Market Integration of Poultry Products in Northwest of

Jafar Haghighat1 and Babak Abdolahi2*

Abstract Although poultry growth have not been much older than a few decades in their industrial form worldwide, they have been able to play an important role to provide the necessary amount of protein in society. Market integration of poultry products has caused an optimized allocation of sources and an increase of efficiency as well as farmers' income by means of caring about price fluctuations. According to the importance of market integration of poultry products, in the present study the following data have been used; Chicken product with weekly prices over the period of 1999-2010 for provinces like Eastern , Western Azerbaijan, Ardebil and Tehran, egg product with weekly prices over the period of 2006-10 for provinces like Eastern Azerbaijan, Western Azerbaijan and Ardebil. In order to achieve the goals of research, "Johansen" method have been used. The results of the research reveal market integration in chicken and egg products market. The law of one price is only true about egg product. In chicken market no providence are weak exogenous and in egg product market Eastern Azerbaijan is weak exogenous. According to the results, executing various policies in markets of each province for chicken product will only be transferred to other provinces in long term.

Keywords: Johansen Method, Market Integration, Northwest of Iran, Poultry Products

1 Department of Agricultural Economics, Collage of Agriculture, University, Tabriz, Islamic Republic of Iran. e-mail: [email protected], telephone number: +98 9144171908 2 Department of Agricultural Economics, Collage of Agriculture, Tabriz University, Tabriz, Islamic Republic of Iran. e-mail: [email protected], telephone number: +98 9354510903 Corresponding author

1

Introduction Since agricultural products are voluminous, corruptible and their production and consumption places are separated, spatial markets are considered important for such products. In fact agricultural products’ market is a market in which producers and consumers are not related directly, so such spatial distribution leads to price difference in different areas. Based on Theories, this price difference equals to caring expense only while factors like poor informing, products garbage and lack of transport equipment cause a significant difference in a product’s price in different areas (Rostamian, 2009). One of features agricultural products own is the state of being voluminous and corruptible considering the fact that production and consumption places are widely separated (Zanias, 1999). Another feature of agricultural product is their price fluctuations. Price stability is very important in developing or developed economics. However, it is unlikely to stabilize price of agricultural products and it is involved with price policies. Price fluctuations in agricultural products are due to a number of factors as follows: Seasonal changes (because of climate changes), cyclic changes (because of pause between decision to produce and production by producers), trend changes (because of climate situation, technical changes, demanders’ population and taste change) and unexpectable changes (natural disasters). That’s why, the relation between spatial distributions of prices means effectiveness of price changes from one place to another which happens constantly. The existence of integration in agricultural products’ market provides the opportunity of stable development in such products. Achieving the following goals seems to be possible in case of markets’ stability: A) Increase of Production and agricultural income Generally increase of risk leads to decrease of activities. The fact is more important in Iran’s agriculture because of low risk – taking tendency. Regional fluctuations of price are expected to be much more than fluctuations of the same product’s price in a country. Price fluctuations in dependent markets would be less than independent markets especially if the mentioned relation were influenced by one another. In addition risk of price would decrease, which leads to producers’ better decision and investment increase. These consequences would cause producers’ income increase. B) Improving policy in agricultural affairs In general, an agricultural policy is successful when there is enough information about the position of production, market and marketing (Aboonoori and Mojaverian, 2002). The present study provides valuable information about marketing balance of agricultural products

