector of (gross) profits Πi(S, h, U, t, c 1,..., c i,..., c N), where S is market size, an exogenous demandshift parameter, h is an index of horizontal product differentiation, U is an index of union power, t is a measure of the intensity of price competition, c i is a vector of parameters specific to firm i (which may include marginal cost, capacity, the number of varieties offered, or the number of plants operated), and N is the number of firms that have entered at stage 1. The parameter t captures the idea that, for any given number of firms N, Πi will depend on the firms' pricing strategies, which will in turn partly depend on exogenous institutional factors, such as the climate of competition policy. In fact, t can be thought of as an inverse measure of the "degree of collusion". 1 It is not, however, equivalent to the price-cost margin, which is 1 It is well known that, under certain conditions, any individually rational and feasible payoff vector can be sustained as an equilibrium of an infinitely repeated pricing game (Fudenberg and Tirole 1991). It seems natural to assume that the climate of competition policy will considerably 2

4 endogenous. Consider now the benchmark case of symmetric single-plant single-product firms. The gross profit function can in this case be written as Πi (S, h, U, t, N). We assume Πi/ S > 0, Πi/ t < 0, Πi/ h > 0, Πi/ U < 0, and also that Πi is decreasing in N. These assumptions are quite plausible, at least for a situation where total sales revenue in the industry depends mainly on market size S, and the marginal cost curve is relatively flat and not much affected by the exogenous variables of the model. For example, in the benchmark case of constant marginal cost and isoelastic industry demand, these assumptions can be easily reformulated in terms of some very plausible hypotheses involving price p. In particular, a sufficient condition for Πi/ S > 0 is p/ S 0; a sufficient condition for Πi/ t < 0 is p/ t < 0; and a sufficient condition for a negative relationship between Πi and N is that price is non-increasing in N, an assumption that has received considerable theoretical and empirical support. Finally, Πi/ h > 0 is a standard result in models of product differentiation, and Πi/ U < 0 is a common result in models of firm-union bargaining and also in empirical studies of the effects of unions. At stage 1 there is free entry, so the equilibrium level of N is given by the largest integer such that Πi(S, h, U, t, N) f, i. Assume that this value of N is unique (this will be the case if the average cost curve of each firm is either U-shaped or everywhere declining). It is then easy to see that an increase in the intensity of price competition t will cause a fall in the equilibrium number of firms N*: gross profit, which has fallen following the increase in t, can only be restored to a level at least equal to f through a fall in N. However, net profit Π* - f will not necessarily fall, and indeed it may rise. Provided that the number of firms is not initially very small, one can expect that net profit will be approximately zero, since the effect of the integer constraint will be minimal. In these circumstances, the rise in the intensity of price competition will have no significant effect on net profit. Since f is assumed exogenous, a affect the probability of any particular outcome being realised (for instance, making collusion illegal will make coordination more difficult and will also increase the expected costs of collusion). 3

5 rise in t will also have no significant effect on gross profit. Note that if the integer constraint is taken into account, little can be said about the precise effect of a rise in t on profit without imposing more structure on the model. Selten examines a specific model and shows that, under certain conditions, both total industry net profit and plant (or firm) net profit are more likely to increase than to decrease following a switch from a collusive to a non-collusive regime (see also Phlips 1995, chapter 3). However, in a context where the integer constraint is the only reason for positive net profit, this can be expected to be small in general, at least at the plant or firm level, so any change may be difficult to identify empirically. Since there is no clear general prediction as to the direction of the change either, it seems legitimate to consider a weaker version of the Selten result as the main testable prediction of the model, namely that a switch of competition regime has no significant effect on plant (firm) net profit or on plant (firm) gross profit, while the effect on industry net profit is ambiguous (however, industry gross profit should fall because of the fall in the number of firms). Clearly, these results depend on two crucial assumptions. The first is that firms under a collusive regime cannot prevent entry of new firms into the industry. It is possible, however, to construct models of collusion where firms adopt trigger strategies that deter entry (Harrington 1989, 1991), or where the collusive strategies result in supranormal profits even though entry is accomodated (Friedman and Thisse 1994). In such models, a decrease in the "degree of collusion" will reduce net and gross profit and will have an ambiguous effect on the number of firms (assuming that the scope for entry deterrence is smaller in a more "competitive" equilibrium). Hence an empirical test of the Selten-Sutton predictions may also be interpreted as a test of alternative theories of collusion. The second key assumption is the symmetry assumption. However, most industries are subject to significant asymmetries, due to a variety of factors including multi-plant or multiproduct firms, or efficiency differences between firms. In the presence of asymmetries, free 4

6 entry is consistent with supranormal profits for all but the marginal firm in an industry, and it is not clear whether any general theoretical predictions can be derived about the long-run equilibrium effect of an intensification of price competition on profits of the average firm. Results from specific models suggest that the effect of tougher price competition on firm profit and even the price-cost margin can be positive in the presence of efficiency differences even when no account is taken of the integer constraint (see Montagna 1995). The intuition is that an exogenous shock which reduces prices in the short run drives the less efficient firms out of the industry, so that at the new long-run equilibrium price may have fallen less than the marginal cost of the average firm in the industry. I have so far focused on the case of an exogenous sunk cost industry. Competition in an industy with significant endogenous sunk costs can be analysed, following Sutton (1991), as a three-stage game (see also Symeonidis 1999a, 1999b). At stage 1 firms decide whether or not to enter at a given sunk cost of entry f. At stage 2 each firm i chooses to incur a sunk cost Ai, which increases the consumers' willingness to pay for the firm's variety or reduces marginal cost. Ai may represent advertising or R&D expenditure. Finally, at stage 3 firms set prices. Assuming single-product single-plant firms and no asymmetries other than in the Ai's, the equilibrium gross profit of firm i in the third-stage subgame can be written as Πi(Ai, A-i, S, h, U, t, N) Ai + f, i, and is again assumed to be increasing in S and h and decreasing in U, t and N. Moreover, Πi/ Ai > 0, since an increase in Ai increases the consumers' willingness to pay for firm i's product or reduces its marginal cost. The equilibrium number of firms N* and levels of endogenous sunk costs Ai* in this model are determined by a free-entry condition and a set of first-order conditions for the optimal choice of the Ai's. The effect of a change in t on N* is not necessarily negative, because it also depends on the direction and the magnitude of the effect of t on the Ai*'s. Whatever happens to N*, however, the effect on net profit Πi* - Ai - f will still be (approximately) zero. Note, however, that, unlike the exogenous sunk cost case, the gross 5

