Model Selection External Thesis Examiner

45 Figure 4 Distortion from Annual Allowable Cut 1978-2007 Source: own calculation based on data from Ministry of Forestry of Indonesia

6.2 Model Selection

In order to analyze the impacts of wood prices on deforestation, this study applies the semi-logarithmic regression some price variables including domestic prices and world price on the distortion rate, as described in the equation 16 in please see chapter 4. There are some assumptions should be fulfilled in order to obtain the unbiased estimation. In this case, since we use only price variables as independent variable i.e. world price and domestic price for any particular wood product which are probably correlated each other, than we consider multicollinearity in setting up the model. Multicollinearity is a condition in which there is high correlation among the independent variables. With the presence of multicollinearity, the reliability of the model as a whole is not affected, but it affects any individual predictor. First step is regressing all the independent variables i.e. domestic price, world price, and export price for each wood products. Then it continues with calculating variance inflation factor VIF to obtain the information about mullticollinearity. Some previous studies consider the problem of multicollnearity when VIF is higher than 10 η‟Brien, β007 . -40000000 -30000000 -20000000 -10000000 10000000 20000000 30000000 40000000 50000000 60000000 70000000 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 46 Table 11 The Result of Regression Analysis before omitting Variable Parameter t-statistics VIF Distortion rate distor Domestic logwood price lnpdl -8.04e+07 -3.97 4.31 Domestic sawnwood price lnpds -0.15e+07 -0.08 2.08 Domestic plywood price lnpdp 4.81e+07 2.36 8.47 World logwood price lnpwl -0.74e+07 -0.27 6.08 World sawnwood price lnpws -6.02e+07 -1.90 15.08 World plywood price lnpwp 0.88e+07 0.23 11.53 Const 3.99e+07 4.64 R 2 0.6873 Adj R 2 0.6057 F6,23 8.4200 Prob F 0.0001 As we can see in the table 11, without considering the existence of multicollinearitiy, the price variables which have statistically significant effect on distortion rate are domestic price of logwood, and domestic price of plywood at five percent level of significance, and world price of sawn wood at ten percent level of significance. However, the first regression model contains a relatively high degree of VIF. According to the calculation, sawn wood world price variable has the highest VIF than is followed by world plywood price variable. To obtain more precise model, we omit these variables which contain high degree of VIF. After omitting those variables, the degree of VIF is reduced for overall model i.e. 3.09 implying that the model does not contain the problem of multicollinearity anymore and may produce better estimation, as shown in the table 12. 47 Table 12 The Result of Regression Analysis after omitting Variable Parameter t-statistics VIF Distortion rate distor Domestic logwood price lnpdl -7.79e+07 -3.72 4.32 Domestic sawnwood price lnpds -0.56e+07 -0.30 4.24 Domestic plywood price lnpdp 2.60e+07 1.72 1.94 World logwood price lnpwl -3.90e+07 -2.37 1.86 Const 4.01e+08 4.47 R 2 0.6305 Adj R 2 0.5714 F4,25 10.6600 Prob F 0.0000 As shown in the table 12, the domestic logwood price lnpdl and the world logwood price lnpwl are statistically significant at five percent level of significant and the domestic plywood price lnpdp at ten percent level of significant. It is shown that the domestic and world price of logwood variables have the negative signs implying that price have the positive correlation with the negative distortion. In other words, higher prices will likely lead to higher deforestation. However, the variable of domestic plywood price has positive sign which means the higher the domestic plywood price, the less the potential deforestation.

6.3 Discussions: The Impacts of Wood Prices on Potential Deforestation in