Page 159 - Contributed Paper Session (CPS) - Volume 3

CPS1979 Francisco N. de los R.
Table 1. Results of Bayes Modelling of Voter Turnout in the Philippine
National and Local Elections of 2016

4. Conclusions, Recommendations and Learning’s
Both Bayesian and non-Bayesian models revealed that there is low
dissimilarity in voter turnout among the 86 areal units contained in the official
Comelec dataset. When spatial variation is taken into account, there is
sufficient basis to say that the spatial variation is low. Thus, it is clear that
Filipinos participated well in the National and Local Elections of 2016 and quite
consistently homogeneous in pattern if taken spatially. As to the statistical
specification of the model, the case of the Philippines requires critical distance
of 500,000 UTM units to assure that areal units have at least one neighbor
based on inter-centroid. A proximity matrix can still be constructed for a
critical distance lower than this value but the algorithm fails to converge due
to provinces without neighbors. For the case of Palawan, one needs to override
the generated proximity matrix to force a neighbor under some special
criterion (here, transportation and trade relation). Localized smoothing is
beyond the scope of this study and is a welcome improvement moving
forward. Moving forward, the dissimilarity index in voter turnout should be
tracked over time to gather insight if the Philippine electorate is indeed
participative. The technique presented here may be also be applied to other
areal information with inherent spatial variation like poverty and health
statistics where strong spatial components are expectedly inherent.

148 | I S I W S C 2 0 1 9