Price transmission in the milk sectors of Poland and Hungary

Abstract

Rapid and thorough changes have recently taken place in the dairy supply chain in the whole of Central and Eastern Europe. Growing concerns have been expressed that these changes may negatively affect farmers' relative position vis-a-vis the downstream industry, owing to market power exercised by the latter. This article investigates the price transmission mechanism in two countries from the region, Poland and Hungary, and contrasts the results with dairy market organisations specific to these countries. Using cointegrated vector autoregression and controlling for potential structural breaks, it is shown that Polish milk prices, unlike Hungarian ones, are characterised by short-term and long-term asymmetries. We discuss a number of potential explanations for these results, and consider, among others, differences between the dairy chain structures and the role of FDI.

Notes

 1. The situation looks somewhat different in the case of Western Europe, where price transmission asymmetries were found for a number of countries and sectors. As regards the dairy sector, which is the interest of this study, Gohin and Guyomard (2000), using a structural model, strongly rejected the hypothesis that French food retail firms behaved competitively and concluded that more than 20% of the wholesale–retail price margins for dairy products could be attributed to oligopoly–oligopsony distortions. Imperfect price transmission in the dairy sector was also found for the UK, Germany and Denmark (London Economics 2003). Similar conclusions could be drawn for the US (Kinnucan and Forker 1987, Lass 2005, Capps and Sherwell 2007) and Brazil (Aguiar and Santana 2002). On the other hand, Serra and Goodwin (2003) found no evidence of price transmission asymmetries in the Spanish dairy and perishable products sectors.

 2. Studies dealing with price transmission and the integration of agro-food markets in Russia should also be noted here (e. g. Goodwin et al. 1999, Loy and Wehrheim 1999).

 3. Provided that they are able to cope with the data frequency problem mentioned earlier.

 4. Although a similar approach has been applied to Western European countries, we are aware of only two such studies, London Economics (2003, 2004).

 5. In countries like Germany and France this share exceeds 80%.

 6. Consider the first order autoregressive process AR(1):

y _t  = ρy _t − 1+e _t t = …, − 1,0,1,2,…, where e _t  is white noise.

The process is considered stationary if |ρ| < 1, thus testing for stationarity is equivalent to testing for unit roots (ρ = 1). Rewriting to obtain

Δy _t  = δy _t − 1+e _t  where δ = 1 − ρ, the test becomes

H ₀:δ = 0 against the alternative H ₁:δ < 0.

 7. For robustness we also employed the Kwiatkowski et al. (1992) unit root test, which, unlike the ADF, has stationarity under the null hypothesis. The results, however, remained unaffected.

 8. Two or more non-stationary variables are cointegrated if there exist one or more linear combinations of the variables that are stationary.

 9. For each possible break point, the ADF statistics corresponding to the residuals of models are computed, then the smallest value is chosen as the test statistic (being the most favourable for the rejection of the null). Critical values are non-standard and are tabulated in Gregory and Hansen (1996).

10. The minimum ADF statistic corresponding to a break point in November 2000 is −  5.951, significant at 1%, thus rejecting the no-cointegration null hypothesis and reinforcing Johansen test results with included structural break dummy.

11. As illustrated in Figure 1, the strong widening of the spread between farm and retail prices happened at the end of 2003 and was caused by expectations of abolition of the national price support system, which took place at the beginning of 2004 (Hockman and Vőneki 2009).

12. Since the presence of cointegration relation implies Granger causality in at least one direction (Lutkepohl and Kratzig 2004) these results are consistent with expectations that could be based on the cointegration analysis.

13. Interpretation of VECM coefficient estimates is not particularly important for the goals of this article; therefore, owing to space limitations, they are not presented. Estimates are however available from the authors upon request.

14. It might also be noted that this difference in farm structure helps to explain the differences in direction of causality between farm and retail prices that could be observed in both countries (Tables 4 and 5).

15. Unfortunately our data limitations do not allow us to test this hypothesis.