7 profit Π* may change significantly, for given f, because of the change in Ai, and the direction of that change is not predictable in general. Two final remarks are in order. First, the use of a two-stage or three-stage game to model what is essentially a dynamic process of repeated interaction may be seen as a limitation of the present theory. This formulation captures, however, the notion that certain long-run decisions constitute a commitment at the time when short-run decisions are taken; hence at any point in time some costs can be thought of as sunk, even though firms in practice incur a continuous stream of expenditures. It should also be emphasised that the capital stock created through sunk expenditures depreciates and has to be renewed. This implies that the driving mechanism of the theoretical model, namely the requirement that gross profit is restored to a level that covers fixed (and sunk) costs, is entirely relevant in the present empirical context: the fact that the capital stock is already in place when the exogenous change in the institutional framework occurs is not relevant, at least for the medium or the long run, because capital depreciates and must be replaced. Second, the assumption of exogeneity of the intensity of competition t is clearly a simplification. It is, however, probably justifiable in the present context for two reasons. First, the key determinant of changes in t during the period under study was the exogenous change in the institutional framework (see section 3 below). Second, cartelisation in UK manufacturing in the 1950s, i.e. the initial value of t, seems to have been a function of exogenous industry-specific factors rather than endogenous variables like concentration or the number of firms (see Symeonidis 1998a and the discussion in section 5). In any case, to the extent that t is observable and the increase in t can be established empirically, it would be easy to derive theoretical predictions conditional on the known change in t, even if t were made endogenous. Suppose that t is a function of the exogenous institutional variable T and a vector of other variables Z,_, which may include N, and consider, for simplicity, the case of an exogenous sunk cost industry with symmetric single-plant firms. N* will then be given by the 6

8 largest integer satisfying Πi[S, h, U, t(t, Z,_), N] f, i, and two conditional predictions follow: (i) a change in T must cause N* to fall if and only if t is larger at the new equilibrium; (ii) since an observed fall in N* following a change in T implies that t must have risen, an observed fall in N* allows us to test the prediction of no significant effect of the increase in t on plant net or gross profit. 3. Competition in UK manufacturing industry. At the time the 1956 Restrictive Trade Practices Act was passed, nearly half of UK manufacturing industry was subject to explicit agreements significantly restricting competition. These were not enforceable at law, but they were not illegal. As a result of the legislation, these agreements were abandoned. This section briefly describes the evolution of competition in UK manufacturing from the 1950s to the early 1970s (a more detailed discussion can be found in Symeonidis 1998b, 2000). The 1956 Act required the registration of restrictive agreements between firms on goods, including both formal, written undertakings as well as informal, verbal or even implied arrangements. Registered agreements were presumed to be against the public interest and should therefore be abandoned, unless they were successfully defended in the newly created Restrictive Practices Court or considered by the Registrar of Restrictive Trading Agreements as not significantly affecting competition. The Act did not then at once make restrictive agreements illegal: an agreement could be upheld if the Court was convinced that it produced positive benefits which outweighed the presumed detriment. Since the attitude of the Court could not be known for certain until the first cases had been heard, the large majority of the existing agreements were registered rather than being immediately dropped or secretly continued. The hard line taken by the Court, however, especially in its initial judgments, induced most cartels to voluntarily abandon their agreements. Of those which were defended in the Court, only a few were upheld. As the first Court cases were heard in 1959, it was not 7

9 until 1959 that industries, on the whole, started cancelling their agreements. A large number of agreements contained minimum or fixed producer prices, conditions of sale, and often ancillary restrictions such as collective exclusive dealing or the maintenance of common resale prices. Some agreements, however, only contained restrictions which were probably much less significant for competition, e.g. conditions of sale. In general, there were no restrictions on media advertising or R&D expenditure. Free entry was a key element in the model of the previous section, so a crucial question is whether entry was restricted in cartelised industries. The evidence from the agreements registered under the 1956 Act, the early Monopolies Commission reports (cf. Guenault and Jackson 1974, Rowley 1966) and the case studies in Swann et al. (1973) suggests that this was not the case. Although practices such as collective exclusive dealing and aggregated rebates were often used by the cartels as a way of limiting outside competition, it is not at all clear that these practices also restricted entry. In most industries the restrictive agreements were operated by trade associations and there were usually no significant restrictions on association membership, so that entry would not be difficult if the entrant was willing to become a party to the agreement. In some industries, on the other hand, the existing association members might reject some applications for membership, even though they would usually accommodate any powerful non-member firm; this could happen in industries where the association firms had some control over distribution channels, usually through agreements with distributors' associations. Even in such cases, entry in the industry would not be restricted unless the barriers to outside competition were fully effective. The above sources contain some information on profitability of firms in cartelised industries. Profits were thought to be more often "reasonable" than "excessive". Also, there were typically marked variations in costs and profits across firms, and this often meant that the profitability of the less or least efficient was "low". In fact, price-setting by the cartels sometimes consisted in a compromise between high-cost and low-cost firms, with prices set at 8

10 a level that allowed the high-cost firms to break even. The lack of excessive profits does not mean that the agreements were not effective. Case-study evidence suggests that in most industries the agreements had been operated honourably. Also, the effectiveness of outside competition was limited in many industries by the fact that the cartels tended to contain most or all of the largest and best known domestic firms, or by practices aimed at discouraging distributors from buying from outside firms, or because imports were hindered by tariffs, transport costs or the operation of international restrictive agreements. Prices of outside firms were typically lower than the cartel prices, although sometimes they were identical or marginally lower. In some industries prices were being adjusted depending on the extent of outside competition, e.g. in particular segments of the market or for particular contracts. Could it be the case that the existence of a price-fixing agreement is not a good overall indicator of collusive conduct in the UK in the 1950s? The above discussion of the evidence on the effectiveness of agreements does not support this view. However, there is also another possible objection, namely that some at least of the industries subject to agreements could be industries where collusion was difficult to sustain otherwise, while many of the industries without agreements may have been practicing tacit collusion. Firstly, it must be remembered that the Act required "implied" arrangements to be registered as well. Secondly, it is difficult to imagine why colluding firms would not want to enter into an explicit agreement given that collusion was not illegal in the 1950s. Thirdly, there is some evidence from a questionnaire survey of competition in UK manufacturing in the 1950s (Lydall 1958) which does not support the above objection. The study by Lydall did not specifically discuss collusive agreements, but some of the information given suggests that firms that perceived their condition as being characterised by "no strong competition" were primarily in industries which had agreements, while firms which thought that they were facing "strong competition" were chiefly in industries without agreements. 9

