China Policy Journal Volume 1, Number 1, Fall 2018 | Page 99
China Policy Journal
with those of 10 international cities
(e.g., New York, Paris). The 2012 version
of AQTI ranked 113 Chinese cities
monitored by the MEP as National Key
Environmental Protection Cities. The
MEP monitors 10 air pollutants (e.g.,
PM 10
, PM 2.5
, SO 2
, NO 2
, and so forth),
and their availability, completeness,
promptness, and user-friendliness on
urban EPBs’ websites were assessed. The
air pollutants were weighted by potential
health impacts and environmental
management practices. The total score
of AQTI ranges from 0 to 100 points,
and higher scores refer to higher levels
of environmental transparency.
IPE also developed the Pollution
Information Transparency Index (PITI)
to assess environmental transparency
of urban EPBs (IPE and NRDC 2009).
PITI evaluated the performance of
EPBs’ online disclosure of pollution-intensive
enterprises, clean production,
environmental impact assessment, and
other pollution-related information. In
this article, we use AQTI and PITI to
measure municipal governments’ environmental
transparency.
Control Variables
Individual-level variables such as gender,
age, education, and income that
may affect people’s subjective air quality
are included in the model. Gender
as a dummy is coded as 1 for male and
0 female. Age is measured by an ordinal
variable ranging from 2 (18–29) to
6 (above 60). Education is denoted by
an ordinal variable with four categories,
ranging from 1 (primary school or
below) to 6 (master’s degree or above).
Monthly income is gauged similarly by
an ordinal variable ranging from 0 (no
fixed income) to 14 (above 30,000 RMB
Yuan).
The demographics of the respondents
are similar with the latest census
data and our sample is largely representative
of the population of the sampled
cities. In the sample, 45.19 percent
of respondents were female, and 55.89
percent had college and above degrees.
The majority of the respondents (45.89
percent) aged between 18 and 29, and
those older than 60 accounted for 7.59
percent. In the sample, 62.84 percent of
the respondents earned monthly income
below 3,000 RMB Yuan, whereas rich
residents with income above 6,000 RMB
Yuan only accounted for 8.30 percent.
Analytical Methods
As our data structure is nested or multilevel
(individual citizens nested in
cities), we adopt MLM to test our hypotheses.
MLM is preferable to estimate
variances at multiple levels. In the model,
individuals are at Level 1 while cities
are at Level 2. MLM can simultaneously
estimate the variances at both Levels
1 and 2 (Raudenbush and Bryk 2002).
We center Level 1 predictors within
the cluster (group mean centering)
and center Level 2 predictors by grand
mean centering, which is appropriate to
estimate the same-level and cross-level
moderating effects in MLM (Enders
and Tofighi 2007).
In order to test our two hypotheses,
we need to estimate the moderating
2013)
96