Page 157 - Contributed Paper Session (CPS) - Volume 3

CPS1979 Francisco N. de los R.
decay term was introduced to the spatial variance matrix in accordance to
Tobler’s principle. The term is ( ) = −3 where is the inter-centroid

distance between areal units and . A piecewise mean function was also
generated for two clusters: voter turnout > 80% where the mean turnout is
83% and voter turnout is 73%. The logit transform of the proportions are
generated from a multivariate Gaussian distribution with a mean of 0.99 (logit
corresponding to the mean turnout of 73%) or 1.73 (logit corresponding to
the mean turnout of 83%). Low, moderate and high (on account of Cotabato
City) spatial variation scenarios were investigated. The Bayesian model was
fitted at = 20,000 , 30,000 40,000 MCMC samples. Parameter estimation
proceeded after 50% burn-in. Data integration, computation of proximity
matrix and testing for spatial autocorrelation were done in Geoda. Modelling
was done in R CARBayes package with extensive use of the S.CARleroux
function.

3. Results
Voter turnout was generally high in the NLE of 2016 as indicated by an
average of 79% across the 86 areal units (Figure 1). The special province of
Cotabato City was outlying with a turnout of only 48%. There was a significant
positive spatial autocorrelation in voter turnout (Moran’s I = 0.22, p = 0.0067).
This indicates that areas with relatively higher voter turnout are spatially close.
A similar conclusion can be said of areas with relatively lower voter turnout.
There is suggestion of parity in voter participation across provinces as
evidenced by a dissimilarity index of D=0.15. The 95% confidence interval is
(0.113, 0.175) based on 10,000 bootstrap samples.

Figure 1. Registered Voters, Actual Voters and Voter Turnout in the
Philippine National and Local Elections of 2016

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