11 Case-study evidence (e.g. Swann et al. 1974) also suggests that price competition intensified in the short run in many industries following the abolition of cartels. This took the form of a fall in list prices or an increase in trade discounts. However, in several cases agreements to exchange information on prices, price changes etc replaced the former restrictive arrangements and were usually successful in restricting competition. In these cases price competition usually emerged after the information agreements were abandoned in the mid-1960s, i.e. about a decade after the 1956 Act was passed, following adverse decisions of the Restrictive Practices Court and the inroduction of the 1968 Restrictive Trade Practices Act, which required registration of information agreements. Swann et al. (1974) also point out that in the longer term there were instances of secret collusion in some industries and parallel pricing in others. However, this does not imply that the intensity of price competition did not increase significantly in the longer term in the large majority of industries with terminated agreements. The cases of detected collusion were not numerous. And while parallel pricing was quite common (Monopolies Commission 1973), it should not be seen as evidence for the existence of collusion, especially since in many industries there was competition in discounts and with respect to sales by tender to large buyers, even though list prices tended to move in parallel. Also, the opening of the British economy during the 1960s and 1970s must have significantly reduced the scope for effective collusion by domestic firms, and the extent of collusion must have been further limited by the abolition of collective exclusive dealing and the right of buyers to complain to the competition authorities when they suspected collusion. Thus, while one cannot rule out cases of ineffective agreements or cases of collusion continuing secretly in the 1960s, all the available evidence suggests that such cases would be exceptional and that the majority of industries with collusive agreements in the 1950s did experience, sooner or later, an intensification of price competition as a result of the 1956 legislation. Hence it seems, on the whole, legitimate to think of this evolution as a change of 10

12 competition regime induced by an exogenous institutional change. 4. Construction of the data set. The empirical analysis in this paper essentially involves a comparison of the evolution of profits over between those industries affected by the restrictive practices legislation and those not affected - controlling for other factors that may have influenced profits during the period examined. Although the Act was introduced in 1956, it had little effect before the first Court cases were heard in Moreover, as competition did not break out immediately in several industries, the impact of the Act was felt at least until the late 1960s or early 1970s. Data on profits and firm/plant numbers are available for five different years, namely 1954, 1958, 1963, 1968 and The industry definitions are sometimes at the threedigit level of aggregation and sometimes at the four-digit level. In particular, in most cases they are the "principal products" within any three-digit "minimum list heading" industry, as defined in the individual industry reports of the UK Census of Production. Whenever a "minimum list heading" industry is not further subdivided in the Census reports, it was used as the industry definition for this paper. Because of changes over time in the industry or principal product definitions, the panel is unbalanced. This section describes the construction of the data set. Profits, number of firms, number of plants, market size. Several profit measures are used in the econometric analysis of section 5: the total industry gross profit, the gross profit divided by the number of plants (or firms), and the price-cost margin. All these were constructed using data on gross output, net output, wages and salaries, the number of plants, and the number of firms from the individual industry reports of the Census of Production (various years) and from 1973 Business Monitors. The figures are for all firms employing 25 or more persons. Industry gross profit is defined as net output minus wages and salaries, and was deflated using either the general producer price index, obtained from the Annual Abstract 11

13 of Statistics, or industry-specific producer price indices. Note that the gross profit includes fixed costs, such as advertising expenditure, R&D expenditure and capital costs. Unfortunately, it also includes some variable costs, the most important of which are employers' contributions to National Insurance, pension funds, etc; data on these are not available for most years, and the implicit assumption has been made that these costs are a constant fraction of net output in any given industry or that changes in these costs are more or less homogeneous across industries and will therefore be picked up by time effects. In addition to being used to construct profit measures, the number of firms and the number of plants were used as dependent variables in some regressions. A limitation of using these variables as measures of market structure is that they are sensitive to the number of small firms (or plants). It is well known that small firms often do not produce core industry products. Moreover, the evidence from the various data sources on competition suggests that the British cartels did not usually include all firms in any given industry and it was often the smaller firms who were not cartel members. Hence the effect of the 1956 Act on many small firms in cartelised industries may have been relatively weak. These problems are somewhat alleviated by the fact that very small firms (namely, firms with less than 25 employees) are not taken into account. Other measures of market structure, such as concentration measures, are not available for the industry categories used here. Market size can be measured either by net output or by sales revenue, deflated either by the general producer price index or by an industry-specific producer price index. Perhaps net output reflects more accurately the extent of profit opportunities in a market, and therefore is a better proxy for market size. On the other hand, its use in profit equations may be inappropriate because of endogeneity problems. Prices may also be endogenous, and this suggests using a general producer price index as a deflator, although this does not control for changes in the relative prices of materials. In any case, all these variables were tried in the empirical models of section 5 and gave broadly similar results. 12

14 Competition. The main source of data on competition were the restrictive agreements registered under the 1956 Act. A number of other sources were also used to identify products subject to unregistered agreements or agreements modified prior to registration, including the various industry reports of the Monopolies Commission; the 1955 Monopolies Commission report on collective discrimination; the 1949 report of the Lloyds' Committee on resale price maintenance; industry studies contained in Burn (1958) and Hart et al. (1973); the Board of Trade annual reports from 1950 to 1956; and the Political & Economic Planning (1957) survey of industrial trade associations (including unpublished background material for this survey). The approach to modelling the competition effect in the present paper involved distinguishing between those industries with a change of competition regime following the 1956 Act and those without a change in regime. The first step in this procedure was to classify all industries in the sample according to their state of competition in the 1950s on the basis of three criteria: the reliability of the data source; the types of restrictions; and the proportion of an industry's total sales covered by products subject to agreements and, for each product, the fraction of the UK market covered by cartel firms. In particular, the various types of restrictions were classified as significant, nonsignificant or uncertain, according to their likely impact on competition. Next, the products which were subject to agreements were assigned to the various industry categories. Now certain products within a particular industry were subject to significant restrictions, while others were not. An industry was classified as collusive in the 1950s if the products subject to significant restrictions accounted for more than 50% of total industry sales. It was classified as competitive if the products subject to significant or uncertain restrictions accounted for less than 20% of industry sales. And it was classified as ambiguous in all remaining cases. 2 It 2 In fact, most industries classified as competitive were free from any restrictive agreements. I have used the 20% cut-off point because in some cases secondary industry products were subject 13

