The importance of market integration

Price transmission refers to the way and the speed in which domestic prices adjust to changes in international prices. ^{Vollrath and Hallahan (2006)} state that the speed and the magnitude of a good’s price response in the region and the price change of the same good in another region depend on market efficiency; efficient markets generate quick and symmetrical responses, while inefficient ones produce an incomplete and asymmetrical price transmission.

Market integration is important because it affects economic growth, induces structural changes, alters the spatial location of the economic activities, and implies an opportunity for customers to acquire goods at competitive prices. Therefore, market integration is an issue of interest for governments, due to its implications for economic welfare (^{Barrett and Li, 2002}).

The Law of One Price (LOP) is the theory used in studies about the market integration of spatially-separated homogeneous products, which guarantees the absence of arbitrage opportunities (^{McNew, 1996}). If this equilibrium condition is met, it is possible to say that the markets are integrated and that price transmission exists. In this context, studying the Mexican and international coffee market integration is relevant, because coffee growers could benefit from the spatial arbitrage and their incomes could improve with expanded opportunities and incentives to strengthen coffee production and trade.

Market integration is defined as the existence of a positive trade flow between two spatially-separated markets (^{Barrett, 2001}). In economy, compliance with the Law of One Price (LOP) is used as an indicator that two geographically separate markets are integrated. The degree of market integration is empirically defined by means of the estimated degree of price transmission, the speed in which the prices of one market are transmitted to another (^{Lence and Falk, 2005}).

This investigation’s hypothesis was that the structural reforms of 1989 regarding coffee export quotas and the trade liberalization process -which began with Mexico’s inclusion, in 1986, in the General Agreement of Tariff and Trade (GATT), and which was deepened with the signature, in 1993, of the North American Free Trade Agreement (NAFTA) between Mexico, Canada and the USA-led to a greater commercial integration between Mexico and the international coffee market. This integration is proved by constant trading, a greater transmission of market signals, and a higher speed in the domestic price adjustment to international price movements. The objective was to estimate the degree of trade integration between both markets and the speed of the coffee price adjustment in Mexico, in view of international price changes.

Shortand long-term dynamic relation

The cointegration between variables -once the unit root existence has been proved- is the obligatory condition for the existence of a long-term equilibrium relationship in the series. A variable vector with unit root is cointegrated if a lineal combination of these variables is stationary (^{Engle and Granger, 1987}). VEC’s importance is derived from the Engle and Granger’s demonstration that, if two variables can be integrated in order one, I (1), and they are cointegrated, they can be shaped under the assumption that they are being generated by an error-correction mechanism. VEC has a higher predictive capacity than other auto-regressive models (^{Engle and Yoo, 1987}).

The structural change happens when the parameters that determine the series are altered. If the series modelled by Y _t = α+ ρY _t−1 +e _t , where ρ<1, is modified in, at least, one of its parameters, (α,ρ,σ ²), a structural change is generated which can affect any or all the model’s parameters. Changes in α alter the series’ average and tendency; modifications in ρ reflect changes in the serial correlation of Y _t , and variance alterations create changes in the series’ volatility.

^{Zivot and Andrews (1992)} proposed an endogenous test to identify structural changes in the series’ intercept or tendency; it is based in a data sequential analysis, in which the whole sample is used and a dummy variable is applied in each observation of the series under evaluation as a potential break. This break is identified, when the Augmented Dickey-Fuller Test’s t-statistic is more negative. Therefore, if the ADF test shows evidence that the series is stationary; there will be a greater possibility of finding breaks. Zivot and Andrews’ t-statistic values (1992) have their own asymptotic theory and critical values^[1]; these are more negative than those used in proceedings that specify structural change date, and consequently they have a greater difficulty rejecting the unit root.

