Developing and Testing a Predictive Model for Political Communication: The Case of Facebook Analysis of the 2016 Philippine Presidential Election Results

Abstract

The focus of the study is to develop and test a predictive model for political communication through the use of Facebook Analytics and determine how Facebook campaign likes and shares, and the presidential debate likes and shares predict the 2016 Philippine Presidential Election.

The data was gathered from the official Facebook page of the candidates through the use of social net importer. The candidates who reached the highest number of engagement based on likes and shares were identified. Then the model was created and tested in order to identify which model has the best fit. Linear regression was used to analyze the statistical relationship of the independent variables to the dependent variable. The model which best predicts the election result has high-r-square results so it can better explain the linear relationship of the variables. Also, it has low Root Mean Square Error (RMSE) which means that it has the smallest error in the regression line and the P-value should not reach the maximum risk of 0.05 to ensure that it is significant.

After testing the data into the model, the presidential debate was the only significant variable in predicting the election result. Presidential candidate Rodrigo Duterte has the highest number of presidential debate post likes that reached up to 1,361,995 and he was declared by COMELEC as the runaway winner who garnered a total of 16,601,997 votes. Hence, this verifies that the model has predicted the 2016 Philippine Presidential Election results.