15 was assumed, for agreements of nationwide application, that the parties accounted for a substantial fraction of the relevant market. For important regional agreements, an estimate of the fraction of industry sales subject to restrictions was made. All industries with ambiguous state of competition in the 1950s were then excluded from the sample. Finally, the dummy variable CHANGE was defined, which takes the value 1 for industries with a change in competition regime sometime after 1954 (or 1958) and 0 otherwise. An analysis of the competition effect on any endogenous variable y could then be performed by testing whether the time effects on y after 1954 (or 1958) are different for the two groups of industries in a regression that also controls for other factors affecting y. This procedure allows for an evaluation of both the short-run and the long-run impact of the 1956 Act. For the final sample used in this paper, industries with ambiguous state of competition in 1958 were excluded, as were industries for which data on profits were not available at least for two of the three core years 1958, 1963 and 1968 (i.e. industries with data only for or were excluded). The final sample contained 207 industries and 782 observations. Setup cost. There are no data for setup costs, so a proxy was constructed. Ideally, one might want to use a measure of minimum efficient scale multiplied by the total value of capital stock in the industry (see Sutton 1991). However, a measure of m.e.s. based on the size distribution of plants could not be used because of data limitations. On the other hand, a simple measure of m.e.s. such as the size of the average plant as a fraction of industry size (i.e. the inverse of the number of plants in the industry) would be inappropriate in the present case to restrictive agreements, although core industry products were not. Similarly, most industries classified as collusive had agreements covering all industry products. I have used the 50% cut-off point because in some cases most core industry products were subject to price-fixing, though some were not. Small variations in the cut-off points (in particular using 10% instead of 20%, or using 40% or 70% instead of 50%) do not significantly affect the results reported in section 5. 14

16 since it is strongly correlated with the number of firms. The proxy used is the capital-labour ratio. This is sometimes used as a proxy for technical economies of scale, but, admittedly, it is a rather imperfect measure of setup cost. Moreover, the available capital stock figures are estimates rather than primary data and are at the three-digit level of aggregation (i.e. for Census "minimum list headings"). Nevertheless, these problems should not be overemphasised. Since an empirical model with industry-specific effects will be used in section 5, one need not assume that the capital-labour ratio is an adequate measure of f; all that is required is that the change in the capital-labour ratio is an adequate measure of the change in f. This still leaves us with three-digit capital-labour ratios, however, while the rest of the data are often at the four-digit level. One way to proceed is to simply use the three-digit capital-labour ratios on the assumption that changes in this ratio across time should be roughly similar for all four-digit industries within any given three-digit industry. An alternative procedure is to adjust the capital stock estimates on the basis of Census data on the fraction of investment accounted for by each four-digit industry within any given three-digit industry. The adjustment adopted in this paper was to multiply the three-digit capital stock by the ratio of four-digit investment to three-digit investment (averaged over two years). This should produce reasonable approximations of capital stock at the four-digit level. As it turned out, the regression results were similar for both sets of capital-labour ratios, although the use of the adjusted data resulted in larger coefficients and t-statistics on the setup cost proxy itself. This suggests that further refinement of the capital stock estimates would not affect very much the results for the variables of interest, namely those capturing the competition effect. The results reported in section 5 below are those obtained using the adjusted capital stock data. Estimates of the capital stock at the three-digit level of aggregation were taken from O'Mahoney and Oulton (1990). I have used their net capital stock figures and defined capital stock as plant and machinery (i.e. buildings and vehicles were not included as these are largely 15

17 recoverable on exit). Data on employment and investment on plant and machinery were taken from the industry reports of the Census of Production (various years). Finally, a remark on the interpretation of the capital-labour ratio in profit equations may be in order. This variable (or the capital-output ratio) has often been used in profitability studies to control for the fact that the endogenous variable (the price-cost margin or the rate of return on capital) includes the gross return to capital. In the present study, the capitallabour ratio is seen as a proxy for setup cost. This is not a real difference, however, since the setup cost is essentially the cost of installing capital (plant and machinery). Union power. The measure used was union density, defined as the number of unionised employees over the total number of employees. Data on this variable for the period examined here are available at a level of aggregation between the two-digit and the three-digit and were taken from Bain and Price (1980). Advertising-sales and R&D-sales ratios. Each industry was classified on the basis of its typical or average advertising-sales ratio and R&D-sales ratio over the relevant period. The objective was to construct a sub-sample of industries that would exclude advertising-intensive and R&D-intensive industries. The reason is that the theoretical predictions for the effect of price competition on gross profit are not so clear for advertising-intensive or R&D-intensive industries as for exogenous sunk cost industries. Hence regressions were run both for the whole sample and for the sub-sample of industries with ADS < 1% and RDS < 1%, where ADS and RDS denote the advertising-sales ratio and the R&D-sales ratio respectively. The procedure for constructing R&D-sales ratios and advertising-sales ratios is the same as in Symeonidis (1999c); see that paper for details and a listing of the sources used. The subsample of exogenous sunk cost industries contained 138 industries and 518 observations. 5. Empirical model and results. To study the effect of the intensification of price competition on market structure and 16

18 performance, I estimate in this section reduced-form equations of the form y = y ( S, h, f, U, t ), 1 where y is a measure of market structure or a gross profit measure. Note that many of the theoretical predictions of section 2 are for net profit, i.e. gross profit minus fixed costs. Unfortunately, it is not possible to construct measures of net profit with the data available. As pointed out above, the capital stock figures are estimates. While the estimated proportional changes in capital stock over time are reasonably accurate, the estimated levels should be treated with caution (see Oulton and O'Mahoney 1990). As a result, only measures of gross profit can be used. While this is a limitation of the present analysis, it should be emphasised that the Selten-Sutton framework also provides predictions regarding gross profit: a key prediction is that industry gross profit divided by the number of plants should not change significantly following an intensification of price competition. This will not be true if, contrary to what Selten assumes, cartels generally deter entry in long-run equilibrium. Moreover, the presence of asymmetries implies that even total industry gross profit may not be significantly affected by cartel laws. If that is the case, and provided that the number of firms has not risen, cartel laws are definitely not bad for business. Prior to examining the impact of the 1956 Act on profits, I analyse the effect on the number of firms (or plants). This is interesting in its own right and complements the analysis presented in Symeonidis (1999c) for changes in concentration. Also, since the main testable hypothesis about profit is that no significant change should be expected, it is necessary to verify (indirectly, i.e. through the observed change in market structure) that the intensity of price competition did actually increase, on the whole, in the previously cartelised industries included in the present sample. Finally, the results for firm numbers will be useful in interpreting the results for profit, especially since both sets of results allow for a comparison of short-run and long-run effects of the legislation. 17