2

and the necessity of governments’ interfering or not interfering in marketing would be revealed. Checking the integration of agricultural products’ market has always been under researchers’ consideration. Several researches have been done about various products For instance, Mojaverian and Amjadi (1997). The mentioned researchers have studied the integration of market and law of one price in rice market of 5 cities in Iran. They have used monthly data’s of price index of rice retail price for five cities; Rasht, Tehran, Isfahan, Mashad and Tabriz. Ravallinon model has been used in order to achieve the goals of study. The results reveal that none of the markets are independent and there is no short run urgent relation among markets and the existence of a long run relation has also been proved among 13 pairs of markets out of 20 pairs. Nanang (2000), in a research called "A multivariate cointegration test of the law of one price for Canadian soft wood lumber market", studied the integration and the law of one price in Lumber market by means of seasonal data’s collected from 5 local markets of Atlantic, Quebec, Ontario, Prairies and British Columbia in the period of 1981–1997 Data’s used in this research are seasonal sale price index of Lumber in these regions. Results of ADF stationary test show that all series are stationary in the first difference and stationary test was done despite intercept and lack of time trend as well as seasonal dummy variables. The existence of only one convergence vector among series was proved by means of Johnson test and λmax and trace statistics which reveals integration of this product’s market in these regions. For the law of one price in the two regions, Likelihood Ratio test was used which was rejected in all cases of price equality between two regions. Peng and Marchant (2003) studied the spatial integration of price among local markets of beef in China. In order to achieve the goal, they considered monthly prices of beef in the period of 1998-2002 and studied price relation among markets in 13 provinces of China. They used Engel – Granger cointegration method and Error correction model to check market integration of beef. Results show that there are long run relations for most regional markets of beef in China and the existence of short run relation was rejected based on Error correction model. Akbarzadeh (2006) has studied the integration of rice market in two provinces Gilan and Tehran. He has used monthly price index of rice retail price for two types of rice, type one and two in 1997–98 as well as Granger causality method to state market integration. The results demonstrate that the integration is suitable for market integration and there is a long run relation between Tehran and Gilan market, besides rice prices in Gilan are influenced by those in Tehran. Falsafian and Zibaei (2006) have studied market integration and the law of

3

one price in cattle and sheep market collected from five provinces as follows: Eastern Azerbaijan – Khorasan, Khuzestan – Kerman and Khorasan – Isfahan. The results reveal that the mentioned markets are integrated for cattle while the law of one price is not true for any of them and also all mentioned markets are integrated for sheep except Khuzestan – Kerman and the law of one price is not for any of the markets. Shah Vali and Bakhshoodeh (2006), studied integration of main fisheries markets in three important regions. In the research, time series statistics of wholesale prices of fish and shrimp have been used in 1987-2002 and Engel – Granger test has been also used for long run relation among markets and the amount of connections among markets was studied by means of MCI. The findings demonstrated that in spite of long run connection among markets, there is no dependence among them in short run. MCI amounts show that the connection between north and south market is less than the connection between north and Shiraz market. Rostamian (2009) has studied the integration of fish market in seven provinces. He has used monthly data’s of retail price index since the first month of spring 2005 till the end of spring 2007. In order to achieve goals of the research Engel – Granger method, Error Correction and Johansen have been used. The results demonstrate that 9 pairs of markets are integrated out of 21 pairs; also among seven time series price, there are only two convergences vectors and the average balance speed ratio for studied markets equals to -0.37.

Method During the 1980s, economists discovered the fact that economic time series were not stationary. In other words usual statistical methods like linear regression are unvalid about non stationary data’s and leads to disappointment or unreliability of methods (Asche et al, 2004). There are two major views for market integration test: Engel and Granger test (1987) and Johansen test (1991). The former test is a two variable method which only studies relations between two series of price. That’s why it is impossible to analyze multi variable models (like the present study in which there are four variables for chicken). In addition this model is sensitive to normalized prices rather than other variables (Vinuya, 2007). Johansen method has been introduced to make up for Engel and Granger method. As we have just used Johansen method in the present study, it will be explained only. The beginning point of Johansen method for the test and stating cointegration relations among time series variables is estimating vector error correction model variables which are introduced as follows:

4

Yt 1Yt1 2Yt2 ...P1YtP1 Ytp Ut (1)

In this relation, Yt and its delays are vectors K×1 are related to pattern Variables such as time series of chicken price in provinces like Eastern and Western Azerbaijan, Ardebil and

