^{1}

^{1}

The purpose of this paper is to test how producers’ and retailers’ prices are horizontally integrated, and to show the direction of causality that exists between producers’ price and retails’ price in Ethiopian milk market. The study was conducted making use of secondary data extracted from Ethiopian central statistics agency. The data was time series having 120 observations of monthly recorded price series of producers and retailers, for the period from January, 2004 to December 2013. For this purpose, descriptive statistics and time series econometrics approach (Johansen’s test for co-integration and Vector Error Correction Model) were employed. The study shows that there is strong long run co-integration between producers’ price and retailers’ price. The policy implication is that the markets are co-integrated in terms of price transmission. However, the causality test shows that retailers are dominant over price determination. In other words, producer’s price is caused by retailers’ price; but producers’ price doesn’t cause retailers’ price. This shows that the market structure is in favor of retailers/traders, which can adversely affect the welfare of producers and consumers.

The achievements of market reforms in the third world countries rely to a large extent on the strength of price signals transmitted within various levels of market. With this regard the coordination between producers, wholesalers and retail prices keep up to be of a widely regarded economic interest. The relationship between farm and retail prices provides insights into marketing efficiency as well as consumer and producer welfare. Regarding to price theory, flexible prices are responsible for efficient resource allocation and price transmission integrates markets spatially [

In bargaining of prices, the point of agreement amongst processors is that well established relationship with retail buyers is essential. Retailers are sometimes considered as business people who bargain hard but realistically; on the other hand, some people think that retailers are dishonest towards buyers and include extra profit margins or do not let them share in rebates. Consumer’s institutions suggest that consumers are not benefited by price advantages of cheap and often subsidized imported dairy products. Consumer institutions indicate that many retailers would keep on their shelves dairy products of relative small processors at low prices as a way of “encouraging” the rest to “toe the line”. In some instances, retailers knowingly stock fresh milk from suppliers.

According to a study done by USAID, in Ethiopia, most dairy products are distributed through supermarkets. Currently in Addis Ababa there are 25 bigger supermarkets owned by 15 companies and out of which 12 are owned by Ethiopian and the other 3 by foreigners. Moreover other supermarkets also exist in major towns of the country which distribute milk and milk products processed by milk processing companies based in Addis Ababa [

The general objective of the study was to investigate milk price co-integration between producers and retailers in Ethiopia. More specifically, the study was undertaken;

・ To assess the short-term and long-term association between producers’ price and retailers’ price

・ To show the direction of causality that exists between producers’ price and retails’ price

Some of the main previous research study findings are highlighted in this research paper. These prominent research papers are further classified as per their focus area:

Tesfu [

Baghestany & Sherafatm [

Barahona et al. [

Octavio, F., Josef, B., & Jesus, C [

Katrakilidis C [

Rumankova L [

Bakucs, Z. Falkowski, J. Ferto, I [

Chalajour and Feizabadi [

Asche, Jaffry & Hartmann [

Bor, Ismihan & Bayaner [

The methodology of the paper is both descriptive and quantitative in nature. The study was conducted making use of secondary data extracted from Ethiopian Central Statistical Agency. The data was time series having 120 observations of monthly recorded price series of producers and retailers, for the period from January, 2004 to December, 2013. The data was analyzed using descriptive statistics and time series econometrics approach via STATA software. The descriptive analysis deals with comparison of trend of price movements for both producers and retailers. This involves graphical presentation of the price series and comparison of price variations using F-statistic. With regard to the quantitative analysis, test of co-integration and Vector Error Correction Model (VECM) were used to show how price in both stage of supply (producers and retailers) are integrated.

This is done to test for presence of non-stationarity co-variance as well as to determine order of integration of each variable. It is often expected that price levels exhibit non-stationary covariance, which may lead to autocorrelation problems in the price response functions. This may result in spurious regression when we estimate the relationship between the price series. Hence, the unit root test was undertaken to know if the monthly market prices are stationary or not, using Augmented Dickey Fuller test [

If we express the two prices (producers’ price and retails’ price) as an autoregressive process of order one as:

where:

α, β, ρ, and θ are constants

ε_{t} and ν_{t} are error terms

The Augmented Dickey-Fuller test involves regressing the first difference of these price series on own lagged values and testing for stationarity or non-stationarity, as shown below.

where:

The set of hypotheses is defined as:

Ho: γ = 0 for producers’ price (i.e. producers price series have a unit root or are non-stationary) and

Ho: φ = 0 for retails’ price (i.e. retails price series have a unit root or are non-stationary)

If the variables are non-stationary (or if we accept the null hypothesis), the co-integration test will follow.