19 Some descriptive statistics on initial levels in profits and market structure are reported in table 1. The table reports means and standard deviations of six different variables in 1958, i.e. prior to any significant impact of the legislation on competition, for industries which experienced a change of competition regime after 1958 as well as for industries not affected by the legislation (the latter group includes a few industries where the agreements continued). The variables are the number of firms NFIRMS, the number of plants NPLANTS, total industry gross profit PROFIT, industry gross profit divided by the number of plants PLANTPROFIT, industry gross profit divided by the number of firms FIRMPROFIT, and the price-cost margin PCM, defined as net output minus wages and salaries divided by sales revenue. For the first five of these six variables a log transformation is used. On the whole, there is no evidence of any significant difference in initial conditions between the two groups of industries. This may seem puzzling, if one expects that price competition has a negative effect on firm numbers. First, note that the group of industries without a change in competition regime includes a few industries which were collusive in the 1950s. More importantly, any effect of price competition on market structure (or profit) may be difficult to identify in a cross-section of industries because of the prevalence of industryspecific characteristics and the links with other variables. For instance, the intensity of price competition and the number of firms are also negatively associated because of a third variable, namely capital intensity, which has a negative effect on firm numbers (see the regression results below) but also increases the likelihood of collusion (see Symeonidis 1998a). I now turn to the econometric analysis of the determinants of market structure and profits. The specifications used are panel data models with an intercept varying over industries to control for industry-specific effects and factors that are unlikely to change significantly over time (such as the degree of horizontal differentiation or the elasticity of demand). Time dummies are also included among the regressors in an attempt to control for other factors that may have affected market structure and profits over this period, such as changes in the tax 18

20 system in the mid-1960s that are thought to have encouraged mergers, economies of scale in distribution and the raising of finance, the progressive opening of the British economy, the UK government's prices and incomes policies between 1965 and 1973, and macroeconomic fluctuations. It is very difficult to measure most of these factors at the industry level, but it seems plausible to assume that their impact would have been more or less equally realised across all industries. In particular, there is no reason to expect a systematic difference between previously collusive and non-collusive industries with respect to these factors. Hence they should be largely captured by the time dummies. 3 The basic specification for the number of firms is ln FIRMS it = α + β β Y 63 + β ln SALES 6 it Y 68 + β + β β CHANGE * Y 63 + β 7 2 ln( K / L) Y 73 + β 10 8 it + β UNION CHANGE * Y54 + CHANGE * Y 68 + β 3 11 it + β 4 Y54 + CHANGE * Y 73 + u it and similarly for the number of plants. SALES is industry sales revenue deflated by the general producer price index for all manufacturing, K/L is the capital-labour ratio, UNION is union density, Y54, Y63, Y68 and Y73 are time dummies for 1954, 1963, 1968 and 1973 respectively, and the interaction terms should capture any differences in the evolution of NFIRMS or 3 It is difficult to control for the impact of the opening of the British economy on market structure and profits in a more satisfactory way. Ideally, one needs some measure of the extent of foreign competition for each industry across time. Two possible candidates are the import penetration ratio and the rate of effective protection. However, there are serious problems, theoretical and practical, with both of these measures. Estimates of effective rates of protection are available at a high level of aggregation and, in any case, only for some years in my sample. The import penetration ratio, on the other hand, is a poor proxy for the extent of foreign competition, since it cannot capture the effect of the mere threat of competitive imports, it does not take into account imports by domestic producers (which may not be in competition with domestic products), and it is itself endogenous; moreover, the industrial classification used in the foreign trade statistics during the period examined in this paper is not easy to match with the one used in the Census of Production. 19

21 NPLANTS after 1958 between industries with a change of competition regime (CHANGE = 1) and industries without such a change (CHANGE = 0). Thus the coefficient on CHANGE*Y63 (CHANGE*Y68, CHANGE*Y73) measures the effect of the 1956 Act between 1958 and 1963 (1968, 1973). The benchmark year is 1958, as it is generally accepted that the Act had little effect before then. The coefficient on CHANGE*Y54 serves as a partial check of this presumption, as well as a check that the evolution of market structure during was not significantly different between the two groups of industries. The basic specification for each of the four profit measures defined above is profit measure it = α + γ γ Y 63 + γ 5 ln SALES it Y 68 + γ + γ γ CHANGE * Y 63 + γ Y 73 + γ 10 ln( K / L) 8 it + γ UNION CHANGE * Y54 + CHANGE * Y 68 + γ 3 11 it + γ Y CHANGE * Y 73 + ε it where the "profit measure" can be lnprofit, lnplantprofit, lnfirmprofit (all three deflated by the general producer price index for all manufacturing) or PCM, and the other variables are as defined above. 4 Note that this specification is very different from those typically used in "traditional" studies of the link between market structure and profitability. In these studies, the relevant dependent variable is the price-cost margin (or the profit rate), not total industry or firm profit. Moreover, these studies include a measure of market structure among the regressors. The above specification, on the other hand, is a reduced-form equation derived from a theoretical model in which market structure and profit are both endogenous. The model's predictions regarding the effect of a change in the intensity of price competition on profit depend on 4 The denominator of PCM is sales revenue. Some studies (e.g. Hart and Morgan 1977, Conyon and Machin 1991) have used net output as the denominator of PCM on the grounds that sales revenue is influenced by input prices, duties and subsidies, and the degree of vertical integration within an industry. These arguments are more important for cross-section studies than for studies using panel data. In any case, regressions using this alternative definition of PCM gave results similar to those reported here. 20

22 allowing market structure to change to restore the long-run equillibrium. It is therefore important for testing these predictions that one does not control for changes in market structure when specifying the profit equation. A possible objection to the above specifications is that some of the independent variables may be endogenous. This is probably not a serious problem for the market size and setup cost proxies, as the variation in these empirical measures across industries and five-year periods is likely to be mainly driven by the variation in the corresponding theoretical variables. 5 CHANGE obviously depends on whether an industry was originally cartelised or not, and that may be endogenous. It is not possible to test formally for exogeneity since there are no appropriate instruments for CHANGE. The fact, however, that there is no evidence from table 1 of any significant difference in market structure or profits in 1958 between industries affected by the legislation and those not affected is reassuring. In addition, it can be argued that even if CHANGE is influenced by certain variables which also affect market structure or profits and are not included in the model, these variables are more likely to be part of the industry-specific effect than of the error term, thus causing no econometric difficulties. Hence, there seems to be no reason to believe that endogeneity of CHANGE may be a serious problem. 6 5 Whether the direct effect of a change in the competition regime is to increase or reduce sales revenue will depend on the elasticity of demand. Although a joint monopoly would operate on the elastic part of the demand curve, its abolition may well move the industry to the inelastic part and sales revenue may rise or fall. Moreover, collusion need not be perfect, so it is not even clear that a collusive industry will be operating on the elastic part of the demand curve. 6 Another reassuring fact is that the results reported in this paper are broadly similar for the whole sample and for the sample of exogenous sunk cost industries. This is important because industries with low advertising or R&D intensity were less likely to be collusive in 1958 (see Symeonidis 1998a). So if some unspecified factor has affected market structure and profits of exogenous sunk cost industries more strongly than those of endogenous sunk cost industries during 21