Tehran,  i for i  1,..., P 1 equals to K  K pattern parameters matrixes; U t of k×1 vector is about pattern error terms.  Matrix contains information about long run balance relations. In fact    ' and in which , adjustment coefficient of imbalance and that is speed of adjustment toward long run equilibrium and  is matrix coefficient of long run equilibrium relationship. The main point in equation one is to state the rank of matrix  which equals to the number of independent convergence vectors. It is clear that if the mentioned rank equals to zero, the matrix is zero matrix and equation one is an ordinary VAR, as opposed to the mentioned vector being stationary, in case matrix  rank equals to "n". In the mean while, if the rank equals to one, there would be a convergence vector and Yt1 would be part of error correction and in other circumstances when matrix  rank is between one and n, there would be several convergence vectors. The number of independent convergence vectors would be estimated by means of significance of matrix  characteristic root test. The rank of a matrix equals to the number of non zero characteristic root of matrix (Noferesti, 1999). In practice, only access to estimations vectors of matrix  and its characteristic root seems possible. For those which show no significant difference with one, the following statistics will be used:

n ˆ trace(r)  T ln(1 i ) (2) ir1 ˆ  max(r,r 1)  T ln(1 r1 ) (3)

In which i includes estimated amounts of characteristic root gained by calculating matrix  which is called Eigen values and T equals to usable observations in estimation. When the suitable quantities of r are clear, these statistics are called trace statistic (λtrace) and maximum eigenvalue statistic (λmax). The former statistic related to the test of the null hypothesis equals to which the number of convergence vectors are less or same as r. The alternative hypothesis is that the numbers of convergence vectors are more than r. The latter statistic related to the test of this null hypothesis equals to which the number of convergence vectors equals to r. the alternative

5

hypothesis is that the number of these vectors equals to r +1. There is an important point; results of λtrace and λmax tests are likely to be paradoxical. In fact λmax test owns a clearer alternative hypothesis. However, in case of such paradox, choosing the minimum convergence vectors would be preferable. In order to state the efficient lag in equation one, Akaike Information Criterion (AIC) and Likelihood Ratio criteria (LR) have been used. For the test of separation of each province’s market from others, LR test has been used which posses  2 distribution. For this reason, we assume each province’s coefficient in matrix  to be zero and we test the reversed from. If being zero gets accepted, the province’s market is separated from others (Vinuya, 2007). If short run changes of a variable is not affected by lack of balance in long run relations and there is no reaction to other pattern variables, cointegration vectors will not be influenced by variables although they are parts of long run relations and affect other variables. In such situation, these variables are called weak exogenous. For weak exogenous test, we assume adjustment speed coefficent of each variable to be zero (α=0) and in case of accepting such assumption according to LR test, the variable would be weak exogenous and is in charge of leading price in province’s market (Asche et al, 2004 and Nanang, 2000). For the law of one price test, there must be a limit on convergence vector and in case of rejecting null hypotheses’ according to LR test; the mentioned law is not performed. Considering that there are more than two variables in the present study (4 variables for chicken and 3 for egg), the limit of a provinces’ coefficients equals to minus one and the addition of other provinces’ coefficients will be considered one; besides the law of one price for an n-variable model is performed only if there are n-1 (number) of cointegration vectors (Asche et al, 2004). If time series data’s are not stationary, cointegration models must be used for market integration and the law of one price. So it is essential to examine series’ stationary first. For such an examination there are various methods among which ADF3 and ERS4 have been selected for the present study. Necessary data’s and information in the present research include weekly price data of chicken in Eastern and Western Azerbaijan, Ardebil and Tehran over the periods 1999/3/23 to 2010/10/12. For egg products weekly data are for Eastern and Western Azerbaijan and Ardebil over the period 2006/3/21 to 2010/10/12.

3 Augmented Dickey-Fuller 1981 4 Elliott, Rothenberg and Stock 1996

6

Discussion Cointegration analysis using Johansen method is in need of all variables to own the same integrated order. Results of ADF and ERS tests are shown in table one. These tests have been done for, level and first difference of time series price logarithm. The efficient lag for tests was selected by means of modified AIC. As it is shown in table one, all time series would be stationary by just once differencing. That’s why mentioned time series possess the essential condition for performing integration test of market by Johansen method.

