Model Selection The Results of Error Correction Model ECM

35 Indonesian government amended the log export ban policy, which allows for export of several types of logwood. Logwood Sample Split Break Point Plywood Sample split Break Point Sawn wood Sample Split Break Point Figure 8 The Results of Chow Test

5.5 Model Selection

According to the results of stability analysis, this study tries to consider the presence of structural changes to obtain more precise analysis. This is conducted by introducing dummy variable in the standard VECM, which explains in which 36 point the structural break takes place. Based on the results of stability analysis, there are several dummy variables which are developed in the analysis: 1 Dlog0912 , which implies the presence of structural break in the last quarter of 2009 for logwood model, and 2 Dsawn0609, which represents the presence of structural break in the third quarter of 2006 for sawn wood model. Nevertheless, after introducing the dummy variables in the VECM for logwood and sawn wood, this study only finds the significant parameter for dummy in the logwood model. While in the sawn wood model, the dummy variable is statistically insignificant and also produces „strange‟ results of price transmission analysis, which might generate miss-understanding of interpretation. In spite of introducing dummy variable in the model, for plywood model, this study does not include any dummy variable since based on the sample split chow- test, it is found that all the parameters are stable and thus it does not show any structural break during the analysis, though this is contradictory with the results of the break point chow-test. According to this, therefore, this study employs the standard ECM for plywood and sawn wood, which assumes all parameters to be constants over time linear error correction model. Furthermore, the number of lags used in the error correction model refers to several criterions: Akaike Information Criterion, Final prediction error, Hannan-Quin criterion, and schwarz criterion which are available in the software package.

5.6 The Results of Error Correction Model ECM

The results of error correction model for logwood, as shown in the table, do not find any statistically significant parameter from different market prices, which indicates there is no short run response between domestic price and world price of logwood. As we can see in the table below, the dummy variable i.e. Dlog0912t-1 is statistically significant at any level of significance, which implies that there is a structural break in the last quarter of 2009. Based on this result, it is estimated the different speed of adjustment of price transmission between world and domestic price of logwood before and after the break period. Both before and after the break period, it is estimated that the speed of adjustment of the domestic logwood price to correct market disequilibrium is slower than the 37 speed of adjustment of the world logwood price. Before the last quarter of 2009, the speed of adjustment of the domestic logwood price to correct market disequilibrium is 5.8 percent, whereas the speed of adjustment of the world logwood price is 10.3 percent. Meanwhile, after the last quarter of 2009, the speed of adjustment of the domestic logwood price is 1.2 percent higher than that before the break period, as well as the speed of adjustment of the world logwood price becomes 2.2 percent higher. In addition, the estimated price adjustment of half-lives which implies the length of time required to adjust for half of the deviations from long-run market equilibrium before the last quarter of 2009 is calculated at about four months, and becomes faster after the last quarter of 2009 at about 3.19 months. Generally, the results from VECM do not confirm the Granger-causality test which tells there is no causality between the domestic logwood price and the world logwood price. Table 7 Error Correction Model for Logwood Variable Parameter Variable Parameter Δdlt Δrwlt ectt-1 -0.058 ectt-1 0.103 Δdlt-1 0.019 Δdlt-1 -0.094 Δrwlt-1 0.010 Δrwlt-1 0.445 Long run dlt-1 Intercept rwlt-1 0.0723 Dlog0912t-1 -0.211 Trendt-1 0.005 The results of error correction model for plywood market, as shown in the table, show a bit different finding than those in the logwood market. It is indicated that in the short-run, the response of the domestic plywood price to the world plywood price changes is estimated to be 3.7 percent, whereas the response of the world plywood price to the domestic plywood price changes is much higher : 23.2 38 percent. Furthermore, the speed of adjustment of the domestic plywood price is also slower than the speed of adjustment of the world price: 6.3 percent and 7.9 percent respectively. Based on this result, the estimated price adjustment of half- lives is a bit similar with that in the logwood market. Contradictory with the results in the logwood market, the VECM confirms the results of Granger- causality test which emphasizes the reciprocal causality from both domestic and world plywood price. Table 8 Error Correction Model for Plywood Variable Parameter Variable Parameter Δdpt Δrwpt ectt-1 -0.063 ectt-1 0.079 Δdpt-1 0.405 Δdpt-1 0.232 Δrwpt-1 0.037 Δrwpt-1 0.206 Long run dpt-1 Intercept 2.102 rwpt-1 0.594 Trendt-1 0.004 The error correction model for sawn wood market shows some different findings compared to logwood and plywood market. It is estimated that in the short-run, there is only response from the world price to the price changes of the domestic price, but the domestic price of sawn wood will not give any response to the changes in the world price. This short run response of the world price is quiet high: 29.4 percent, which is statistically significant at any level of significance. Furthermore, the speed of adjustment of the domestic sawn wood price to correct the market disequilibrium is faster than the speed of adjustment of the world sawn wood price: 7.02 percent and is 3.7 percent respectively. This situation is different with that in the logwood and plywood market which estimates that the speed of adjustment of the domestic price is slower than that of the world price. In addition, the estimated price adjustment of half-lives is calculated at six months. 39 Similar to the findings in the plywood model, the VECM confirms the results of Granger-causality test as well, which emphasizes both the domestic and the world price have an influence to each other. Table 9 Error Correction Model for Sawn wood Variable Parameter Variable Parameter Δdst Δrwst ectt-1 -0.072 ectt-1 0.037 Δdst-1 0.277 Δdst-1 0.294 Δrwst-1 -0.070 Δrwst-1 0.119 Long run dst-1 Intercept 3.509 rwst-1 0.350 Trendt-1 0.004

5.7 Residual Analysis