After the stationarity test, we need to examine the existence of co-integration between the two variables. In this case, we search for the existence of the number of co-integrated vectors, r, within Johansen and Juselius’ [

where: P_{t} = is a (2 × 1) vector matrix of the producers and retail prices

e_{t} = are Gaussian residuals.

j = no. of lags in observation

The VAR in Equation (3) can be re-parameterized into a VECM form as:

where:

B_{j} is a (2 × 2) matrix of the short-run parameters,

Following Johansen’s procedure, the trace and maximum eigenvalue statistics are used to determine the rank of

In the third stage of our approach, we have to define the direction of causality between the two variables. Therefore, we implement a complete dynamic Granger-Engle VECM test of the following form as indicated in Reziti and Panagopoulos, [

where Z_{t}_{1−1} and _{1} and ᴨ_{2} are their respective coefficients and the β are short-run coefficients.

The set of hypotheses and options which are now available are as follows:

a) ᴨ_{1} ≠ 0_{ }and ᴨ_{2} ≠ 0 (a feedback long-run relationship between the two variables)

b) ᴨ_{1} = 0 and ᴨ_{2} ≠ 0 (producers’ price causes retails’ price in the long-run)

c) ᴨ_{1} ≠ 0 and ᴨ_{2} = 0 (retails’ price causes producers’ price in the long-run)

In this section, comparative analysis of variation of the two price series is reported. The trends of producers’ price and retailers’ price is presented using

The overall trends of the variation in the two price series were found to be the same as indicated by the F-statistics shown in

This section involves three steps/points of analysis including test of stationarity, test of co-integration, and test of causality

This test was undertaken to know if the variables have unit roots or not (if they are stationary or not), as well as to determine their order of integration, individually. As indicated in

To test for co-integration or fit co-integrating VECMs, we must first specify how many lags to include. Nielsen (2001) has shown that the methods implemented in lag-order selection statistics for VARs and VECMs can be used to determine the lag order for a VAR model with (1) variables. Accordingly, the lag-order selection statistics (LR, FPE, AIC, HQIC and SBIC) were computed. All these statistics show the same result that three lags

Variables | Observations | Standard Error | St. Deviation |
---|---|---|---|

Producers price | 120 | 0.2520926 | 2.761536 |

Retails price | 120 | 0.2593678 | 2.841232 |

Combined | 240 | 0.1841997 | 2.85361 |

Degree of freedom = 119, | F = 0.9447 |

Stage of supply | Producers | Retailers |
---|---|---|

Intercept | 0.1637712 | 2649805 |

P-value | (0.27) | (0.024) |

Price t-1 | −0.2707294 | −0.1750158 |

P-value | (0.000) | (0.001) |

First diff of price | 0.1122096 | 0.107915 |

P-value | (0.231) | (0.247) |

Trend | 0.020669 | 0.0144725 |

P-value | (0.000) | (0.001) |

L | 1 | 1 |

MacKinnon approximate P-value for Z(t) | 0.0046 | 0.0525 |

should be used in the estimation of the co-integration equation.

Once the number of lags was determined, the Johansen and Juselius’ framework was implemented to determine the number of co-integration equations. The estimation result is presented in

By now, we have assured that there is co-integration between the two price series, and we have identified that there is no more than one co-integrating equation. Given this, we need to test which price causes the other. This was analyzed using Engel Granger-Vector Error Correction Model, as applied by Reziti and Panagopoulos [

The negative sign of the estimate of the coefficient of adjustment parameter on producers price (adjustment_{p}) shows that when the average price of producer is too high, it quickly falls back toward the equilibrium level. Similarly, the positive sign of the estimated coefficient of adjustment parameter on retails price (adjustment_{r}) implies that when the average price of the producers are high, the retailers average price will quickly adjusts by rising toward the equilibrium level.