Table one: Results of variable stationary test. Variable Amount of Critical Amount of Critical ADF value in ERS value in 1 statistic 1 statistic percent percent+ Chicken price logarithm in Eastern Azerbaijan -0.72 (16)++ -3.99 1.08 (16) -2.56 (EA) First difference Chicken price logarithm in EA -16.46 -2.56 -16.46 -2.56 (1)*** (1)*** Chicken price logarithm in Western Azerbaijan -0.72 (16) -3.99 -0.72 (16) -2.56 (WA) First difference Chicken price logarithm in WA -16.46 -2.56 -16.46 -2.56 (1)*** (1)*** Chicken price logarithm in (A) -0.42 (16) -3.99 1.30 (16) -2.56 First difference Chicken price logarithm in A -12.68 -2.56 -12.67 -2.56 (2)*** (2)*** Chicken price logarithm in Tehran (T) -0.79 (12) -3.99 0.85 (12) -2.56 First difference Chicken price logarithm in T -14.82 -2.56 -14.83 -2.56 (1)*** (1)*** Egg price logarithm in Eastern Azerbaijan (EA) -2.79 (0) -3.99 -0.08 (1) -2.57 First difference egg price logarithm in EA -7.63 (2)*** -2.57 -7.69 (2)*** -2.57 Egg price logarithm in Western Azerbaijan -2.42 (12) -3.99 -0.28 (1) -2.57 (WA) First difference egg price logarithm in WA -9.36 (1)*** -2.57 -9.38 (1)*** -2.57 Egg price logarithm in Ardabil (A) -2.79 (14) -3.99 -0.21 (14) -2.57 First difference egg price logarithm in A -7.72 (2)*** -2.57 -7.77 (2)*** -2.57 + Critical value with intercept and trend for level of variables and without intercept and trend for first different of variables ++ Optimal lags in parentheses ***: Significant in level of one percent

Johansen cointegration test was estimated according to the quality of data’s with unrestricted intercept and no trend for chicken and with unrestricted intercept and restricted trend for egg, by means of AIC and LR; in addition to four lags for chicken and three for egg. The results have been demonstrated in tables 2 and 3. According to maximum eigenvalue statistic with a probability of 5 percent, three cointegration vectors among four time series for chicken and

7

two vectors among three variables for egg were proved. These results are signs of chicken and egg market’s integration among studied provinces. The results of LM test demonstrate lack of correlation in error terms of estimated models. Cointegration vectors for chicken product are as follow: LWA= -8.43 LA – 0.76 LEA + 10.11 LT LA= 0.53 LWA + 0.95 LEA – 0.35 LT LEA= 0.84 LWA + 0.12 LT + 0.11 LA Based on the above consequences only the third vector has an economical interpretation which tells one percent price increase in Ardebil, Tehran and Western Azerbaijan, lead to 0.11, 0.12 and 0.84 price increase in chicken price in Eastern Azerbaijan respectively. Cointegration vectors for egg product are as follow: LWA= 0.81 LA + 0.19 LWA LA= -0.43 LWA + 1.47 LEA Based on the above consequences, only the first vector possesses an economical interpretation which tells one percent price increase in Eastern Azerbaijan and Ardebil lead to 0.19 and 0.81 price increase of egg in Western Azerbaijan.

Table 2: results of Johansen test for chicken market. test Null hypotheses alternative statistic Critical value probability in five percent Maximum r  0 r  1 69.39*** 27.58 0.000 eigenvalue r  1 r  2 55.69*** 21.13 0.000 test r  2 r  3 36.46*** 14.26 0.000 r  3 r  4 2.32 3.84 0.127 Trace test r  0 r  1 163.87*** 47.85 0.000 r  1 r  2 94.48*** 29.79 0.000 r  2 r  3 38.79*** 15.49 0.000 r  3 r  4 2.32 2.84 0.127 LM  Stat : 15.51 (1) prob: 0.46 LM  Stat : 21.99 (52) prob: 0.14 ***: significant in level of one percent

Table 3: results of Johansen test for egg market. test Null hypotheses alternative statistic Critical value probability in five percent Maximum r  0 r  1 52.18*** 25.82 0.000

8

eigenvalue r  1 r  2 25.44*** 19.38 0.005 test r  2 r  3 8.93 12.51 0.184 Trace test r  0 r  1 86.57*** 42.91 0.000 r  1 r  2 34.38*** 25.87 0.003 r  2 r  3 8.93 12.51 0.184 LM  Stat : 10.99 (1) prob: 0.28 LM  Stat : 11.68 (52) prob: 0.23 ***: significant in level of one percent

