ANALISIS ARAH KAUSALITAS (CAUSAL ORDERING)

Abstract

When researchers test a model that represents the effect an independent variable on
another—dependent-- variable, many researchers commonly do not further investigate
about the correctness of the causal direction of the model. Hypothesis testing of such
model is generally done by assuring that the model coefficients are statistically
significant assuming that the direction of the causality is indeed correct. Hence, the
direction of the causality of these models is simply ex ante assumed, which means that the
direction could be incorrectly stated. The effect of this mistake could be enormous,
particularly if findings of the study, which adopt an incorrect causal order, are used for
policy makingt. This study discusses two approaches in testing the causal ordering of a
model, i.e., the Granger and Sim’s tests as well as SCDTs test of causality, which could
be either used in an experimental or nonexperimental setting. Findings of two empirical
researches written by Gudono (2006) and Chong and Chong (2002) are discussed and
used as an illustration.
(Keywords: causal ordering, lagged- regression, the sequential Chi-Square Differences tests (SCDTs), Type I, Type II, and Type III errors).