During the second stage, the short- and long-term relationships between domestic and international prices of coffee were estimated; according to ^{Baffes (1991)}, this relationship is calculated using the VEC model of equation (1):

However, while estimating the equation (1), an identification problem arises; therefore, its parameters were modified in order to find out the longand short-term effect, and the estimate was carried out based on the specifications of equation (2). On this matter, ^{Baffes and Ajwad (1998)} showed the following differential: Pt1-Pt2=εt; therefore, this implies that β ₀ =0, and that, therefore, there is a long-term relationship.

In (2), ft=st+pt* and zt+pt+st in which: p _t , pt*and s _t indicate, respectively, the price of coffee in Mexico, the international price logarithm, and the Mexico-USA nominal exchange rate.

Within the VEC methodology framework, the stationarity of the differential (p _t−1 −β ₁ Z _t−1 ) implies than an error-correction mechanism (ECM) exists and therefore α ₂ must be significantly different from zero (^{Baffes and Ajwad, 1998}). In this model, α ₂ is interpreted as transmission to internal price, derived from a change in international price adjusted according to the exchange rate within the first period, an effect known as short-term effect. The most important feature of the VEC model is related to the interpretation of the α ₃ parameter, which accounts for the way in which the difference between the two prices (internal and international adjusted according to the exchange rate) is eliminated in each subsequent period, an effect that is known as error-correction or adjustment speed.

The short-term coefficient can take any value, but the error-correction value must have a 0 to 2 absolute value. As the latter is closer to the unit, adjustment speed will be greater. A symmetrical value with regard to the unit (e.g., 0.8 and 1.2) indicates that the adjustment speed is the same, but with different trajectories (monotonous, in the first case; oscillatory, in the second case) (^{Baffles and Ajwad, 1998}). Long-term convergence requires that α ₃ must sufficiently differ from zero, without any condition above the α ₂ parameter.

The VEC model is useful because its parameters -or one of their functions-are directly interpreted in terms of the links between prices. The model helps to determine if the law of one price works in a given market and, additionally, to find out at which speed does the internal price adjusts to external changes. Therefore, if n is the period in which an adjustment k percentage takes place (^{Baffles and Ajwad, 1998}), the adjustment accumulated in the n period is given by: k=1−(1−α ₂)(1−α ₃) ⁿ . In the previous expression, n can be solved as follows:

Equation (3) is interpreted as the number of periods required to achieve a certain percentage of k adjustment.

Price transmission and adjustment speed

According to the results’ analysis, during the 1981:01-2014:12 period, the transmission elasticity from international prices to domestic prices amounted to 0.13, and that, in face of a movement in international prices, domestic prices needed 24 periods to return to their equilibrium level or long-term trend. Elasticity (α ₂) indicates that price transmission was increasingly modified throughout the period under study; from 0.39 (1981-1986) up to 0.67 (2000-2012) (Table 2).

The above-mentioned results are consistent with the findings of ^{Krivonos (2004)} who studied price transmission between international and domestic coffee markets, including Mexico, using price data provided by the International Coffee Organization (ICO), and an error-correction model. Krivonos found that price transmission increased after 1993 (when the market was liberalized), whereas economic interventionism policies were in place before that year; meanwhile, price transmission elasticity decreased in the periods before (0.28) and after (0.44) the liberalization of the coffee market. In our study, we calculated that, before and after the reforms, values were 0.35 and 0.61. A possible explanation for these differences is that -although both studies use the same international pricethe price received by growers is different.

The main coincidence between both studies is that movements in international prices have a significant influence in the domestic prices of coffee. This result is similar to the one obtained by ^{Mundlak and Larson (1992)} who carried out an empiric analysis in 58 countries, using FAO’s data and VEC, and found out that global price movements are the main factor that explains internal price variations: international price movements account for 68 % of domestic variations.

The speed at which internal prices adjust to changes in external prices is given by the α ₃ coefficient. At the start of the period, adjustment speed was low during the two sub-periods prior to 1994 -characterized by economic interventionism- and it constantly increased during the three periods after 1994; this indicates that domestic price was increasingly dependent on international price. This result does not agree with that of Krisvonos’ (2004): while he concluded that adjustment speed between the periods analyzed changed from -0.18 to -0.14, our results show that they changed from 0.18 to 0.32. Additionally, our results are consistent with economic theory, according to which high price transmission is related to high adjustment speed, and that the latter is related to more efficient markets (^{Barret, 2001}).

