Unit Root Test Co-integration Analysis

30 Figure 5 Plot Series of Logwood Figure 6 Plot Series of Plywood Figure 7 Plot Series of Sawn wood

5.1 Unit Root Test

Stationarity is a necessary condition in the time series analysis to produce unbiased estimation. Data is stationary when there is no unit root included, thus 31 testing of the stationarity is usually known as a unit root test. As most recent studies use, this study employed Augmented Dickey Fuller ADF test. All price variables are transformed into logarithmic form. As shown in the table 3, we can see that the values of test statistics of all variables are higher than -1.94 at the level implying that the null hypothesis of non-stationarity cannot be rejected for five percent level of significance. While for the first difference, the results are contradictory i.e. all value of test statistics are even less than -2.56 which means that the null hypothesis of non-stationairity is rejected at the one percent level of significance. According to these results, it can be concluded that all price variables in the model are not stationary at level but stationary in the first difference i.e. ~I1. Table 3 ADF test Result Variable Value of test statistics Number of Lags Akaike Level First Difference Level First Difference dl_log 2.9823 -12.7408 dp_log 2.5193 -6.8294 1 ds_log 3.6914 -8.4119 rwl_log 0.3503 -8.3923 1 rwp_log 0.3515 -10.7933 rws_log 1.1805 -14.4396

5.2 Co-integration Analysis

With the presence of un-stationary price series, than we proceed co- integration testing. As mentioned in the chapter 4, there are two techniques to check the co-integration relationships of wood price series between world and domestic market : two-step Engel- Granger and Johansen techniques. Based on the former technique, to check the co-integration relationship is by employing ADF test for the residuals from the regression between a set prices series i.e. world price and domestic price for any particular wood product as we can see in equation 2 in the chapter 4. The results of ADF test showed that all of the value of 32 test statistics for the residuals is statistically significant for rejecting the null hypothesis of non-stationarity. According to this, it can be concluded that based on the two-step Engel Granger approach, there are co-integration relationships between world price and domestic price for all wood products studied in this study i.e. logwood, plywood, and sawn wood. Table 4 The Result of Engel-Granger-Two-Step Procedure Wood Product Variable Parameter ADF Test for ut Logwood dlt -2.6431 Intercept 4.506 rwlt 0.157 Trendt 0.007 Plywood dp_logt -6.8727 Intercept 0.477 rwp_logt 2.668 Trendt 0.005 Sawn wood dst -4.7252 Intercept 0.220 rwst 4.132 Trendt 0.005 the asterisk denotes the value of test statistics in the ADF test Generally, the results from Johansen technique showed the similar conclusion with the Engel-Granger approach, implying the existence of co- integration for all wood product prices between world and domestic market. As we can see in the table 5, for plywood and sawn wood, the null hypothesis of r = 0, which implies that there is no co-integration relationship, is rejected at both five percent and one percent level of significance for all wood products. Meanwhile, for logwood, the null hypothesis of r = 0 is rejected for ten percent level of significance. The null hypothesis of one co-integrating relation i.e. r = 1, is not rejected for all wood products. Table 5 Results of Johansen Trace test 33 Wood Product r LR p-value Number of Lags Logwood 24.80 0.0659 2 1 5.92 0.4813 Plywood 34.99 0.0021 2 1 7.65 0.2905 Sawn wood 42.43 0.0001 1 1 4.47 0.6771

5.3 Granger Causality Test