China Policy Journal Volume 1, Number 1, Fall 2018 | Page 41

China Policy Journal
reports, and at the end of June (around
week 26 of the year), they have to submit
the allowances valid during year T-1 to
comply with their targets of year T-1. So,
the trading of emission allowances is relatively
active between April and June of
year T. We ran an auto-regression (AR)
model of logarithmic CEA price data
and used a Wald test to see if there is a
structural break of the coefficients after
the week 26 of each year. The results in
Table 3 show that the AR coefficients of
lnPrice t
do have structural changes either
after 2014 Week 26, or after 2015 Week
26, or after 2016 Week 26, in all ETS pilots
except Chongqing. Chongqing ETS
has very low-level trading activities and
its CEA price might be artificially set at
a certain level during weeks when trading
volumes were zero, not reflecting
the market demand and supply. Thus, it
is not surprising that Chongqing CEA
price does not exhibit compliance break
as other piloting ETSs.
2.2. Methods
This study uses time-series analysis
techniques. The Akaike Information
Criterion (AIC) is used for choosing the
number of lags in all tests. Augmented
Dickey-Fuller (ADF) tests are applied
to check the unit roots of variables. This
study performed three forms of ADF
tests: random walk model with drift,
random walk model with a trend, and
pure simple random walk model. The
Philips-Perron (PP) test is performed
with and without trend as a robustness
check.
To investigate the relationships
between energy prices and CEA prices,
this study first uses Johansen’s test to see
if there are long-run equilibrium relations
between energy prices and CEA
prices. Second, it performed time-series
regressions using the technique of multivariate
OLS or NW-OLS which works
with stationary data series. If there is a
significant serial correlation problem,
the Newey-West (NW) heteroscedasticity-and-autocorrelation-consistent
estimator
will be used. If not, the regressions
will be estimated with the robust estimator.
Three types of post-estimation tests
were performed: Durbin’s alternative test
for serial correlation, the Breush-Godfrey
serial correlation Lagrange Multiplier
test, and the joint F-test.
Therefore, the role played by
Brent crude oil price and coal price on
CEA price is estimated using the first
specification (Eq.1):
The second specification (Eq.2) takes account of stock index:
The third specification (Eq.3) adds the price data of LNG, which is only available
since 2014 week 1.
38