2012 PROCEEDINGS - Public Relations Society of America

Economic return was measured by the US imports from the client country. Imports from the four
client countries to the US were collected from foreign trade data available at the US Census
Bureau. These data, which include government and non-government shipments of goods, were
collected from the documents of the US Customs and Border Protection Agency of the
Department of Homeland Security. The dataset ―reflects the total arrivals of merchandise from
foreign countries that immediately enters the consumption channels, warehouses, or Foreign
Trade Zones‖ (US Census Bureau, 2012). Although these datasets contain reporting and data
capture errors, these errors can be treated as random. Matching public relations expenditure data,
the semi-annual dollar amount of the US imports from Japan, Liberia, Belgium, and Philippines
from 1996 to 2009 were collected.
Time series analysis
Time-series test analyzes the data observed sequentially in time. The time-series is
assumed to be stationary if the trend of the series can be forecasted based on time and if the
covariance between two time points are equal for every equal time difference (Box, Jenkins, &
Reinsel, 2008). Otherwise, the series is called non-stationary. This study conducted a unit root
test, called the augmented Dickey-Fuller (ADF) test, to examine whether each time-series (public
relations expenditure and the US imports) is stationary or non-stationary. To convert nonstationary
time series to stationary, the process called ‗difference‘ was applied (Box et al., 2008).
For the SAS procedure, the command, respectively, DIF (Variable) and DIF2 (Variable)
transform non-stationary series to 1 st order and 2 nd order difference of the series (SAS Institution,
2012).
After several tests, this study applied first order difference to the series of public relations
expenditure of and the US imports from Japan, the US imports from Belgium, and the US
imports from Philippines. Table 2 shows the result of the unit root test for each series. For the
model specification, this study checked the White Noise for the residuals after the model
constructed (Box et al., 2008; Brocklebank & Dickey, 2003).
[INSERT TABLE 2 ABOUT HERE]
This study focuses on the causal relationship between the four client countries‘ public
relations expenditure in the US and the US imports from those countries over 28 time points for
fourteen years. This study conducts a Vector Autoregressive (VAR) model to analyze the causal
relationship. The VAR model is suitable when the input series cannot be assumed to be
independent to each other and if the feedback from the output series to the input series is
expected (Brocklebank & Dickey, 2003). We chose this model because public relations
expenditure each year is hard to be independent to each other and it is not clear whether there is a
feedback from economic returns to public relations expenditures.
Based on VAR model, the Granger Causality Test and the Toda & Yamamoto Causality
Test were used for the analysis of causality. The Granger causality test is an effective tool to test
whether the direction of causal relationship is one-way or two-way. Toda and Yamamoto is an
effective tool to detect a long-term effect of time-series as it checks the result for all the different
lags.
Therefore, the present study tried both result for comparison and transformed time-series
of each country to natural logarithm in order to stabilize the variance, which is generally applied
to the analysis of the economic data (Cohen, Cohen, West, & Aiken, 2003). To select the optimal
lag length for the model, the Akaike Information Criterion (AIC) is checked with VAR model
based on the result of the unit root test presented on Table 2. The result of the AIC is shown in
Table 3.
110