2 PRICE TRANSMISSION AND MARKET INTEGRATION

A wide economic literature has studied the relationship between prices, either spatial or vertical. Concerning the former, a wide recent critical review is in Fackler and Goodwin (2001). The premises of full price transmission and market integration correspond to those of the standard competition model: in a frictionless undistorted world, the Law of One Price (LOP) is supposed to regulate spatial price relations, while pricing along production chains will depend exclusively on production costs, with all firms producing on the highest isoquant compatible with their isocost lines.

In fact, however, the literature on price transmission indicates that there are at least six groups of factors affecting it.^[1].

Transport and transaction costs: where the latter can be classified, following Williamson, into the three groups of information, negotiation, and monitoring and enforcement costs. These can act as wedges between prices in different markets, which need to be overcome by the total price differences between two locations or industries to allow for arbitrage and integration to take place between two markets. Their treatment is simple if they can be assumed to be stationary, proportional to traded quantities rather than fixed, and if they can be assumed to be additive rather than multiplicative. If this is not the case, modelling price transmission and integration requires non linear models, or linear models including thresholds (McNew, 1996; Barrett and Li, 2002; Brooks and Melyukhina, 2003).

Market power: along production chains some agents may behave as price makers while some other as price takers, depending on the degree of concentration of each industry. It may be the case that e.g. input price increased in an industry may be passed over to consumers, while input price decreases can be captured in the mark-ups of the industry (Wohlgenant, 1999; Azzam, 1999; Goodwin and Holt, 1999; Dhar and Cotterill, 1999; Mc Corriston et al., 2001).

Increasing returns to scale in production: along the same lines, they may be at the origin of market power, although as has been shown, their effect on vertical price transmission is different from that of market power (Mc Corriston et al., 2001).

Product homogeneity and differentiation: the degree of substitutability in consumption between similar goods produced in different countries may affect market integration and price transmission. This type of evidence can be addressed through the introduction of the so-called Armington assumption, of less than infinite substitutability in consumption between goods produced in different countries.

Exchange rates: the extent to which changes in the exchange rates are “passed through” on output prices has been studied in relation to the ability of firms to discriminate prices across destinations (pricing-to-market behaviour), to market structure, to product non-homogeneity, and the adjustment costs of firms (Dornbush, 1987; Froot and Klempeter, 1989; Knetter, 1993).

Border and domestic policies: those that directly affect spatial price transmission are trade policies, although domestic policies affecting price formation do also affect both vertical and spatial price relations (Mundlak and Larson, 1992; Zanias, 1993; Baffes and Ajwad, 2001; Thompson et al., 2002; Sharma, 2003). Among border measures, non tariff barriers may have strong effects on price transmission: this is the case of variable tariffs; tariff rate quota, prohibitive tariffs, and technical barriers. Ad valorem and fixed tariffs, instead, should behave exactly like proportional and fixed transaction costs respectively.

All these elements can affect both spatial and vertical price relations; nonetheless, the second, the third and the fourth have been mostly investigated with reference to vertical price transmission, while the last one has been mostly studied with reference to spatial price transmission.

A general distinction may be drawn in the literature in terms of the extent to which contributions deal explicitly with the deviations from the competitive model. Many papers are aimed at checking the consistency of the empirical evidence with the competitive framework, without proposing explicitly an alternative behavioural model, rather attempting to infer behavioural evidence from the data. The relatively more recent among such contributions focus their attention on the dynamics of the transmission process, using the properties of co-integrated time series, and the related econometrics. In such type of work, which may be considered to be following a “non structural” approach, factors determining transmission are treated as external prior information - rather than the outcome of a theoretical framework - to be confirmed by the results. In other words, the data are the starting point, while the tests indicate the extent to which prices adjust toward an equilibrium. Evidence showing that this is not the case is interpreted in terms of one or more of the highlighted factors that can affect price transmission depending on the specific context. From this point of view, testing for price transmission can be interpreted as an exercise to check the degree of efficiency of the markets, in terms of their being close to the competitive model, or as a tests for market integration, following the definition offered by Barrett and Li (2002).

Comprehensive analytical framework for this econometrics approach can be found in Balcombe and Morrison (2002), and Rapsomanikis et al. (2003). Co-integration between the price series analyzed implies that two prices may behave in a different way in the short run, but that they will converge toward a common behaviour in the long run. If this property is verified, the characteristics of the dynamic relationship between the prices can be described by an Error Correction Model (ECM). Despite a number of caveats (Barrett and Li, 2002; Rapsomanikis et al., 2003), the short-run adjustment parameter of this type of model can be interpreted as a measure of the speed of price transmission, while the long run multiplier can be interpreted as a measure of the degree of price transmission of one price to the other (Prakash, 1999). The properties of co-integrated series also imply the existence of a causality relation, as defined by Granger, that can be tested by assessing if the past observations of one of the two prices (fail to) predict those of the other. Therefore, most analyses start by investigating the dynamic properties of the price series, through tests for the presence of unit roots, and then proceed with co-integration tests, and with the specification of ECMs.

