Structure–conduct–performance approach

33 firms have the ability to influence price and sell their output at prices above marginal costs. This idea was formalised by Lerner 1934, p. 161 with an index known as the Lerner index p c L p − ≡ , where p is the output price and c is the marginal production cost. A higher Lerner index is interpreted as a higher degree of market power: This interpretation needs to be used with caution, because the price–cost margins that determine the index can increase either with an increase in price or a decrease in marginal cost. The interpretation will be appropriate if the increase in the Lerner index is triggered by an increase in output prices. If the increase stems from a decrease in marginal costs, a higher Lerner index may reflect higher efficiency rather than market power. This ambiguity may arise if the observations are derived from single-period equilibria. The one-shot game framework of such equilibria precludes both the possibility of new entrants to the markets and firms’ consideration of their rivals’ responses. In a multi- period case, positive price–cost margins will attract new entrants to the markets, or give incentives for rivals to increase their output quantity. If there are no barriers to entry—as in competitive markets—this process may continue until prices equal marginal costs again in equilibrium. Therefore, the existence of a positive price–cost margin can only be considered as evidence of market power if its occurrence is persistent over time. Although the Lerner index has been broadly accepted as a good measure of market power, most studies do not directly use it because marginal costs are usually difficult to determine. As an alternative, many models have been developed in order to measure market power. These models can be divided into the structure–conduct–performance SCP and the new empirical industrial organization NEIO approaches Tirole 1988.

3.2 Structure–conduct–performance approach

The structure–conduct–performance SCP approach, introduced by Mason 1939; 1949, suggests that evidence of market power can be concluded from a positive relationship between structure and performance. In this approach, market structure variables— measured by market concentration, product differentiation, vertical integration, or Universitas Sumatera Utara 34 barriers to entry—are treated as exogenous variables. Market performance variables—for which rate of return, price–cost margins or Tobin’s q index can be used as proxies—are treated as the dependent variable. The structure variable is regressed on the performance variable to estimate the market power in the industries. As well as evidence of market power, this estimation also provides an explanation of the impact of changes in market structures on changes in market performance. Therefore, this model is often used to evaluate the impact on market performance of policies that change market structure. The SCP approach has been criticised in at least three aspects. First, market structure variables are, in fact, not always exogenous. For example, market concentration—as one of the market structure variables—may be affected by firms’ efficiency. Firms with higher efficiency will have lower costs, enabling them to sell output at lower prices. Less efficient firms can not do so without facing a loss, forcing them to exit from the market, which leads to an increase in market concentration. In this case, the relationship between market performance and market structure is the inverse of the idea in the SCP approach, that is, the former variable is determined by the latter one Caves and Porter 1977, p. 241; Bresnahan 1989; Delorme et al. 2002, p. 13. As a result, suggesting that market power exists from a positive regression coefficient in such conditions may be misleading. Second, both market structure and market performance variables are often difficult to measure. In market structure variables, market concentration is widely measured with the Herfindahl–Hirschman index HHI. However, this index has been widely criticised for its sensitivity to the relevant market definition in both geographical boundaries and product homogeneity Lijesen 2003, p. 123. Product differentiation is also often difficult to measure. Data on product differentiation strategies such as advertising, marketing or technical change are rarely available. In addition, there is product differentiation that is unique to market leaders in the buyers’ perception, which is more difficult to measure. Even when in some cases these variables can be measured, their degree of differentiation may vary, as some firms may produce a product that is more differentiated than others Rhoades 1985, pp. 344-347. Measuring vertical integration or barriers to entry is often problematic due to the lack of transaction costs data and the ambiguous interpretation. Economies of scale are not seen as a barrier to entry for the Chicago school because they Universitas Sumatera Utara 35 affect equally the costs of both new and old firms when they have access to the same technology. In contrast, they are treated as a barrier to entry in the classic limit pricing model because it is assumed that incumbents will maintain their outputs and thus constrain the market available to new entrants Gilbert 1989, p. 113; Meyer 2004, p. 328. In the market performance variables, the rate of returns is often problematic, as capital data are usually reported using accounting rather than economic concepts. Price–cost margins are also difficult to calculate because marginal cost data are rarely available. As an alternative, average variable costs are often used as a proxy for marginal costs. However, unless the data are derived from a long run equilibrium, average costs may differ from marginal costs. Similarly, measuring replacement costs and expenditures on intangible assets such as advertising and research and development in the Tobin’s q index, is often difficult Boyer 1996, p. 116; Carlton and Perloff 2005, chapter 8; Perloff et al. 2005, chapter 2. Therefore, the market performance measures may be inaccurate. Third, the SCP approach assumes that various industries—data from which are used as samples in the regression—have the same structure–performance relationships. In fact, each industry may have important idiosyncrasies. While in a regression analysis, the relationship between variables is supposed to be a causal effect between a dependent or explained variable and independent or explanatory variable, this needs not necessarily be true if data are collected from various industries with different structure and performance relationships. The relationship between variables could be interpreted as only correlation or descriptive analysis. As a consequence, a positive coefficient from such a relationship may not necessarily provide evidence of market power Carlton and Perloff 2005, chapter 8; Perloff et al. 2005, chapter 2. In the recent SCP models, the endogeneity problem is addressed by using a simultaneous- equations technique Delorme et al. 2002. Another way is by carefully choosing the market—whose structure is set by the government rather than by other variables in the system—to avoid the endogeneity problem in the market structure variable Brown and Brown 1998. However, even in these studies, the important idiosyncrasies and Universitas Sumatera Utara 36 performance measures issues have not been considered. Therefore, the results may still be implausible.

3.3 New empirical industrial organization approach