23 The model was estimated for the whole sample as well as for the sub-sample of exogenous sunk cost industries and the results are reported in tables All the results presented are for a fixed effects specification. 8 The t-statistics are based on heteroskedasticityconsistent standard errors, adjusted for finite-sample bias following MacKinnon and White (1985). Two different R 2 's are reported: the first is derived from transforming the data to obtain deviations from industry means and applying OLS to the transformed data, while the second is from applying OLS to the untransformed data after including a set of industry dummies among the regressors (the LSDV model). Table 2 contains regression results for lnnfirms and lnnplants. Note that in the regressions using the whole sample interaction variables are used to control for possible differences between exogenous sunk cost, advertising-intensive and R&D-intensive industries regarding the effect of market size on the number of firms or plants (cf. Sutton 1991). In particular, ADV*lnSALES (RD*lnSALES) is equal to 0 for industries with typical or average advertising-sales ratio (R&D-sales ratio) over the period lower than 2% and to lnsales for the 1960s, the results of the present paper would be biased. However, there is no reason to expect such a differential impact of some unspecified factor across groups of industries. And the fact that the results are robust to including or excluding advertising-intensive and R&D-intensive industries also suggests that no such bias exists. 7 As can be seen in these tables, the exclusion of industries with ADS or RDS > 1% does not very much affect the coefficients and t-statistics on the competition variables. One reason for this may be that there were relatively few industries with high (i.e. larger than 2%) advertising or R&D intensity and a change of competition regime, and almost none of these has very high (larger than 5%, say) advertising or R&D intensity. 8 The Hausman test typically rejects the random effects model. In any case, the results of the random effects model with respect to the competition effect were similar to those obtained from the fixed effects specification. 22

24 industries with advertising-sales ratio (R&D-sales ratio) higher than 2%. 9 The results in table 2 suggest that the 1956 legislation had a strong and statistically significant negative effect on the number of firms in the long run. This effect was only partly realised by 1963, and it was mostly realised by The magnitude of the coefficient on CHANGE*Y73 implies that the intensification of price competition following the 1956 legislation has reduced, on average, the number of firms by about 15% between 1958 and This may understate the impact of price competition on the number of firms to the extent that there is measurement error in the construction of CHANGE as a result of ineffective or unregistered agreements. The effect of the intensification of price competition on the number of plants is somewhat weaker: the coefficients on CHANGE*Y68 and CHANGE*Y73 in regressions with lnnplants are negative but smaller in absolute value than in regressions with lnnfirms, and not always significant at the 5% level. This implies that, while much of the structural adjustment in British manufacturing following the 1956 legislation took the form of exit, mergers were also part of the process (so the reduction in firm numbers was more pronounced than the reduction in plant numbers). Finally, note that market size has a strong positive effect on the number of firms or plants, although this effect is less pronounced in advertisingintensive industries Regressions were also run with alternative interaction variables, namely using 1% instead of 2% as the cutoff point, but these were not statistically significant. In preliminary regressions, interaction variables for advertising intensity and R&D intensity were also included for the competition effect, but they were not statistically significant, either individually or jointly. Note that whether an industry's typical advertising-sales or R&D-sales ratio is higher or lower than 2% is largely determined by exogenous characteristics, namely "advertising effectiveness" and "technological opportunity" respectively. Hence the interaction variables are exogenous. 10 It is to be expected that differences across classes of industries with respect to the market size effect on firm numbers will not be significant, since many small firms in advertising-intensive and 23

25 Table 3 presents results for lnplantprofit and lnfirmprofit. The first of these variables may be the one more closely associated with the theoretical model of section 2. This is because the capital-labour ratio is meant to control for setup costs at the plant level (note that the coefficient on lnk/l, which is everywhere positive and statistically significant, is larger in the regressions for lnplantprofit). To the extent that the plant-to-firm ratio increases, gross profit per firm could rise relative to gross profit per plant if there are significant economies of multi-plant operation. The results in table 3 suggest that the 1956 Act had no statistically significant impact on the gross profit of the average plant in the long run, as predicted by the theory of section 2. Moreover, there is no evidence of any statistically significant effect of the Act on firm gross profit either. A remarkable feature of the results in tables 2 and 3 is the magnitude of the coefficients on the time dummies. After controlling for market size, the capital-labour ratio, union density, the intensity of price competition and industry effects, the number of plants or firms in a given industry in 1973 was on average about 30-40% lower than in 1958 and the gross profit of the average plant or firm was more than 60% higher in 1973 than in This evolution seems in several cases to have continued a trend present during as well. To some extent these high coefficients must be due to the crudeness of the setup cost proxy used here: as K/L has been increasing across industries throughout the period, it is correlated with the time dummies; so to the extent that K/L is only an imperfect proxy for setup costs and setup costs have also been increasing, their effect could be partly picked up by the time dummies. Note, in this respect, that K/L is not statistically significant or only marginally significant at the 5% R&D-intensive industries spend little or nothing on advertising or R&D (so the endogenous sunk cost model analysed in Sutton 1991 is not relevant for these firms). Also, small firms often do not produce core industry products; for instance, in most engineering industries (many of which are R&D-intensive), there is a large number of small non-r&d-intensive firms producing parts and accessories. 24