Among econometric applications, some were directly aimed at verifying the LOP for commodity prices. Examples are Ravallion (1986), Ardeni (1989), Baffes (1991), Mundlak and Larson (1992), Gardner and Brooks (1994), Goletti and Babu (1994),Mohanty et al. (1998b), Yang et al. (2001), Baffes and Ajwad (2001), Barrett (2001). Results appear controversial, and sensitive to the techniques employed. Interesting extensions of the econometric approach allow for transmission to be affected by the presence of asymmetric response, by thresholds, and for a fractional order of integration of the price series.

Threshold models were introduced by Enders and Silkos (1999) and quite widely applied to agricultural price series (Goodwin and Piggott, 1999; Thompson and Bohl 1999; Goodwin and Harper, 2000; Mainardi, 2001; Abdulai, 2002; Meyer, 2002; Sephton, 2003). This type of model is aimed at testing for the presence of non linear transaction costs, and in general for the existence of price bands within which there is no transmission.

Models with asymmetric adjustment have been frequently employed to test for the presence of market power, drawing on the idea that agents holding market power will pass-through only (or mostly) positive input changes. Examples include Morissett, (1998), based on a static framework for analyzing annual data. In dynamic applications, the short-run adjustment term is substituted by two separate coefficients accounting, respectively, for negative and positive deviations from the long run equilibrium. This allows testing for asymmetry in terms of rejection of the restriction that the two coefficients are equal. Applications of this type can be found, among others, in Goodwin and Holt, (1999); Abdulai (2000), Meyers and von Cramon, (2000), Kuiper et al. (2002), Rapsomanikis et al. (2003). A simpler method is adopted by Prakash et al. (2001), based on the significance of a dummy variable accounting for positive residuals in the static regression between the two price series involved. The idea is that if this variable is significantly different from zero, and if the ECM coefficient of the model including this variable is greater than the one without the dummy variable, transmission is asymmetric, since positive shocks are passed through faster than negative ones.

Alternative to the econometric approach, there are a (relatively limited) number of contributions that, instead of testing for transmission and market integration, attempt to derive explicitly an alternative behavioural rule, different from those implied by the simple equalization of price behaviour. Contributions following this “structural” approach are relatively less frequent in the analysis of spatial transmission, and more frequent in the analysis of vertical transmission; in this field there are a number of industrial economics applications, addressing the presence of market power and/or of increasing returns to scale in production. Examples are McCorriston et al. (2000), showing, on the one hand, how market power is expected to reduce the degree of price transmission compared to competitive markets, given that producers will be able to gain extra profits by holding prices higher than in competitive market conditions; while increasing returns to scale, on the other hand, are able to increase the degree of price transmission beyond the level of perfect competition. In the same vein, Dhar and Cotteril (1999) propose a structural model with strategic behaviour to estimate price transmission along a dairy supply chain; and Acharya (2000) shows how asymmetric price transmission behaviour can be explained by the existence of market power along the food chain.

Concerning spatial relations, interesting work has been devoted to specify conditions allowing for market integration and price transmission to take place within competitive models. McNew (1996) provides an example of this approach; an ad hoc spatial trade model for an homogenous product, driven by transport costs, is employed to estimate the probability of trade between markets on the basis of a competitive market structure in which the LOP holds. In the same line, Barrett and Li (2002) develop a spatial model, also driven by transaction cost, in which, among other things, they highlight two issues: firstly, the possibility that price transmission occurs in absence of trade (segmented equilibrium), and that trade takes place in absence of price transmission (imperfect market integration); and, secondly, that most econometric applications are in fact aimed at testing the most restrictive condition, in which both market integration and a competitive equilibrium are verified. These applications, therefore, would not be capable of fully capturing the “messy” character of market relationships (Barrett and Li, 2002), arising from treating price transmission mostly as a linear phenomenon, and from neglecting to consider the erratic nature that transaction costs can assume.

As usual, each approach has its own merits and drawbacks. The econometric applications, and especially the most recent ones, have analyzed mostly the dynamics of price transmission, while elaborating less from a theoretical point of view. At the same time, models attempting to develop behaviour rules governing pricing and market relations - either in a competitive environment with transaction costs or under imperfect competition and increasing returns to scale - appear more specific and more demanding in terms of data.

Given the purpose of this paper - which is that of providing evidence of transmission for a wide set of prices, both within the same areas along the production chains, and between local and world reference prices - it is chosen a simple econometric framework. As seen, despite its limitations, this approach can provide at least an useful starting point for more in-depth investigations to be conducted on specific cases, together with a set of background information to be used in the development of structural models. This latter topic is discussed in next section.