To admit market’s integration test by LR test, we assume each province’s coefficient to be zero and being non zero for these coefficient was tested. If being null hypothesis is accepted for this test, province’s market is separated with other markets then. The results for chicken and egg products have been shown in tables 4 and 5 respectively. Results of LR test in probability level one percent are significant and it admits consequences of Johansen cointegration test. So markets of the four provinces are integrated for chicken and egg product and there are long run relations in province’s market. According to the fact that there must be three and two cointegration vectors for a four and three variable model respectively for the law of one price to be true, the law of one price test was done separately for products and the null hypothesis of test was significant in level one percent for chicken market but for egg market it was not significant even in level ten percent (table 6).

Table 4: Results of LR test for provinces separable in chicken market. Null hypotheses Statistic amount Critical value in P Value one percent Ardabil province coefficient=0 59.78*** 11.34 0.000 Western Azerbaijan province 37.76*** 11.34 0.000 coefficient=0 Eastern Azerbaijan province 41.14*** 11.34 0.000 coefficient=0 Tehran province coefficient=0 65.20*** 11.34 0.000 ***: significant in level of one percent

Table 5: Results of LR test for provinces separable in egg market. Null hypotheses Statistic amount Critical value in P Value one percent Ardabil province coefficient=0 29.61*** 9.21 0.000 Western Azerbaijan province 38.50*** 9.21 0.000 coefficient=0 Eastern Azerbaijan province 15.58*** 9.21 0.000 coefficient=0 ***: significant in level of one percent

9

Table 6: Results of the law of one price test. Mar Null hypotheses Statistic P Value ket amount Chi Ardabil province coefficient=-1, sum of other provinces 35.24*** 0.000 cke coefficient=1 n Egg Ardabil province coefficient=-1, sum of other provinces 0.39 0.819 coefficient=1 ***: significant in level of one percent

For examining the independence of one time series on others, in other words that province’s leadership in chicken and egg markets, weak exogenous test was used. This test’s null hypothesis of being can be true as long as α equals to zero. In such situation, the province’s price would be independent on other provinces’ and is in charge of price leadership. The results are shown in tables 7 and 8. Based on the consequences, the assumption of being zero for chicken markets for all provinces is significant in level one percent. So no provinces are weak exogenous and are in charge of price in chicken market. For egg market, the assumption of being zero in Western Azerbaijan and Ardebil is significant in level one percent but not significant even in level ten percent for Eastern Azerbaijan; That’s why it is weak exogenous to model and in charge of price leadership in egg market.

Table 7: Results of LR test for week exogenous in chicken market. Variable Null hypotheses Statistic P Value amount Ardabil province Ardabil province coefficient=0 22.42*** 0.000 Western Azerbaijan Western Azerbaijan province 33.08*** 0.000 province coefficient=0 Eastern Azerbaijan Eastern Azerbaijan province 17.51*** 0.000 province coefficient=0 Tehran province Tehran province coefficient=0 18.00*** 0.000 ***: significant in level of one percent

Table 8: Results of LR test for week exogenous in egg market. Variable Null hypotheses Statistic P Value amount Ardabil province Ardabil province coefficient=0 10.75*** 0.004 Western Azerbaijan Western Azerbaijan province 22.27*** 0.000 province coefficient=0 Eastern Azerbaijan Eastern Azerbaijan province 1.49 0.470