The time required to fully return to the longterm equilibrium situation underwent a gradual modification since 1994. In the period of economic interventionism in coffee prices, domestic prices required 11-32 periods to complete their adjustment, while in the liberalization period, it was reduced up to five periods (Table 2).

Using cointegration analysis, several studies found out that, once the quota period came to its end, the domestic price of coffee became more sensitive to changes in international prices, during the trade liberalization period (after the export quota system was abolished). Additionally, an international price shock had more persistent impact during the quota period, with regard to the post-abolition period (^{Lee and Gómez, 2013}; ^{Mofya-Mukuka and Abdulai, 2013}; ^{Worako et al., 2008}; ^{Krivonos, 2004}).

The proportion of the price paid to coffee growers to the global price of coffee increased during the reform period (^{Worako et al., 2008}; ^{Krivonos, 2004}), which implies that commercialization margins for growers increased in the countries analyzed in these studies. In Mexico, the proportion of the average rural price of coffee to the international price was 1.04, during the price stabilization period (price control), and 0.83, during the price liberalization period.

The trade liberalization process did not contribute to the improvement of signals in the Mexican coffee market and, therefore, Mexican growers obtain a lesser margin with regards to the international price of coffee, in both absolute and relative terms. This may be the result of the lack of an appropriate market intelligence system, the lack of access to credit, the high commercialization costs (mainly transportation), and the high costs of transactions related to the search for buyers and the enforcement of contracts (^{AMECAFE, 2014}).

The elasticity of the non-elastic demand price of coffee means that a drop in international prices leads to a decrease in exportation profits (^{Hallarn, 2004}). In this sense, the impact of the variability of international prices on domestic prices depends on the kind and speed of price transmission. This information enables the design of more effective price stabilization measures or of measures that mitigate the negative effects of variability on international prices.

Coefficients were significant in all models (p≤0.05). The model’s assumptions were verified, in order to trust in the results of the regressions that were carried out. The Portmanteau test (^{Box and Pierce, 1970}) did not reject the null hypothesis of normality (Chi2 probability=0.285). With regard to the presence of heteroscedasticity, the Breusch-Pagan test showed that the constant variance hypothesis (Prob≥0.083) with a 5 % significance was not rejected. The ADF and PP tests showed that these are I (1) series. This fact indirectly proves the series’ linearity assumption, since their first difference is stationary (average and constant variance).

This research empirically showed, for the first time, that the liberalization of coffee trade -characterized by the abolition of the export quota systemcreated a greater integration between the Mexican and the international markets. Likewise, it showed a longterm relationship between price series, diminished trade flows, and an increase in the adjustment speed of internal prices, in response to changes in international prices, which implies greater trade integration.

The results of our study implicitly match the conclusions of ^{Zorya et al. (2014)}, ^{Dawe (2009)}, and ^{Timmer (2004)}: if the long-term trend of international prices is slowly and imperfectly transmitted to domestic markets, consumers and growers take decisions based on prices that do not represent their actual social costs and benefits and any major and sustained deviation between domestic and international prices leads to social welfare inefficiencies and losses.

Coffee growers could benefit from special arbitrage and their income could be increased as a result of better trade opportunities and production incentives. As adjustment speed did in fact increase during the liberalization period, growers and traders can take more efficient decisions. Price increase movements -such as those of 2007-2009- affect Mexican consumers in the supply chain (processors and ultimate consumers), although primary growers may be benefited. However, several factors preserve the imperfect transmission in certain rural areas in Mexico, including inadequate market intelligence, and a trade infrastructure that causes high transaction costs. This could have been the cause that the grain price increases in 2011 were not fully reflected in growers’ income.