ArticlePDF Available

How Does Air Pollution Influence Housing Prices in the Bay Area?

Authors:

Abstract and Figures

In this paper we examine the effects of localized air pollution measurements on the housing prices in Oakland, CA. With high-resolution air pollution measurements for NO, NO2, and BC, we can assess the ambient air quality on a parcel-by-parcel basis within the study domain. We combine a spatial lag model with an instrumental variable method to consider both the spatial autocorrelation and endogeneity effects between housing prices and air pollution concentrations. To the best of our knowledge, this is the first work in this field that combines both spatial autocorrelation and endogeneity effects in one model with accurate air pollution concentration measurements for each individual parcel. We found a positive spatial autocorrelation with housing prices using Moral’s I (value of 0.276) with the total sample number of 26,386. Somewhat surprisingly, we found a positive relationship between air pollution and housing prices. There are several possible explanations for this finding. Homeowners in high demand, low-stock housing areas, such as our study, may be insensitive to air pollution when the overall ambient air quality is relatively good. It is also possible that under clean air conditions, low variability in pollutant concentrations has little effect on property values. These hypotheses could be verified with more high-resolution air pollution measurements with a diversity of regions.
Content may be subject to copyright.
International Journal of
Environmental Research
and Public Health
Article
How Does Air Pollution Influence Housing Prices in the
Bay Area?
Minmeng Tang 1,* and Deb Niemeier 2


Citation: Tang, M.; Niemeier, D.
How Does Air Pollution Influence
Housing Prices in the Bay Area? Int. J.
Environ. Res. Public Health 2021,18,
12195. https://doi.org/10.3390/
ijerph182212195
Academic Editor: Elena Rada
Received: 16 September 2021
Accepted: 17 November 2021
Published: 20 November 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1Department of Land, Air and Water Resources, University of California, Davis, One Shields Ave,
Davis, CA 95616, USA
2Department of Civil and Environmental Engineering, University of Maryland, 1173 Glenn Martin Hall,
College Park, MD 20742, USA; niemeier@umd.edu
*Correspondence: mmtang@ucdavis.edu
Abstract:
In this paper we examine the effects of localized air pollution measurements on the housing
prices in Oakland, CA. With high-resolution air pollution measurements for NO, NO
2
, and BC, we
can assess the ambient air quality on a parcel-by-parcel basis within the study domain. We combine
a spatial lag model with an instrumental variable method to consider both the spatial autocorrelation
and endogeneity effects between housing prices and air pollution concentrations. To the best of
our knowledge, this is the first work in this field that combines both spatial autocorrelation and
endogeneity effects in one model with accurate air pollution concentration measurements for each
individual parcel. We found a positive spatial autocorrelation with housing prices using Moral’s I
(value of 0.276) with the total sample number of 26,386. Somewhat surprisingly, we found a positive
relationship between air pollution and housing prices. There are several possible explanations for
this finding. Homeowners in high demand, low-stock housing areas, such as our study, may be
insensitive to air pollution when the overall ambient air quality is relatively good. It is also possible
that under clean air conditions, low variability in pollutant concentrations has little effect on property
values. These hypotheses could be verified with more high-resolution air pollution measurements
with a diversity of regions.
Keywords:
air pollution; housing price; instrumental variable; spatial autocorrelation; spatial
lag model
1. Introduction
Air pollution is not only a major global risk resulting in high incidences of illness and
deaths [
1
,
2
], but can also produce external damages to different economic sectors, including
manufacturing, agriculture, transportation, and utilities [
3
]. In the U.S., air pollution costs
were roughly equivalent to about 5% of the yearly gross domestic product (GDP) in 2014 [
4
].
One sector we might expect to be highly sensitive to air quality is housing, and there are a
number of studies both nationally and globally focusing on the relationship between air
quality and housing prices.