26 level in some regressions. Moreover, the time dummies may be capturing the effect of scale economies not directly associated with the cost of plant and machinery, for instance scale economies in marketing or the raising of finance. While the results in table 3 are consistent with Selten's predictions, the discussion in the previous paragraph and the fact that the overall changes in lnplantprofit and lnfirmprofit during the period examined were probably mainly driven by the large decline in firm and plant numbers suggest that it is necessary to also look at other profit measures to assess the impact of the UK cartel laws on firms' profits. Table 4 reports regression results using lnprofit or PCM as dependent variables. The coefficients on CHANGE*Y68 and CHANGE*Y73 are nowhere statistically significant, even at the 10% level. The failure to detect any long-run effect of the intensity of price competition on industry gross profit may seem puzzling. However, it is not inconsistent with the Selten-Sutton framework. For instance, the existence of asymmetries between firms may explain the result, as pointed out in section 2. Or it may be the case that any effect is very small and therefore difficult to identify empirically. On the other hand, the results are consistent with the Selten-Sutton model in the sense that the primary effect of an intensification of price competition in the long run is on market structure, not on profits. It is also of interest to compare the short-run and long-run impact of the legislation. As can be seen in table 2, it was between 1963 and 1968 that most of the restructuring of previously cartelised industries occurred. Also, the price-cost margin declined, on average, between 1958 and 1963 in these industries, before recovering during (see table 4). The evolution of margins is thus consistent with the evolution of market structure in previously collusive industries: a moderate effect on the number of firms by 1963, at which date several industries were in short-run disequilibrium with reduced margins; then a significant negative effect on firm numbers between 1963 and 1968, leading to a rise in margins of those industries. 25

27 It should also be noted that the coefficient on CHANGE*Y63 in regressions using PCM, although statistically significant at the 5% or 10% level, is nevertheless not large: about one percentage point. It has to be borne in mind that only a subset of the previously cartelised industries were in short-run disequilibrium in 1963; in several industries competition had not yet emerged, and there must have also been several industries where competition had emerged but much of the adjustment of market structure had already taken place. Thus the magnitude of the coefficient on CHANGE*Y63 should not be taken as a measure of the fall in the pricecost margin following a change of competitive regime and prior to any adjustment of market structure; this should be quite higher than one percentage point. The above arguments may also help explain why no short-run negative effect of the Act could be identified on lnplantprofit or lnfirmprofit (in addition to the fact that these variables are heavily influenced by changes in firm or plant numbers and are thus less appropriate than PCM for identifying disequilibrium effects). Finally, it may be argued that the profit equations estimated above may inadequately control for industry-specific factors that cause departures from long-run equilibrium. Hence regressions were also run including the variable ÄlnSALES among the regressors, defined for industry i in year t as the change in lnsales in the five-year period preceding year t. The coefficient of this variable was everywhere positive and sometimes statistically significant, but the rest of the results were not affected. A disadvantage of this alternative specification is that the first-year observation for each industry cannot be used (since ÄlnSALES is not available), and this implies dropping all 1954 observations; this is why results have been reported here using the more restricted specification. 6. Concluding remarks. This paper has analysed the impact on profits and market structure of the introduction of cartel laws in the UK in the late 1950s using a panel data set of manufacturing industries 26

28 over the period The only previous statistical analysis of the effect of the 1956 Act on profitability is O'Brien et al. (1979). They used firm-level data for a sample of about 30 industries and found no difference in profitability during the 1960s between industries affected by the 1956 Act and industries not affected. The results of the present paper, which uses industry-level data for a much larger sample of industries, support the hypothesis of a negative effect of the intensity of price competition on the number of firms and of no significant effect on profits. The former result is consistent with the finding of a positive impact of the Act on concentration in Symeonidis (1999c). Moreover, the comparison of short-run and long-run effects seems to suggest a link between changes in market structure and profitability: in the short-run, when the number of firms does not fall much, profit margins are adversely affected; but they recover in the long run, once the number of firms falls. These results are consistent with models that endogenise market structure by means of a free-entry condition and consequently tend to emphasise the effect of firm conduct on structure rather than on performance. It may also be worth emphasising that the present study is not a direct test of the freeentry zero-profit condition. What is tested here is not whether net profit is approximately zero in the long run, a task which is not feasible without firm-level data on profits and capital costs. What this paper tests is whether there is a change in any of several different profit measures in the short run and in the long run following a change in firm conduct. Thus, the results presented here would also be consistent with a model predicting that net profit is consistently larger than zero, even for the marginal firm, by a fraction å, and that å is not much affected by a change in conduct. It is not clear, however, what mechanism could account for the existence, across industries, of supranormal profits which are relatively stable irrespective of whether firms collude to fix prices or not. 27

29 Table 1. Initial (1958) conditions in the two groups of industries. All industries Industries with ADS, RDS < 1% CHANGE=1 CHANGE=0 CHANGE=1 CHANGE=0 No. of industries: Mean (St. dev.) of lnnfirms 3.88 (0.98) 3.98 (1.15) 3.92 (0.96) 4.25 (1.08) Mean (St. dev.) of lnnplants 4.30 (0.95) 4.31 (1.11) 4.33 (0.93) 4.57 (1.00) Mean (St. dev.) of lnprofit 8.64 (1.10) 8.67 (1.17) 8.52 (1.00) 8.39 (1.05) Mean (St.dev.) of lnplantprofit 4.34 (0.88) 4.36 (1.13) 4.19 (0.83) 3.82 (0.87) Mean (St. dev.) of lnfirmprofit 4.76 (0.97) 4.70 (1.23) 4.60 (0.92) 4.15 (1.00) Mean (St. dev.) of PCM 0.17 (0.06) 0.18 (0.08) 0.17 (0.06) 0.16 (0.06) Note: The figures are based on industries with unambiguous state of competition in CHANGE takes the value 1 for industries affected by the 1956 Act and 0 for those not affected. 28

30 Table 2. Regression results for lnnfirms and lnnplants. (Fixed effects estimation.) Dependent variable: lnnfirms Dependent variable: lnnplants No. of observations: No. of industries: No. of industries with change of regime: All industries ADS, RDS <1% All industries ADS, RDS <1% lnsales (13.97) lnk/l (-1.45) UNION (-1.13) Y (5.12) Y (-7.94) Y (-9.76) Y (-6.78) CHANGE*Y (-0.002) CHANGE*Y (-1.25) CHANGE*Y (-2.19) CHANGE*Y (-2.04) AD*lnSALES (-3.66) RD*lnSALES (0.95) R R 2 LSDV (12.76) (21.22) (-2.00) (-3.08) (-1.23) (-0.81) (3.13) (4.61) (-6.62) (-4.30) (-7.45) (-6.71) (-5.28) (-6.77) (0.55) (-1.71) (-0.95) (-1.14) (-2.08) (-1.97) (-2.38) (-1.46) (-4.47) (-0.67) (19.69) (-2.86) (-0.45) (2.52) (-3.92) (-5.40) (-5.32) (-0.80) (-0.84) (-1.78) (-2.00) Hausman: Prob. value: Note: t-statistics based on heteroskedasticity-consistent standard errors in parentheses. 29

