Given that most data employed in the econometric analysis is expressed in logarithms to reduce data variability, the estimated parameters can directly be interpreted as “transmission elasticities” of one price with respect to another. The dangers involved in this interpretation are clearly highlighted in the literature, particularly by the critics of the econometric approach: in fact such parameters can be
affected by factors that do not prevent market integration or the transmission of price signals (Mc New, 1996; Barrett and Li, 2002; Brooks and Melyukhina, 2003), so that a low parameter may arise between two markets which are in fact integrated;
smaller than one even if price transmission and market integration are complete: a transmission elasticity will be equal to one only if there are no fixed elements involved in the transaction, but only ad valorem ones (Sharma, 2003).
Given these caveats, still the value of the parameters and their significance level provides information about the extent to which markets share the same price shocks or, conversely, the extent to which they are “messy” according to Barrett and Li’s (2002) expression. In other words, a transmission parameter summarizes the overall effect of a set of factors affecting price signals, including transaction costs that may be stationary, the existence of market power among the agents involved in transactions, the existence of non-constant returns to scale, the degree of product homogeneity, the changes of the exchange rates, and the effects of border and domestic policies. Since most estimations include a constant term, they should include only the effects of those elements that change proportionally with prices, without accounting for the interaction between the effects of each of those elements.
Can transmission parameters be safely plugged in projection and simulation models to reproduce the functioning of the involved markets? At least, in the price transmission equations the coefficient of the constant term may be included together with the transmission elasticity, to account for the fixed effects separately from the proportional ones. But a more complete answer depends on the aim of the projection or simulation model involved; and the more these models are capable of reproducing explicitly factors affecting price transmission, the higher may be their usefulness for projections and simulations.
For instance, in multi-markets models aimed at policy simulation, the inclusion of a set of variables representing a policy tool should be preferred to the inclusion of a simple transmission elasticity, which summarizes the effect of many factors, since in the first case it is possible to assess the effect of a change in the policy itself on the transmission of prices, while in the second it will not be possible to separate the effect of the policy change from one taking place in the other factors affecting transmission.
As an example, suppose that a country changes the level at which it operates the floor price for wheat. If this policy is represented with an equation that triggers stock accumulation when the price falls below the floor level, while transport costs are dealt by within a different constant term that determines a wedge between the domestic price and the world price, it will be possible to assess separately the effect of a 10 percent reduction in the floor price from that of a 10 percent reduction in transport costs; whereas, if all is summarized in a spatial price transmission elasticity, it will only be possible to represent both changes as some percentage change in the elasticity.
Despite this limitation, however, transmission elasticities can serve as background information for policy modellers, to understand how specific factors, and their interaction, are affecting transmission. In the above example, an estimate of the price transmission elasticity between a world reference price and a domestic price for wheat can help understanding if the adopted modelling strategy is in fact reproducing the functioning of that particular market, or if, e.g. transmission is in fact very incomplete and asymmetric due to, for example, a combination of high transaction costs, and a highly concentrated market. The calibration of policy analysis models could benefit from information on transmission parameters: it will be possible to check the extent to which the results generated by the structural model are consistent with the overall price transmission observed in the econometric exercise.
In fact, large size equilibrium models currently employed in agricultural market projection and policy simulation tend to use policy variables rather than transmission elasticities, and/or to supplement policy variables with the transmission elasticities where less information is available, or where domestic prices are not defined. A review of the spatial price transmission mechanism in such models can be found in Cluff (2003). That paper shows how, for example, in the FAO WFM model price transmission for countries with no WTO commitments is modelled without qualifying the cause affecting it, simply with a price transmission elasticity derived in some cases from estimation, but often from expert judgments or calibration. In the same model, policies are assumed to be the major determinant of price transmission for countries that have undertaken WTO commitments, together with a residual term accounting for transaction costs. In the FAPRI modelling system, policies are kept separated from the other elements that affect price transmission, which are included in a constant term. Even this term, however, includes policies affecting prices as a fixed - rather than proportional - element. A more interesting approach is proposed by the USDA CCLS model, which allows for a variety of factors to explicitly affect price transmission, such as transport costs, the exchange rate, and trade policies, depending on the specific market. AGLINK, the OECD agricultural model, also takes a similar approach. As a matter of fact, even in well researched and known markets such as, for example, wheat in the EU and the United States the degree of price transmission hypothesised by different models is widely variable (Cluff, 2003), reflecting the relative importance attached by each researcher to different elements, if not measurement errors. Each of these models seems to be setting price transmission according to the main variables of interest: where policy analysis are the major objective, price transmission is mainly affected by policy measure, while other elements assume a less important role: either generic transmission measures are utilized to handle cases in which no information is available (e.g. an average for similar countries), or elements different from policies are included as fixed terms, or as calibration residuals.
Apart from the size of the parameters and their form, price transmission in large-size equilibrium models depends also on the nature of the trade component. Where trade is modelled as a residual of domestic supply and demand, and only net positions are generated endogenously - as is the case for models quoted above - price transmission for one market will defined by a single set of parameters for each product. If the trade component, instead, generates a set of bilateral trade flows, then a specific set of transmission parameters are to be defined for each of such bilateral flows.
This is the case of spatial equilibrium policy models (Anania, 2001), in which transmission is affected separately by transaction costs and policies, allowing the effects of changes in both groups of variables to be simulated separately. Another example is that of models in which bilateral flows are defined through the elasticity of substitution between imported and domestic products - following the so-called Armington assumption - in which case transmission is affected by the size assigned to this parameter. The implementation of this approach can be problematic, as it may be difficult to assign credible values to the elasticities of substitution, especially for a wide set of markets; in the case of the GTAP model, for instance, these are assumed to be homogenous across products, to keep the modelling simple (Hertel, 1997).
Market structure as a factor affecting price transmission appears very rarely in transmission equations of large size equilibrium models. Examples can be found among the GTAP application (Francois et al, 2003) where the assumption of increasing returns to scale allows for spatial transmission to be governed by a non competitive pricing rule. Another example can be found in the partial model proposed by Moro et al. (2002), in which price transmission include policies as price wedges, while wedges between producer and consumer prices, representing vertical transmission, are driven by Herfindhal indexes, related to the degree of concentration of that particular market. Within the limitations implied by such simple modelling, the representation allows for a separate simulation of the effects of a change in the market structure of an industry and those in the policy setting.
To sum up, rather than directly providing parameters to be inserted in policy analysis models, evidence on price transmission may be employed to check the consistency of the results of such models. Ideally, by including an explicit modelling of those factors that affect transmission - such as policies, the exchange rate, transaction costs, quality differentials, the degree of concentration etc. - equilibrium models should be able to reproduce a degree of transmission consistent with the one found in econometric exercises; the inclusion of a transmission elasticity appears mostly as a measure of the lack of interest of the modeller in the factors affecting transmission.