10

province coefficient=0 ***: significant in level of one percent

Consequences and suggestions The present research has studied market integration for poultry products in northwest of Iran using weekly data’s gained from cattle affairs supportive association and agricultural jihad organization over the periods of 199/3/23 to 2010/10/12. It has also studied market integration and the law of one price in provinces like Eastern and Western Azerbaijan, Ardebil and Tehran for chicken product; Eastern and Western Azerbaijan and Ardebil for egg product over the period of 2006/3/21 to 2010/10/12. Johansen econometric model has been used in order to achieve the goals. The advantages of the mentioned model are that it checks more than two variables in the model and the vital condition to use this model is the equality of variables stationary degree. As the studied variables had I(1), we were allowed to them. It can be consequence for chicken product that the market in Eastern and Western Azerbaijan, Ardebil and Tehran are integrated and significant in level one percent. Besides, LR test for each province to be separated in level one percent was significant and it admits four markets’ integration for this product, which shows price long run transfer. Law of one price has not been true for the markets of four provinces and price changes are not transferred in short run. Moreover results of weak exogenous test reveals lack of price leadership. Prices in Western Azerbaijan, Tehran and Ardebil have positive effects on the one in Eastern Azerbaijan. It could be concluded for egg product that the market in Eastern and Western Azerbaijan and Ardebil were integrated and significant in level one percent. Furthermore, LR test was significant in level one percent for each province’s separation and it admits that markets in the three provinces are integrated which reveals that prices are transferred in long run. The law of one price is true for markets of three provinces and price changes are transferred in short run. Eastern Azerbaijan is also in charge of price leadership, and prices in Eastern Azerbaijan and Ardebil affect Western Azerbaijan positively. The following tips could be recommended: As the markets of chicken product are integrated among provinces and the law of one price is not true in them, these markets must be considered as a single market in long run while individual ones in short run. The government could execute long run policies in each

11

province and wait for the consequences in others; but short run policies must be executed individually. As egg market is integrated among provinces and law of one price is true in them, these provinces are considered as one single market and it is highly efficient. That’s, why if the government execute short run policies, they would be transferred to other provinces, too. In other words it is not necessary to execute policies in all provinces.

12

References

- Aboonoori. A and Mojaverian. M. 2002. Investigation the Law of one price in crop product of Iran. Business Research Letter, 25: 85- 126. (In Persian) - Agricultural statistical year book. 2007 and 2008. (In Persian) - Akbarzadeh. M. 2006. Investigate rice market integration in Iran. 5th biennial agriculture economic conference. Iran. (In Persian) - Asche. F, Gordon. D.V and Hannesson. R. 2004. Test for market integration and the low of one price: The market for Whitefish in France. Marine Resource Economic, 19:195- 210. - Cattle affairs supportive association. 2010. www.iranslal.com - Dickey. D.A and Fuller. W.A. 1981. Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica, 49: 1057-1072. - Elliot. G, T.J. Rothenberg. and J.H. Stock. 1996. Efficient tests for an autoregressive unit root. Econometrica, 64: 813-836. - Engle, R.F. and Granger, C.W.J. 1987. Cointegration and correction: representations, estimation and testing. Econometrica, 55: 251-276. - Falsafian. A and Zibaei. M. 2006. Market integration and the law of one prices meat market of cattle and sheep in selected provinces. Since and Agricultural Industry, 19 (1): 173-180. (In Persian) - Johansen, S. 1991. Determination of cointegration rank in the presence of a linear trend. Oxford Bulletin of Econometric and Statistics 54: 383-397. - Johansen, S. and K. Juselius. 1990. Maximum likelihood estimation and inference on cointegration with application to the demand for money. Oxford Bulletin of Economics and Statistics, 52:169-210. - Mojaverian. M and Amjadi. A. 1997. Investigate spatial market integration and the law of one price. Agricultural economic and development Journal, 18: 165-187. (In Persian) - Nanang, D.M. 2000. A multivariate cointegration test of the low of one price for Canadian softwood lumber markets. Forest Policy and Economics, 1:347-355. - Noferesti. M. 1999. Unit root and cointegration in econometric. Second publication, Ghazal press, Tehran. (In Persian) - Peng, X. and Marchant, M. 2003. Spatial price linkages between Chinese regional beef markets. Southern Agricultural Economics Association Annual Meeting.

13

- Rostamian. R. 2009. Investigate market integration of fish meet in Iran. 7th biennial agriculture economic conference. Iran. (In Persian) - Shah Vali. A and Bakhshoodeh. M. 2006. Investigate integration fishery market in Iran. Economical research, 15: 69-85. (In Persian) - Vinuya. F.D. 2007. Testing for market integration and the law of one price in world Shrimp markets. Aquaculture Economics and Management, 11:243-265. - Zanias, G.P.1999. Seasonality and spatial integration in agricultural (product) market. Agricultural Economics, 20:225-262.

14