_{p}) has coefficient of −0.3738948 and P-value of 0.0000 implying that it is significant at 1% level of significance.

Rank | Eigen value | Trace | Max | ||
---|---|---|---|---|---|

Statistics | 5% critical value | Statistics | 5% critical value | ||

r = 0 | ------ | 36.3710 | 18.17 | 26.5210 | 16.87 |

r ≤ 1 | 0.19978 | 9.8500 | 3.74 | 9.8500 | 3.74 |

r ≤ 2 | 0.07944 | ----------- | --------- | -------- | -------- |

Number of obs = 119 Lags = 1 |

Dependent variable | Independent variables | Coefficient | Standard error | P-value | |||
---|---|---|---|---|---|---|---|

∆Pproducers | Adjustment_{p} | −0.3738948 | 0.0983138 | 0.000 | |||

Trend_{ } | 0.000069 | 0.0020562 | 0.973 | ||||

Constant_{p} | −0.0012291 | 0.1405618 | 0.993 | ||||

∆Pretails | Adjustment_{r} | 0.024 7243 | 0.0639977 | 0.699 | |||

Trend | 0.0010429 | 0.0013385 | 0.436 | ||||

Constant_{r} | 0.0172156 | 0.0914992 | 0.851 | ||||

No. of obs = 119 | |||||||

R-sq | chi2 | P > chi2 | |||||

Pproducers | 0.1203 | 15.86640 | 0.0012 | ||||

Pretails | 0.0287 | 3.424623 | 0.3307 | ||||

On the other hand, the adjustment parameter on retail price (i.e. adjustment_{r}) has coefficient of 0.0247243 and P-value of 0.699, implying that it is not significant even at 10% level of significance. This reveals that there is only one way of causality. In other word; the causality test shows that retailers are dominant over price determination. That means producer’s price is determined by retailers price or retailers price causes producer’s price. But, producer’s price does not cause retailers price.

Coefficient of the adjustment parameter when producers’ price becomes dependent variable, i.e. −0.3738948 shows the speed of adjustment of producers’ price when there is change in retailers’ price. It shows that for 100% change in price of retailers, the producers’ price changes by about 37% in a month. This implies it takes more than two months for the producers’ price to fully adjust if there is no additional shock in retail price. On the other hand, coefficient of the adjustment parameter when retailers’ price is independent variable, i.e. 0.0247243 shows that for 100% change in price of producers, retailers’ price changes by about 2.4% which is very insignificant, revealing that producers’ price does not significantly causes retailers’ price.

This study, by taking average prices of producers and retailers from nine (9) different regions in Ethiopia, undertakes the milk price co-integration analysis. The methodology of the paper is both descriptive and quantitative in nature. The descriptive analysis indicates that the two price series are moving together which can be a sign that there is an association between producers’ price and retailers’ price, for the period between 2004 and 2013. This situation is also revealed by the trends of the percentage change in the two price series. Secondly, result of the F-statistics shows that the extents of variations in the two prices are the same which is another indication that the two prices are associated. With regard to the quantitative analysis, the two prices were found to be associated which indicated the possibility of co-integration between them. This is also affirmed by the Johansen’s test of co-integration implying that there is possibility of long-run relationship between the two price series, as they are expected to have common stochastic trend overtime. The policy implication is that the markets are co-integrated in terms of price transmission. However, the negative sign of the estimate of the coefficient of adjustment parameter on producers price (adjustment_{p}) shows that when the average price of producer is too high, it quickly falls back toward the equilibrium level. Similarly, the positive sign of the estimated coefficient of adjustment parameter on retails’ price (adjustment_{r}) implies that when the average price of the producers are high, the retailers’ average price will quickly adjust by rising toward the equilibrium level. This shows the market structure is in favor of retailers/traders, which can adversely affect the welfare of producers and consumers. Thus, the concerned authorities and other stakeholders will have to take necessary action to improve market coordination among farmers, distributors, retailers and customers. In this manner, the production capacity as well as market sustainability of farmers and customers can be assured as well.

The authors would like to gratefully acknowledge the financial support of Diredawa University. The authors also grateful to Mr. Yonas Abera, and Mr. Besufikad Regassa for their valuable comments and suggestions to improve the quality of the paper.