Estimation Model Determination in Panel Data

α 2-4 = coefficient X 2-4 = Independent variable

3. Common Model

This model assumes that both intercept coefficient and slope coefficient are constant in cross of place and time. Equation: Y it = α 1 + α 2 X 2 + α 3 X 3 + α 4 X 4 + ….. + µ 3.3 In which: Y it = Dependent variable α 1 = constant α 2-4 = coefficient X 2-4 = Independent variable µ = error

3.5.2 Estimation Model Determination in Panel Data

From three panel data method approaches, there are two approaches that are commonly used to estimate the regression model in panel data, they are fixed effect model and random effect model. To determine the method between pooled least square and fixed effect we test it through F test, and Hausman test will be used to determine between random effect and fixed effect Winarno, 2009. In fixed effect, the general regression equation panel data is Gujarati, 2003: Y it = β 1 + β 2 X it + β 3 X 3it + ... + β n X nit + u it 3.4 F test could be used to determine between pooled least square PLS model and fixed effect model. The equation is Gujarati,2003 1 2 2 2 k n R m R R ur r ur F − − − = 3.5 In which: R 2 r = R 2 PLS model R 2 ur = R 2 FEM model m = restricted variabel total n = sample total k = explained variable total Null Hypothesis of restricted F-test, is: H = Pooled Least Square restricted model H 1 = Fixed Effect unrestricted model From the equation above, we can get the result of value F statistics F table in certain degree of freedom α . This result allows us to reject null hypothesis H which states to choose PLS model, because of the above result, we should choose H 1 which states that we must use Fixed Effect model estimation in this research. Haussman test is used to determine between fixed effect method and random effect method. Chi Square value equation of Hausman test, is: Matrix b_diff = b_fixed – b_random Matrix var_diff = cov_fixed – cov_random Matrix qform = transposeb_diffinversevar_diffb_diff Null Hypothesis of Haussman test, is: H = random effect H 1 = fixed effect Suppose Chi Square statistic Chi Square table or in other word, while p- value 0.005. It means that we should reject null hypothesis H and determine that fixed effect model is the suitable model to use Winarno, 2009. Hausman test is also available through Eviews-6 command program. Slope coefficient is constant but there is a variation between individual intercept.

3.5.3 Econometric Model of the Research Study