China Policy Journal Volume 1, Number 1, Fall 2018 | Page 100
Subjective and Objective Air Quality in Urban China
effects of Level 2 variables (environment
transparency) on the relationship
between Level 2 independent variable
(objective air pollution) and Level 1 dependent
variable (subjective air quality).
The first case is referred to the moderating
effects at the same level (means
as outcomes model) while the other case
is cross-level (slopes as outcomes model).
We follow standard procedures to
detect moderating effects in MLMs and
probe graphically, which enables us to
intuitively interpret the results (Preacher,
Curran, and Bauer 2006). Varying-intercept
model is used to estimate
the direct effects while varying-intercept,
varying-coefficient model is used
to estimate the moderating effects.
Results
Descriptive Statistics
Table 1 reports the descriptive
statistics of our key variables. It
reveals that the sampled cities
vary substantially in objective air pollution
indicators. Air grade in 14 sampled
cities (or 43.75 percent) was rated
as III, with the other 18 cities graded as
II. Environmental transparency varies
drastically across sampled cities with
Table 1. Descriptive Statistics and Correlation Matrices
Variable Observations Mean Std. Dev. Min Max Correlation
Air quality satisfaction 25,139 6.328 2.075 1 10 1
SO 2
32 0.039 0.014 0.007 0.062 −0.0989*
NO 2
32 0.044 0.012 0.015 0.064 −0.0815*
PM 10
32 0.094 0.022 0.044 0.145 −0.182*
Air grade (III=1) 32 0.438 0.504 0 1 −0.153*
AQTI 31 32.355 19.607 9 76 −0.096*
PITI 31 50.555 17.216 23.2 83.7 −0.0051
Gender (male=1) 25,222 0.548 0.498 0 1 −0.0119
Age 24,939 3.069 1.260 2 6 0.117*
Education 24,988 3.712 1.250 1 6 −0.0438*
Income 23,595 3.348 2.780 0 14 −0.0464*
Note: The last column denotes the correlation coefficients between air quality satisfaction (the
dependent variable) and all independent variables. *p<0.05.
AQTI scores ranging from 9 to 76 and
PITI scores between 23.2 and 83.7.
The last column in Table 1 shows
the correlations between our independent
variables and the dependent variable.
The results suggest all air pollutants
are negatively associated with air
quality satisfaction and statistically
significant at the 0.05 level. The measures
of environmental transparency
are both negatively correlated with subjective
air quality measures, albeit only
the correlation coefficient of AQTI is
97