31 Table 3. Regression results for lnplantprofit and lnfirmprofit. (Fixed effects estimation.) Dependent variable: lnplantprofit Dependent variable: lnfirmprofit No. of observations: No. of industries: No. of industries with change of regime: All industries ADS, RDS <1% All industries ADS, RDS <1% lnsales (6.91) lnk/l (4.18) UNION (-2.00) Y (-1.97) Y (9.06) Y (9.36) Y (9.43) CHANGE*Y (0.56) CHANGE*Y (-0.40) CHANGE*Y (1.30) CHANGE*Y (0.62) AD*lnSALES (2.22) RD*lnSALES (-0.67) R R 2 LSDV (7.30) (7.92) (3.26) (2.98) (-2.16) (-1.32) (0.02) (-2.30) (8.71) (11.59) (9.50) (11.47) (8.96) (9.32) (-0.27) (-0.53) (-0.67) (-0.26) (0.47) (1.63) (1.02) (1.37) (1.96) (-1.73) (8.02) (2.60) (-1.05) (-0.50) (10.96) (11.01) (9.27) (-1.22) (-0.57) (0.91) (1.63) Hausman: Prob. value: Note: t-statistics based on heteroskedasticity-consistent standard errors in parentheses. 30

32 Table 4. Regression results for lnprofit and PCM. (Fixed effects estimation.) Dependent variable: lnprofit Dependent variable: PCM No. of observations: No. of industries: No. of industries with change of regime: All industries ADS, RDS <1% All industries ADS, RDS <1% lnsales (22.54) lnk/l (2.63) UNION (-3.56) Y (1.75) Y (8.36) Y (6.07) Y (4.91) CHANGE*Y (-0.70) CHANGE*Y (-1.51) CHANGE*Y (-0.07) CHANGE*Y (-0.08) AD*lnSALES (-0.55) RD*lnSALES (-1.67) R R 2 LSDV (25.06) (1.32) (1.38) (3.38) (-3.59) (-4.19) (2.91) (1.81) (8.81) (7.81) (7.62) (5.65) (6.09) (4.34) (-1.06) (-0.21) (-1.65) (-2.03) (-1.19) (0.22) (-0.22) (0.44) (-0.68) (-1.37) (0.55) (2.02) (-3.51) (2.63) (7.54) (6.42) (5.37) (-0.60) (-1.72) (-0.39) (0.27) Hausman: Prob. value: Note: t-statistics based on heteroskedasticity-consistent standard errors in parentheses. 31

33 REFERENCES Bain G.S. and R. Price, 1980, Profiles of Union Growth (Oxford: Basil Blackwell). Burn, D., ed., 1958, The Structure of British Industry, 2 vols (Cambridge: Cambridge University Press). Conyon, M. and S. Machin, 1991, Market structure and the empirical specification of profit margins, Economics Letters 35, Friedman J.W. and J.-F. Thisse, 1994, Sustainable collusion in oligopoly with free entry, European Economic Review 38, Fudenberg, D. and J. Tirole, 1991, Game Theory (Cambridge, MA: M.I.T. Press). Guenault, P.H. and J.M. Jackson, 1974, The Control of Monopoly in the United Kingdom, 2nd ed. (London: Longman). Harrington, J.E., 1989, Collusion and predation under (almost) free entry, International Journal of Industrial Organization 7, Harrington, J.E., 1991, The joint profit maximum as a free-entry equilibrium outcome, European Economic Review 35, Hart, P.E. and E. Morgan, 1977, Market structure and economic performance in the United Kingdom, Journal of Industrial Economics 25, Hart, P.E., M.A. Utton and G. Walshe, 1973, Mergers and Concentration in British Industry (Cambridge: Cambridge University Press). Lydall, H.F., 1958, Aspects of competition in manufacturing industry, Bulletin of the Oxford University Institute of Statistics 20, MacKinnon, J.G. and H. White, 1985, Some Heteroskedasticity Consistent Covariance Matrix Estimators with Improved Finite Sample Properties, Journal of Econometrics 29, Monopolies Commission, 1973, Parallel Pricing, Cmnd (London, H.M.S.O). Montagna, C., 1995, Monopolistic competition with firm-specific costs, Oxford Economic Papers 47, O'Brien, D.P., W.S. Howe, D.M. Wright and R.J. O'Brien, 1979, Competition Policy, Profitability and Growth (London: MacMillan). O'Mahoney, M. and N. Oulton, 1990, Industry-Level Estimates of the Capital Stock in UK Manufacturing, , NIESR discussion paper no Phlips, L., 1995, Competition Policy: A Game-Theoretic Perspective (Cambridge: Cambridge University Press). Political & Economic Planning, 1957, Industrial Trade Associations (London: George Allen & Unwin). 32

34 Rowley, C.K., 1966, The British Monopolies Commission (London, George Allen). Selten, R., 1984, Are cartel laws bad for business?, in H. Hauptmann, W. Krelle and K.C. Mosler, eds., Operations Research and Economic Theory (Berlin/Heidelberg: Springer-Verlag). Sutton, J., 1991, Sunk Costs and Market Structure (Cambridge, MA: M.I.T. Press). Sutton, J., 1997, Game-theoretic models of market structure, in D.M. Kreps and K.F. Wallis, eds., Advances in Economics and Econometrics: Theory and Applications (Cambridge: Cambridge University Press). Sutton, J., 1998, Technology and Market Structure (Cambridge, MA: M.I.T. Press). Swann, D., D.P. O'Brien, W.P.J. Maunder and W.S. Howe, 1973, Competition in British Industry: Case Studies of the Effects of Restrictive Practices Legislation (Dept. of Economics, Loughborough University of Technology). Swann, D., D.P. O'Brien, W.P.J. Maunder and W.S. Howe, 1974, Competition in British Industry (London: George Allen). Symeonidis, G., 1998a, In which industries is collusion more likely? Evidence from the UK, Dept. of Economics discussion paper no. 487, University of Essex. Symeonidis, G., 1998b, The evolution of UK cartel policy and its impact on market conduct and structure, in S. Martin, ed., Competition Policies in Europe (Amsterdam: North- Holland). Symeonidis, G., 1999a, Price and non-price competition with endogenous market structure, forthcoming in Journal of Economics and Management Strategy. Symeonidis, G., 1999b, Price competition, non-price competition and market structure: theory and evidence from the UK", forthcoming in Economica. Symeonidis, G., 1999c, Price competition and market structure: the effect of cartel policy on concentration in the UK", forthcoming in Journal of Industrial Economics. Symeonidis, G., 2000, The Effects of Competition (Cambridge, MA: M.I.T. Press, forthcoming). 33