The literature mainly relies on the construction of the hedonic price models to evaluate
the effect of air pollution on housing prices. The hedonic price model, commonly used in
economy, focuses on the relation between price and other corresponding features [
5
]. It
has been widely used in the housing market studies [
6
,
7
]. Some new regression methods,
such as neural network, quantile regression, and semi-log regression, have also been
applied in housing price prediction studies [
8
]. For the studies focusing on how air
pollution influences housing prices, we can divide the body of research based on the
approach. The first category uses an instrumental variable to address endogeneity effects
and frequently uses a variable that is not related to housing prices but directly related
to air pollution as the instrumental variable to determine the exogeneous part of the
variability from air pollution [
9
,
10
]. The endogeneity effect means correlation between
Int. J. Environ. Res. Public Health 2021,18, 12195. https://doi.org/10.3390/ijerph182212195 https://www.mdpi.com/journal/ijerph
Int. J. Environ. Res. Public Health 2021,18, 12195 2 of 13
the explanatory variable and the error term, which leads to the biased estimates using
the ordinary least square estimation method. The second group uses spatial econometric
models and hedonic price models to understand air pollution’s influence on housing
prices, accounting for spatial autocorrelation of housing prices. Spatial autocorrelation is a
term that is used to describe the systematic spatial variation in a variable. For example,
positive spatial autocorrelation, which is a more common situation, means that sites that
are located close together tend to have similar values. The most common spatial hedonic
models are the spatial lag model (SLM) [
11
,
12
], spatial error model (SEM) [
11
,
12
], spatial
Durbin model (SDM) [
13
], geographically weighted regression (GWR) [
14
,
15
], and quantile
regression models (QRM) [
15
]. The results from the literature are inconclusive: some
of the studies conclude that air pollution concentrations do not significantly influence
housing prices [
11
,
12
,
16
], while others find that air pollution concentrations negatively and
significantly influence housing prices [10,1720].
Previous studies have produced inconclusive findings, in part because there were
limitations to the approaches. For example, nearly all the studies consider only spatial
autocorrelation or endogeneity effects. Most studies rely on Moran’s I to measure spatial
autocorrelation [
21
], and results from cities in both China and the U.S. suggest that there
are positive and significant spatial autocorrelations in housing prices [
22
,
23
]. When air
pollution is added to the mix, the endogeneity effect on housing prices results in model
estimation and causal inference biases [
9
,
15
,
17
]. We depart from previous studies by
constructing a hedonic price model combining both spatial autocorrelation and endogeneity
effects to examine the relationship between housing prices and air pollution. To the best
of our knowledge, this is the first study combining these two effects to comprehensively
understand how air pollution influences housing prices. We also introduce high-resolution
air pollution mapping data into housing valuation studies. Prior research relied on air
pollutant data from a limited number of stationary monitors to underpin estimation for a
large region or a city. Our high-resolution mobile-based air pollution mapping data cover
every street within the study domain, which allows us to draw on much more accurate
ambient air quality measurements for each property.
2. Materials and Methods
2.1. Study Area
Our study domain includes three major areas within Oakland, California: West
Oakland (WO), Downtown Oakland (DO), and East Oakland (EO) (Figure 1). The WO
and DO areas together cover about 15 km
2
, with residential, commercial, and industrial
blocks, and the EO area covers about 15 km
2
with a mix of industrial and residential blocks.
The WO and DO areas have a total population of about 25,000, and the EO area has a total
population of about 58,000 [24].
2.2. Pollutant Concentration and Housing Valuation Data
Two Google street view mapping vehicles, carrying Aclima environmental intelligence
sensors, were deployed in the study area between June 2015 and May 2016. The dataset
covers the measurements of weekday daytime concentrations of black carbon (BC), nitric
oxide (NO), and nitrogen dioxide (NO
2
) with one second temporal resolution within the
study area (Figure 1). A mobile-based data reduction and aggregation algorithm was
developed by Apte et al. [
25
] to average the instantaneous measurements into median
annual weekday concentrations with 30 m resolution [
26
]. We used the high-resolution air
pollution concentration product from Apte et al.’s [
25
] supporting information as ambient
air pollution measurement in our study. Since meta-analyses have demonstrated that
the spatial extent of mobile sources is in the order of 100–400 m for particulate matter
and 200–500 m for NO
2
[
27
,
28
], we selected 400 m as the buffer size and calculated the
mean air pollution concentrations within the buffer area of each property to represent
the ambient air pollution concentrations. We also calculated air pollution concentrations
with a 100 m buffer and without any buffer. The results and conclusions were the same as
Int. J. Environ. Res. Public Health 2021,18, 12195 3 of 13
those produced with the
400 m
buffer. For the purposes of this paper, we used the
400 m
buffer air pollution concentrations calculated to ensure that we incorporate proximate
roadway-generated air pollution.
Int. J. Environ. Res. Public Health 2021, 18, 12195 3 of 14
Figure 1. Study domain.
2.2. Pollutant Concentration and Housing Valuation Data
Two Google street view mapping vehicles, carrying Aclima environmental intelli-
gence sensors, were deployed in the study area between June 2015 and May 2016. The
dataset covers the measurements of weekday daytime concentrations of black carbon
(BC), nitric oxide (NO), and nitrogen dioxide (NO2) with one second temporal resolution
within the study area (Figure 1). A mobile-based data reduction and aggregation algo-
rithm was developed by Apte et al. [25] to average the instantaneous measurements into
median annual weekday concentrations with 30 m resolution [26]. We used the high-res-
olution air pollution concentration product from Apte et al.’s [25] supporting information
as ambient air pollution measurement in our study. Since meta-analyses have demon-
strated that the spatial extent of mobile sources is in the order of 100400 m for particulate
matter and 200500 m for NO2 [27,28], we selected 400 m as the buffer size and calculated
the mean air pollution concentrations within the buffer area of each property to represent
the ambient air pollution concentrations. We also calculated air pollution concentrations
with a 100 m buffer and without any buffer. The results and conclusions were the same as
those produced with the 400 m buffer. For the purposes of this paper, we used the 400 m
buffer air pollution concentrations calculated to ensure that we incorporate proximate
roadway-generated air pollution.
The housing valuation data (shown in Figure 2) were provided by Estated, Inc.
(https://estated.com/ accessed on 13 August 2020), and include land, improvement, and
total value for every property within our study domain. The value of each property is
calculated based on tax assessment as provided by the county assessor. For each property,
the detailed structure information includes year built, stories, room counts, parking type,
construction type, and total area. Finally, sociodemographic variables at the census tract
level influencing housing price, including population density, income, and non-employ-
ment rate, were assembled using the 2016 American Community Survey.
West Oakland
Downtown
Oakland
East Oakland
Figure 1. Study domain.
The housing valuation data (shown in Figure 2) were provided by Estated, Inc.
(
Boulder
, CO, USA), (https://estated.com/ accessed on 13 August 2020), and include
land, improvement, and total value for every property within our study domain. The value
of each property is calculated based on tax assessment as provided by the county assessor.
For each property, the detailed structure information includes year built, stories, room
counts, parking type, construction type, and total area. Finally, sociodemographic variables
at the census tract level influencing housing price, including population density, income,
and non-employment rate, were assembled using the 2016 American Community Survey.
2.3. Methods
Following Kim et al.’s study [
16
], in which the SLM model specification outperformed
SEM on housing data in Korea, we constructed a spatial lag model (SLM) with an additional
instrumental variable to include both the spatial autocorrelation and endogeneity effects
(Equation (1)):
y=Xβ+λWy +ε,εN0, σ2(1)
where yis the logarithm of housing price, X’s are independent variables including an
instrumental variable,
β
are the estimated coefficients, Wis the non-stochastic spatial
weight matrix, Wy represents the spatial lag of the dependent variables, and
ε
is the error
term. For the spatial weight matrix, there are no widely accepted spatial structures for
housing price data, but some studies use the queen contiguity weighting matrix since it is
representative for contiguity-based weighting matrices [
12
]. Therefore, we used the queen
contiguity weighting matrix.
Int. J. Environ. Res. Public Health 2021,18, 12195 4 of 13
Int. J. Environ. Res. Public Health 2021, 18, 12195 4 of 14
Figure 2. Housing price spatial distribution in the study domain.
2.3. Methods
Following Kim et al.’s study [16], in which the SLM model specification outper-
formed SEM on housing data in Korea, we constructed a spatial lag model (SLM) with an
additional instrumental variable to include both the spatial autocorrelation and endoge-
neity effects (Equation (1)):
𝑦 = 𝑋𝛽 + 𝜆𝑊𝑦 + 𝜀, 𝜀~𝑁(0, 𝜎2)
(1)
where y is the logarithm of housing price, Xs are independent variables including an in-
strumental variable, β are the estimated coefficients, W is the non-stochastic spatial weight
matrix, Wy represents the spatial lag of the dependent variables, and ε is the error term.
For the spatial weight matrix, there are no widely accepted spatial structures for housing
price data, but some studies use the queen contiguity weighting matrix since it is repre-
sentative for contiguity-based weighting matrices [12]. Therefore, we used the queen con-
tiguity weighting matrix.
To address the endogeneity concern between housing prices and air pollution con-
centrations, we combined the instrumental variable (IV) method together with the SLM.
We used the mean of the median vehicle speed within the buffer area as the instrumental
variable, which is positively related to air pollution concentrations but is not correlated
with housing prices.
The spatial lag term in Equation (1) is an endogenous variable, and the instrumental
variable is an additional endogenous variable, which can result in difficulty in estimating
the model coefficients due to the extra endogenous variable. We used a two-step general-
ized moments (GM) and instrumental variable (IV) method to estimate the coefficients in
Equation (1) [2933]. All the calculations were conducted in R [34] and the two-step
GM/IV method is available in sphet package with function spreg [35,36].
Figure 2. Housing price spatial distribution in the study domain.
To address the endogeneity concern between housing prices and air pollution con-
centrations, we combined the instrumental variable (IV) method together with the SLM.
We used the mean of the median vehicle speed within the buffer area as the instrumental
variable, which is positively related to air pollution concentrations but is not correlated
with housing prices.
The spatial lag term in Equation (1) is an endogenous variable, and the instrumental
variable is an additional endogenous variable, which can result in difficulty in estimating
the model coefficients due to the extra endogenous variable. We used a two-step gener-
alized moments (GM) and instrumental variable (IV) method to estimate the coefficients
in Equation (1) [
29
33
]. All the calculations were conducted in R [
34
] and the two-step
GM/IV method is available in sphet package with function spreg [35,36].
3. Results and Discussion
3.1. Variable Distribution
Housing prices are not normally distributed (Figure 3A), so we applied the logarithm
transformation of housing prices (Figure 3B). The NO, NO
2
, and BC concentrations in
Figure 3C–E are the average of measurements within the 400 m buffer of each parcel.
We also tested using the logarithm transformation of NO, NO
2
, and BC concentrations
as input to our model, which obtained very similar results as using the concentrations
without transformation. Therefore, no transformation was applied to the NO, NO
2
, and BC
concentrations. Most parcels have NO concentrations less than 40 ppb, NO
2
concentrations
less than 25 ppb, and BC concentrations less than 1.5
µ
g/m
3
. The summary statistics of
housing prices and air pollution concentrations are shown in Table 1below.
Int. J. Environ. Res. Public Health 2021,18, 12195 5 of 13
Int. J. Environ. Res. Public Health 2021, 18, 12195 5 of 14
3. Results and Discussion
3.1. Variable Distribution
Housing prices are not normally distributed (Figure 3A), so we applied the logarithm
transformation of housing prices (Figure 3B). The NO, NO2, and BC concentrations in Fig-
ure 3CE are the average of measurements within the 400 m buffer of each parcel. We also
tested using the logarithm transformation of NO, NO2, and BC concentrations as input to
our model, which obtained very similar results as using the concentrations without trans-
formation. Therefore, no transformation was applied to the NO, NO2, and BC concentra-
tions. Most parcels have NO concentrations less than 40 ppb, NO2 concentrations less than
25 ppb, and BC concentrations less than 1.5 µ g/m3. The summary statistics of housing
prices and air pollution concentrations are shown in Table 1 below.
Figure 3. Distributions of housing prices (A), logarithm transformed of housing prices (B), and concentrations of NO (C),
NO2 (D), and BC (E).
Figure 3.
Distributions of housing prices (
A
), logarithm transformed of housing prices (
B
), and concentrations of NO (
C
),
NO2(D), and BC (E).
Table 1. Summary statistics of housing prices and air pollution concentration data.
Housing Price, USD NO Concentration,
ppb
NO2Concentration,
ppb
BC Concentration,
µg/m3
Sample size 26,386 a26,210 a26,210 a26,210 a
Mean 275,664.1 10.293 12.121 0.457
Median 227,788.4 6.632 9.883 0.393
Standard deviation 200,586.2 8.68 5.07 0.23
a
Some apartments have multiple stories at the same locations, which makes the number of air pollution data less than the number of
housing price data.
Int. J. Environ. Res. Public Health 2021,18, 12195 6 of 13
3.2. Spatial Autocorrelation
We used the Moran scatter plot to examine the spatial autocorrelation of housing
prices and three air pollutants within the study domain (Figure 4). For the Moran scatter
plot, it shows the relation between the spatially lagged variable and the original variable,
which suggests how housing prices and air pollutant concentrations are related to their
surrounding neighbors. The Moran’s I test is commonly used in geography-related fields
to quantify spatial autocorrelation. For the Moran’s I test, the test statistic is represented by
the slope of the fitted line in the Moran scatter plot (Figure 4), which measures how one
object is similar to its surroundings. We also used the permutation-based random Moran’s
I test, which uses the Monte-Carlo simulation method to randomly shuffle the data and
calculate the Moran’s I statistic for each random shuffle and compare it with the actual
Moran’s I statistic. The results of both Moran’s I tests for housing prices and the three
pollutants are shown in Table 2. The housing price has a Moran’s I value equal to 0.276,
suggesting a positive spatial autocorrelation. All of the pollutants have Moran’s I values
close to 0.99, suggesting highly positive autocorrelation. Therefore, including the spatial
autocorrelation term into the hedonic price model is necessary.
Variables
NO Concentration
NO2 Concentration
BC Concentration
Intercept
2.9196 ***
(0.4688)
2.5027 ***
(0.45954)
2.8232 ***
(0.46502)
Year Built
0.0070745 ***
(0.00033103)
0.0068693 ***
(0.00032984)
0.0070433 ***
(0.0003305)
Figure 4.
Moran’s I scatter plots of housing prices (
A
), NO (
B
), NO
2
(
C
), and BC (
D
) concentrations (blue lines are the linear
regression lines between variables and the lagged variables; the slopes of blue lines are the Moral’s I statistic).
Int. J. Environ. Res. Public Health 2021,18, 12195 7 of 13
Table 2. Moran’s I test results for housing prices and three pollutants.
Housing Price NO Concentration NO2Concentration BC Concentration
Moran’s I test statistic 0.27643 0.98498 0.9927 0.99127
Analytical method p-value <0.001 <0.001 <0.001 <0.001
Monte-Carlo-based p-value <0.001 <0.001 <0.001 <0.001
3.3. Spatial Lag Model Results
The model results of all three pollutants were very similar (Table 3). As expected,
the year the home was built negatively influences the housing price, and garage, bath
number, total area, and median income positively influence the housing price. Air pollution
concentrations positively and significantly influence housing prices, which we will discuss
in greater detail in Section 3.4.
3.4. Discussion
Based on the coefficients from the SLM model in Table 3, they suggest that all three
pollutants (NO, NO
2
, and BC) have a positive and significant effect on housing prices. This
is unexpected and we have a few speculations as to why this occurs. First, the air pollution
concentrations are low throughout the area. The average concentration of NO is 10.29 ppb,
NO
2
is 12.12 ppb, and BC is 0.46
µ
g/m
3
. We also included the air pollution concentrations
of BC and PM
2.5
from a stationary monitoring station (Oakland-West site) located in the
center of
West Oakland (https://ww3.arb.ca.gov/qaweb/iframe_site.php?s_arb_code = 60349
,
accessed on 15 September 2020). For the stationary data, we calculated the mean values of the
hourly measurements between June 2015 and May 2016, which covers the same time range
(9 am to 5 pm) of the mobile air pollution measurement in our study. The mean concentra-
tions of BC and PM
2.5
from the stationary monitor are 0.59 and 8.36
µ
g/m
3
, respectively.
The BC concentrations are close between the stationary monitor measurement and the
mobile measurement we used in this study, which provides a general estimate about the
PM
2.5
concentrations across our study domain. Comparing these to the National Ambient
Air Quality Standards (NAAQS), the annual standard of NO
2
is at a level of
53 ppb
, and
the annual standard of PM
2.5
is 12.0
µ
g/m
3
for a primary source and
15.0 µg/m3
for a
secondary source [
37
]. It is possible that when the ambient air quality is relatively clean,
affordability dominates the need to pay a housing premium for even cleaner air. Of the
19 papers we found on housing and air quality, 10 papers were relevant to our research.
Among these, the findings are mixed (Table 4). Three show insignificant effects of air
pollution on housing price. In the remaining seven papers, the air pollution concentrations
have significant and negative effect on house prices.
All 10 papers we found used the hedonic price model to study the impact of air pollu-
tion on housing prices. Their conclusions were derived from the model coefficients. If the
regression coefficient of air pollution is statistically significantly less than zero, air pollution
negatively influences housing prices, and if the coefficient is significantly greater than
zero, air pollution positively influences housing prices. If the coefficient is not significantly
different from zero, air pollution is not significantly influencing housing prices.
As we noted in the introduction, even though all of the papers used the hedonic price
model, the authors relied on different methods to emphasize different effects (e.g., instrumen-
tal variable (IV), spatial lag model (SLM), spatial error model (SEM), fixed effect, etc.).
In Table 4, among the three studies with insignificant results about air pollution
influencing housing prices, they all take the spatial autocorrelation effect into consideration
when constructing the hedonic price model. In one study, the authors argue that the
insignificant effect of NO
x
concentrations on housing prices is due to the fact that NO
x
does not tend to exceed the standard; on the contrary, SO
2
shows a significant and negative
impact on housing prices in the same study, because SO
2
has exceeded the official air
quality standard over a long period of time [
16
]. The other two studies believe that the
insignificant results are caused by either an insufficient degree of efficiency [
11
] or that the
Int. J. Environ. Res. Public Health 2021,18, 12195 8 of 13
change of air pollution concentration is more important than air pollution concentration
itself [12].
Table 3. Results of models with different pollutants a.
Variables NO Concentration NO2Concentration BC Concentration
Intercept 2.9196 ***
(0.4688)
2.5027 ***
(0.45954)
2.8232 ***
(0.46502)
Year Built 0.0070745 ***
(0.00033103)
0.0068693 ***
(0.00032984)
0.0070433 ***
(0.0003305)
Effective Year Built 0.010137 ***
(0.0003401)
0.010268 ***
(0.00033955)
0.010166 ***
(0.00033986)
Construction type: concrete 0.014669
(0.061875)
0.0076211
(0.061662)
0.0041038
(0.061783)
Construction type: frame 0.3531 ***
(0.020988)
0.32364 ***
(0.021187)
0.34251 ***
(0.021)
Construction type: masonry 0.36805 ***
(0.075532)
0.32321 ***
(0.074869)
0.35448 ***
(0.075232)
Other rooms:
gym
0.10416 **
(0.043856)
0.080572 *
(0.04356)
0.092731 **
(0.043693)
Other rooms:
office
0.17428
(0.40627)
0.18374
(0.4051)
0. 18428
(0.40593)
Parking type:
Carport
0.027576
(0.029369)
0.016681
(0.029342)
0.020942
(0.029382)
Parking type:
garage
0.051695 ***
(0.010082)
0.061715 ***
(0.010229)
0.055929 ***
(0.010152)
Parking type:
Mixed
0.0064772
(0.041319)
0.0046338
(0.041243)
0.00011624
(0.04131)
Stories 0.020611 ***
(0.0021083)
0.017629 ***
(0.0021295)
0.020372 ***
(0.0021063)
Rooms 0.0095808 **
(0.0040749)
0.0096307 **
(0.0060424)
0.094721 **
(0.0040706)
Beds 0.010024
(0.0064244)
0.0092185
(0.0064047)
0.010209
(0.0064193)
Baths 0.084969 ***
(0.0086318)
0.082202 ***
(0.0086111)
0.08463 ***
(0.0086243)
Total area 0.00027102 ***
(0.000014127)
0.00026972 ***
(0.000014055)
0.0002711 ***
(0.000014107)
Population density 0.000014708 ***2.7835 ×106)0.00017308 ***
2.8254 ×106)
0.000018455 ***
2.9621 ×106)
Median income 5.0184 ×106***
3.0363 ×107)
5.244 ×106***
2.9925 ×107)
5.3215 ×106***
2.9968 ×107)
Non-employment rate 0.07601
(0.050197)
0.031651
(0.050804)
0.081003
(0.04948)
NO concentration 0.0054361 ***
(0.00082701) - -
NO2concentration - 0.013246 ***
(0.0016209) -
BC concentration - - 0.22871 ***
(0.03212)
lambda 0.21710 ***
(0.019776)
0.18774 ***
(0.020526)
0.20761 ***
(0.020015)
R20.3183 0.3175 0.3178
a*** significant at less than 0.1%, ** significant at less than 5%, * significant at 10%. (): standard error.
Int. J. Environ. Res. Public Health 2021,18, 12195 9 of 13
Table 4. Literature review summary.
Location Air Pollution Concentrations
Method
Air Pollution
Impact on
Housing Price
CO, µg/m3NO2,µg/m3O3,µg/m3PM2.5,µg/m3PM10 ,µg/m3SO2,µg/m3TSP, µg/m3BC, µg/m3NO, µg/m3
Seoul, Korea
(Kim & Yoon, 2019) 45.611 SDM insignificant
Seoul, Korea (C.W. Kim,
Phipps, & Anselin, 2003)
45.57 a
SLM, SEM insignificant
82.95 negative
18 districts in Warsaw,
Poland (Ligus &
Peternek, 2017)
__ __ Linear, Logarithm,
SLM, SEM insignificant b
Beijing, China
(Mei, et al. 2020)
1399.1
Fixed effect
negative
60.34 negative
53.66 positive
88.24 negative
111.27 negative
20.5 negative
286 prefectural cities in China
(Chen & Jin, 2019) 64.81 IV negative
288 Chinese cities
(Huang & Lanz, 2018) 77.44
IV and
discontinuity
regression
negative
3 largest cities in Mexico
(Gonzalez, Leipnik &
Mazumder, 2013)
38.5, 51.7, 84 IV negative
Metro areas US
(Bayer et al., 2009)
42.21 (1990),
33.87 (2000) IV negative
All counties in USA
(Chay & Greenstone, 2005)
64.1 (1970),
56.3 (1980)
quasi-
experimental
discontinuity
regression
negative
Lebanon (Marrouch &
Sayour, 2021) 27.67 Fixed effect negative
Oakland, CA, USA 22.79 0.457 12.86 IV and SLM positive
aPaper reported NOx concentration in ppb and we converted it to µg/m3with NO2molecular weight. bInsignificant in most districts, some districts are positive or negative.
Int. J. Environ. Res. Public Health 2021,18, 12195 10 of 13
In examining the literature, the results are suggestive that air pollution’s effects tend
to be insignificant when overall ambient air pollution concentrations are relatively low. In
our study, the average air pollution concentrations across all of our sample observations
were the lowest among these studies. It is possible that affordability is more important
than a housing premium for even cleaner air when the ambient air quality is already good.
Therefore, the positive and significant coefficients of air pollution on housing prices may
be reasonable in areas with good air quality.
A second possible reason why we find counterintuitive results may be due to the very
low variability in pollutants and housing prices. Within our buffer, standard deviations of
NO, NO
2
, and BC concentrations were 8.68 ppb, 5.07 ppb, and 0.23
µ
g/m
3
, respectively.
We compared the distribution of the three pollutants in our study with one stationary
monitoring measurement located in the center of West Oakland (WO) in Figure 5. For
the stationary data, we used the hourly measurements of NO, NO
2
, and BC from the
above-mentioned Oakland-west site, covering the same date and time range of the mobile
air pollution measurement in our study. Our data variability is close to the variation of
air pollution concentrations at a single location. Low variability may lead to positive and
significant coefficients even if the results are not significant.
Int. J. Environ. Res. Public Health 2021, 18, 12195 11 of 14
A second possible reason why we find counterintuitive results may be due to the
very low variability in pollutants and housing prices. Within our buffer, standard devia-
tions of NO, NO2, and BC concentrations were 8.68 ppb, 5.07 ppb, and 0.23 µg/m3, respec-
tively. We compared the distribution of the three pollutants in our study with one station-
ary monitoring measurement located in the center of West Oakland (WO) in Figure 5. For
the stationary data, we used the hourly measurements of NO, NO2, and BC from the
above-mentioned Oakland-west site, covering the same date and time range of the mobile
air pollution measurement in our study. Our data variability is close to the variation of air
pollution concentrations at a single location. Low variability may lead to positive and sig-
nificant coefficients even if the results are not significant.
Figure 5. Pollutant distributions comparison between our work and one stationary monitor (NO and NO2 are in the unit
of ppb, BC is in the unit of µg/m3).
4. Limitations and Conclusions
To the best of our knowledge, this is the first study that combines a spatial lag model
with an instrumental variable method to capture both the spatial autocorrelation and en-
dogeneity effects of the relationship between housing prices and air pollution. Our study
Figure 5.
Pollutant distributions’ comparison between our work and one stationary monitor (NO and NO
2
are in the unit of
ppb, BC is in the unit of µg/m3).
Int. J. Environ. Res. Public Health 2021,18, 12195 11 of 13
4. Limitations and Conclusions
To the best of our knowledge, this is the first study that combines a spatial lag model
with an instrumental variable method to capture both the spatial autocorrelation and
endogeneity effects of the relationship between housing prices and air pollution. Our
study demonstrated the use of high-resolution air pollution mapping data to quantify the
localized ambient air quality at the parcel level, which can largely improve the accuracy
of pollution cost valuation. The results of our work have significant policy implications,
as the air quality regulations require accurate understanding of the potential pollution
cost. Furthermore, our work is helpful and instructive in environmental justice studies,
especially in the local scale. It helps to identify the vulnerable populations in the complex
urban environment.
Our results are counterintuitive, suggesting that air pollution positively influences
housing prices in Oakland, CA. We believe that this counterintuitive result arises from
two possible explanations. First, the results suggest that people may be insensitive to air
quality if the overall ambient air quality is good, which is consistent with other literature
we reviewed. Second, our study focused on a relatively small study domain, where the
variability of air pollution concentrations and housing prices is low. The low variability of
variables may lead to significant result even though the true influence is not significant.
Our results indicate that a larger, multi-regional study is probably the best way to
determine the relationship between air pollution and housing prices. This study was
conducted based on the 2016 data. Recently, data from high-resolution air pollution mea-
surements have been expanding quickly. The Google Earth Outreach team has conducted
the high-resolution air pollution measurement in Houston, London, Copenhagen, and
Amsterdam. These data may prove useful to better understand how air pollution affects
housing prices. The framework and method of this paper can be applied to multi-regions
when the coverage of the high-resolution air pollution measurement is widely expanded.
With more diversified regions covered in the analysis, we will be able to more accurately
and precisely understand how air pollution concentrations influence housing prices.
Author Contributions:
Data collection, M.T.; analysis and interpretation of results, M.T. and D.N.;
draft manuscript preparation, M.T. and D.N. All authors have read and agreed to the published
version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement:
Restrictions apply to the availability of housing price data. Housing
price data were obtained from Estated, Inc. (https://estated.com/ accessed on 13 August 2020) and
are available from the corresponding author with the permission of Estated, Inc. Air pollution data
are publicly available in the supplementary material of the paper at https://pubs.acs.org/doi/10.102
1/acs.est.7b00891 accessed on 13 August 2020. Social demographic data are from the 2016 American
Community Survey, which is publicly available at https://www.census.gov/programs-surveys/acs
accessed on 13 August 2020.
Conflicts of Interest: There are no conflicts of interest of this research.
References
1.
Apte, J.S.; Marshall, J.D.; Cohen, A.J.; Brauer, M. Addressing Global Mortality from Ambient PM2.5. Environ. Sci. Technol.
2015,49, 8057–8066. [CrossRef] [PubMed]
2.
Lim, S.S.; Vos, T.; Flaxman, A.D.; Danaei, G.; Shibuya, K.; Adair-Rohani, H.; AlMazroa, M.A.; Amann, M.; Anderson, H.R.;
Andrews, K.G.; et al. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor
clusters in 21 regions, 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet
2012
,380, 2224–2260.
[CrossRef]
3.
Robinson, E. How Much Does Air Pollution Cost the U.S.? Available online: https://earth.stanford.edu/news/how-much-does-
air-pollution-cost-us#gs.6njexm (accessed on 30 August 2020).
4.
Tschofen, P.; Azevedo, I.L.; Muller, N.Z. Fine particulate matter damages and value added in the US economy. Proc. Natl. Acad.
Sci. USA 2019,116, 19857–19862. [CrossRef]
5.
Rosen, S. Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. J. Political Econ.
1974
,82, 34–55.
[CrossRef]
Int. J. Environ. Res. Public Health 2021,18, 12195 12 of 13
6. Ekeland, I.; Heckman, J.J.; Nesheim, L. Identifying Hedonic Models. Am. