Cointegration and Error Correction Models 1

2 Producer Price-Wholesaler Price The second relationship is between producer price and wholesaler price. Table 7 shows that there are no sufficient evidences that producer price and wholesaler price are cointegrated. The Johansen Trace test cannot reject the null hypothesis of no cointegrating vector. It indicates that producer price and rice miller price are not integrated. Price determination of wholesaler price is not related with producer price, and vice versa. Each price is determined by its previous prices autoregressive. They behave in the different way both in the short-run and the long-run. Table 7 Cointegration test for producer price-wholesaler price 3 lags Johansen Trace Test Level of significant r0 LR P-value 90 95 99 16,95 0,1358 17,98 20,16 24,69 1 4,68 0,3317 7,60 9,14 12,53 The factors which may influence this relationship are transaction costs, market information, market power, and market operation policy. Transaction costs to deliver rice product from producer to wholesaler are high because the rice product has to pass the other marketing intitution, the rice miller. The rice miller cost potentially increases the transaction costs to deliver rice from producer to wholesaler. As we stated before, transaction costs potentially inhibit the price transmission. Transaction costs consist of processing cost, storage cost, transportation cost, and marketing margin between producer and wholesaler. The indirect interaction also influences the transfer of market information. Farmers as producer commonly do not have access about the demand and the price in wholesaler market. Therefore, there are no price adjustments between producer and wholesaler. In addition, wholesaler can store the rice product in warehouse and manage its supply. If we compare the price volatilities of them, wholesaler price volatility 18,1 is lower than producer prices volatility 24,9. This ability became market power for wholesaler to control its price and to inhibit the high seasonal price volatility from producer price. Therefore, they get price which is not really big different either in harvest time or lean time. The Market Operation policy by Bulog to reduce high price also help wholesaler to have relatively stable price. In this case we use Cipinang Central Market price as wholesaler price. The rice price in Cipinang Central Market is the reference market price for the stability and availibility of rice stock in Indonesia. So, Bulog always try to mantain the stability and the availibility of rice stock in Cipinang Market by Market Operation policy. When stocks from wholesale market are secure for retail market, the retailer price for consumer are also secure to be stable. The Table 8 shows the Error Correction Model result for producer- wholesaler. We can see there are no significant adjustments each other. This supports the conclusion that producer and wholesaler are not integrated. Table 8 Error Correction Model for producer price-wholesaler price Dependent variable ΔWP t ΔPP t Independent variable parameter t-value parameter t-value ECM t-1 -0,207 -3,346 -0,008 -0,200 ΔWP t-1 0,200 2,014 0,072 1,128 ΔPP t-1 0,114 0,670 0,085 0,781 ΔWP t-2 -0,069 -0,713 -0,024 -0,382 ΔPP t-2 -0,215 -1,386 -0,376 -3,765 ΔWP t-3 0,086 0,929 0,001 0,017 ΔPP t-3 -0,145 -0,860 0,047 0,434 Long-run Equilibrium: WP t-1 = -868,476 + 1,640 PP t-1 + u t-1 -4,867 -19,865 Note: One , two , and three asterisks indicate rejection Ho: = 0 at 10, 5, and 1 level of significance, respectively. Critical values for 10 = 1,645; 5 = 1,96; and 1 = 2,576 . 3 Producer Price-Retailer Price The third relationship is between producer price and retailer price. The Table 9 provides sufficient evidences that producer price and the retailer price are cointegrated. The Johansen Trace test rejects the null hypothesis of no cointegration, but fails to reject the null hypothesis of one cointegrating vector. It indicates that producer price and retailer price are integrated. Table 9 Cointegration test for producer price-retailer price 3 lags Johansen Trace Test Level of significant r0 LR P-value 90 95 99 28,76 0,0020 17,98 20,16 24,69 1 7,59 0,1006 7,60 9,14 12,53 But if we see the long-run equilibrium in the Error Correction Model see Tabel 10, the relationship between producer price and retailer price is implausible. It is not accordance with the theory and unrealistic to be interpreted economically the relationship has negative sign, it should have positive sign. Therefore we cannot acknowledge that producer price and retailer price are integrated. Transaction costs are indicated to be one of the factors which influence this relationship. The transaction costs between producer in the upstream and retailer in the downstream are high. The rice product from producer has to pass rice miller and wholesaler before come to retailer. There are processing cost, storage cost, transportation cost, and marketing margin which may strongly interfere the price transmission between producer and retailer. There are also no communication infrastructures to tranfer market information between producer and retailer. So, the imperfect market information mistify the market signal for price adjustment. Producer price and retailer price are the significant determinant for price determination in their own market, producer price in the upstream market and retailer price in the downstream market. They behave in different way based on the conditions and the characteristics in the upstream market and the downstream market. They are independent each other. Producer price has high price volatility due to the seasonal price, meanwhile retailer price is strongly influenced by price stabilization policy. Therefore, there is no price transmission between producer price and retailer price. Table 10 Error Correction Model for producer price-retailer price Dependent variable ΔRP t ΔPP t Independent variable parameter t-value parameter t-value ECM t-1 0,005 5,018 0,003 2,453 ΔRP t-1 0,388 3,884 0,330 3,000 ΔPP t-1 0,217 2,393 -0,034 -0,341 ΔRP t-2 -0,374 -3,565 -0,473 -4,094 ΔPP t-2 -0,022 -0,257 -0,269 -2,853 ΔRP t-3 -0,144 -1,512 -0,058 -0,556 ΔPP t-3 0,245 2,715 0,178 1,791 Long-run Equilibrium: RP t-1 = -4179,049 - 4,564 PP t-1 + u t-1 -0,933 2,011 Note: One , two , and three asterisks indicate rejection Ho: = 0 at 10, 5, and 1 level of significance, respectively. Critical values for 10 = 1,645; 5 = 1,96; and 1 = 2,576. 4 Rice Miller Price-Wholesaler Price The fourth relationship is between rice miller price and wholesaler price. Table 11 shows there are no sufficient evidences that rice miller price and wholesaler price are cointegrated. Table 11 Cointegration test for rice miller price-wholesaler price 3 lags Johansen Trace Test Level of significant r0 LR P-value 90 95 99 17,00 0,1340 17,98 20,16 24,69 1 5,86 0,2093 7,60 9,14 12,53 The Johansen Trace test cannot reject the null hypothesis of no cointegration vector. This indicates that rice miller price and wholesaler price are not integrated. Each price is determined by its previous own prices autoregressive. Since cointegration result shows there is no cointegration between rice miller and wholesaler, it is not valid to interpret the Error Correction Model. So we do not need to interpret the Error Correction Model result for this case. Table 12 Error Correction Model for rice miller price-wholesaler price Dependent variable ΔWP t ΔMP t Independent variable parameter t-value parameter t-value ECM t-1 -0,194 -3,275 -0,018 -0,515 ΔWP t-1 0,172 1,700 0,046 0,791 ΔMP t-1 0,212 1,134 0,263 2,432 ΔWP t-2 -0,073 -0,745 -0,020 -0,355 ΔMP t-2 -0,236 -1,412 -0,477 -4,930 ΔWP t-3 0,090 0,933 0,013 0,227 ΔMP t-3 -0,119 -0,615 0,134 1,200 Long-run Equilibrium: WP t-1 = -883,319 + 1,610 MP t-1 + u t-1 -4,473 -18,021 Note: One , two , and three asterisks indicate rejection Ho: = 0 at 10, 5, and 1 level of significance, respectively. Critical values for 10 = 1,645; 5 = 1,96; and 1 = 2,576. Transaction costs to deliver rice product from rice millers in production areas to wholesale market in Jakarta, the Cipinang Market, are high. Transportation cost is indicated as the biggest contributor for the high transaction costs. In addition, the capability of wholesaler to store the stock in warehouse and manage time to release supply also influence the rice price determination in wholesale market. Market operation and import policy also help to mantain stable prices. Therefore the rice price in wholesale market can be controlled and it does not follow rice miller price changes. Rice miller price refers to producer price, meanwhile wholesaler price refers to retailer price. They are independent each other. 5 Rice Miller Price-Retailer Price The fifth relationship is between rice miller price and retailer price. Table 13 shows that there are strong evidences that the rice miller price and retailer price are cointegrated. The Johansen Trace test rejects the null hypothesis of no cointegration, but fails to reject the null hypothesis of one cointegrating vector. The cointegration indicates that rice miller price and retailer price are integrated. Table 13 Cointegration test for rice miller price-retailer price 3 lags Johansen Trace Test Level of significant r0 LR P-value 90 95 99 28,29 0,0025 17,98 20,16 24,69 1 6,52 0,1593 7,60 9,14 12,53 On the contrary, the Error Correction Model shows that it has the implausible long-run equilibrium because it is not accordance with the fact. The long-run equlibrium in the model says that when rice miller price increases IDR 1.000, the retailer price will increase IDR 9.241. The fact is that the retailer price is not too big different with the rice miller price. So, we cannot acknowledge that rice miller price and retailer price are integrated. The factors which may influence this relationship are the high transaction costs, imperfect market information, market power, and price stabilization policy. The rice product needs to pass the other marketing institution, namely wholesaler. There are also processing cost, storage cost, transportation cost, and marketing margin which contribute to the high transaction costs. In other side, there are no communication infrastructures to tranfer the market information between rice miller and retailer price. The imperfect market informations mistify the market signal and inhibit the price adjustment between them. The rice miller price determination is more influenced by producer price. Meanwhile, retailer price is more influenced by price stabilization policy. Therefore there is no price transmission between rice miller price and retailer price. Table 14 Error Correction Model for rice miller price-retailer price Dependent variable ΔRP t ΔMP t Independent variable parameter t-value parameter t-value ECM t-1 0,003 5,202 0,002 2,656 ΔRP t-1 0,338 3,356 0,280 2,731 ΔMP t-1 0,332 3,330 0,126 1,245 ΔRP t-2 -0,383 -3,657 -0,412 -3,863 ΔMP t-2 -0,071 -0,775 -0,367 -3,959 ΔRP t-3 -0,182 -1,905 -0,129 -1,323 ΔMP t-3 0,378 3,708 0,306 2,951 Long-run Equilibrium: RP t-1 = -8308,793+ 9,241 MP t-1 + u t-1 -1,215 2,700 Note: One , two , and three asterisks indicate rejection Ho: = 0 at 10, 5, and 1 level of significance, respectively. Critical values for 10 = 1,645; 5 = 1,96; and 1 = 2,576. 6 Wholesaler Price-Retailer Price The last relationship is between wholesaler price and retailer price. Table 15 shows that there are strong evidences that wholesaler price and retailer price are cointegrated. The Johansen Trace test rejects the null hypothesis of no cointegration, and fails to reject the null hypothesis of one cointegrating vector. The cointegration indicates that wholesaler price and retailer price are integrated. Table 15 Cointegration test for wholesaler price-retailer price 2 lags Johansen Trace Test Level of significant r0 LR P-value 90 95 99 29,64 0,0014 17,98 20,16 24,69 1 11,58 0,0159 7,60 9,14 12,53 Table 16 shows the result of Error Correction Model. In the short-run, the shocks from the retail market are transmitted instantaneously to wholesaler price significantly. The different sign of the coefficient retailer price changes show there is complex relationship between wholesaler price and retailer price. The coefficient of error correction term shows it will adjust 0,044 of the divergence to long-run equilibrium each month. The long-run equilibrium is achieved when the wholesaler price increases about IDR 1.000, the retailer price will increase about IDR 1.391. Table 16 Error Correction Model for wholesaler price-retailer price Dependent variable ΔRP t ΔWP t Independent variable parameter t-value parameter t-value ECM t-1 0,038 4,861 0,044 3,135 ΔRP t-1 0,411 4,366 0,350 2,051 ΔWP t-1 0,142 2,760 0,098 1,056 ΔRP t-2 -0,347 -3,969 -0,505 -3,185 ΔWP t-2 0,014 0,266 -0,101 -1,062 Long-run Equilibrium: RP t-1 = 1.391,230+ 1,220 WP t-1 + u t-1 1,848 -6,719 Note: One , two , and three asterisks indicate rejection Ho: = 0 at 10, 5, and 1 level of significance, respectively. Critical values for 10 = 1,645; 5 = 1,96; and 1 = 2,576. Wholesaler and retailer do direct interaction in the trade. So they have strong linkage in product flow which influence the price determination in both markets. The transaction costs factor between them are indicated to be relatively low, because commonly wholesalers and retailers take place in one region. There is no significant trade barrier between them, even the price stabilization policy in the retail market become incentive for wholesaler. The price stabilization policy creates stable price in retail market along the year. It stimulates wholesaler to keep on adjusting the demand and the price in the retail market continuously along the year. Retailer also keep on adjusting the price in the wholesaler market, because the price in wholesaler is the input cost for retailer and the supply from wholesaler to retailer will influence the price determination in the retail market.

5.3 Discussion and Policy Implications

The results of this study are interesting, there are some results appropriate with the initial expectation, but there are also some unexpected results. The fact the data have been collected from various institutions with possible difference way of agregating data. This is might be the source of bias. Agregated data have potential tendency of being less variative compared to individual data. So there is always a trade off of loosening variability in using agregated data instead of individual data. There might be human or systematic error in data collection. However this study has no sources to confirm such kinds of biases. Therefore this study assumed that all possibility of data collection and systematic error did not happened. In this way we keep our results the way they are found out. Table 17 The summarize of market integration relationships Model Producer Rice miller Wholesaler Retailer 1 2 3 4 5 6 Table 17 depicts that the rice market chain in Indonesia is segmented. The price changes from the retail market are not transmitted to the production point completely in the market chain, and vice versa from producer price to retailer price. There are only two cointegrating relationships between producer price and rice miller price and between wholesaler price and retailer price. Both producer- rice miller and wholesaler-retailer have direct interaction. In addition, there are two price determinations in the rice market chain in the upstream and in the downstream of market chain. Both of them have strong influence for price determination in the each stages, producer and rice miller in the upstream market and wholesaler and retailer in the downstream market. Therefore, it is like there are two markets which have own price and independent each other, whereas there are rice product flows between them. The other relationships in the market chain confirm that the Indonesian rice market chain is segmented. The relationship between rice miller price and wholesaler price in the mid of rice market chain is not integrated, even though they interact directly. It is like a bridge that connects the point of production in the upstream and the retail market in the downstream is broken. The three relationships which cross this bridge are also not connected. We can see that producer price is not integrated with wholesaler price, as well as rice miller price and retailer price. These findings are plausible which also confirm the previous results. The factors which determine the price transmission and market integration are not explained by models, but we can analyze by referring to the real conditions of market chain then confirm to the literature reviews. Based on the results analysis above, we found that the factors which may influence the price transmission in Indonesian rice market chain are seasonal price effects, market power to manage supply, trade barriers, direct or indirect interaction related to product flows effects on price determination, transaction costs, transparancy of market information, and price stabilization policy. Price determination in the farm market is strongly influenced by seasonal rice production. The high production in the first harvest and low production in the second and the third harvest create high volatility of producer price. There are no arrangements on cropping patterns in production areas, so farmers grow paddy together simultaneously during the rainy season. Meanwhile in the dry season, not all farmers grow paddy due to the limitation of water. There has been no for new technology to grow paddy in dry season or in dry land which can produce rice as good as in the rainy season and can be applied massively. The domestic purchasing by Bulog cannot dampen the high volatility of seasonal price significantly. Rice miller price will be influenced by producer price with lower price volatility than producer price, but then rice miller price is not transmitted to wholesaler price. Wholesaler blocks the seasonal price effects transmitted to retailer price or consumers. They have ability to storing rice for the stocks then manage the supply to provide stable supply for the retail market. The high price volatility of producer price inhibits the price transmission along the rice market chain. Therefore, the policy implication based on this factor is that the government should manage the price stabilization for producer price through overcome the unstable supply along the year. This policy can be implemented through managing the cropping pattern in the rice production areas. In addition, as the typical smallholder farmers in the developing countries, farmers do not have strong power to store the rice production in harvest time. They always sell their rice production immediately in harvest time due to immediate cash needs. They cannot control rice production supply into market as well as to control the rice price. Strengthening farmers’ institution is also needed to do. Reinforces “Lumbung padi” for the farmers may can be solution to control the rice supply from production. This is the traditional warehouse which is used by Indonesian ancient farmers in the past to storing the yield in the harvest time for consumption in the dry season. When farmers can control the rice production and can provide stable rice supply into market, government does not need to do price stabilization policy in the consumer market. How they interact in the trade also influence the price transmission. Direct or indirect interaction among market institutions related to the product flows effects on price determination and the transaction costs. The market institutions which interact directly have strong linkages in product flows. The strong linkage in product flows influence the price determination in each market related to supply and demand which determine the equilibrium price in the market, with no trade barrier assumption. Direct or indirect interaction also influence the transaction costs. The marketing institutions which do not interact directly will need more transaction costs for transportation cost, processing cost, storage cost, and marketing profit because the rice products have to pass the other marketing institutions. The big transaction costs inhibit the price transmission in the rice