Page 18 - Contributed Paper Session (CPS) - Volume 7

CPS2014 Ma. S.B.P. et al.
while VAR(1) produces high MAPE for time series of shorter length. The
procedure remains robust even for shorter t. The estimated MAPE by the
procedure is recorded to be at most 15 % which is observed at = 30.
Lower APB is observed in the estimates when < for = 20 while range
of APB remains the same for = 30,50 when < or > . It is also observed
that the APB of autocorrelation coefficient components of small values (11 =
0.10 , 12 = 0.15,22 = 0.20) are relatively larger than with the APB of larger
coefficient (21=0.95). The MAPE is robust with the choice of p and m over the
different series length.
Infusing the misspecification error leads to minimal changes in the
expected MAPE. The expected MAPE is also relatively higher for t=30. The
estimated MAPE of the proposed procedure is robust and remains superior
over VAR(1) in the presence of misspecification error. The procedure
consistently produces low MAPE across the specified values of t while VAR(1)
has the highest MAPE when = 20. The APB of the estimates produced by the
proposed procedure are also robust in the presence of misspecification error
except when < for = 50 wherein changes in APB are relatively higher.
Moreover, the standard error of the estimates are lower for = 20 compared
to the standard error of the estimates produced by VAR(1).
Generally, simulation results show that the postulated model is fairly robust
to misspecification error. Furthermore, the predictive ability of the estimated
model is better compared to VAR(1) over different lengths of time series.
The proposed estimation procedure is applied in a series of short annual
data set from 1995 to 2015. The goal of the analysis is to determine how the
contemporaneous effects of the total number of graduates from the University
of the Philippines (UP) system (excluding UP Manila), the budget allocated for
the UP system, and the board exam passing rate of UP for various Professional
Regulation Commission (PRC) licensure examinations simultaneously affect
the Real Gross Domestic Product (GDP) and Gross Value Added (GVA) in
Agriculture, Hunting, Fishing, and Forestry from 1995 to 2015.
The contemporaneous effects of the input series are accounted by taking
up to the fourth lagged values of the input series. The first sparse principal
component is considered in the analysis as it explains 83.32% of the variability
of the input series as shown in Table 1.
The estimated output autocorrelation measures the relationship between
the bivariate components 1,(( )) and 2,(log( )) with their past
values and with the past value of the other output component (1, 2,−1
2, 1,−1). The derived components of the estimate as shown in Table 2
explains the following relationship given the contemporaneous effects of the
input series up to the fourth lag: a 36.49 % increase in the growth rate of real
GDP at results from a proportionate increase in GDP growth rate at − 1; a
36.29 % increase in GDP growth rate at results from a proportionate increase

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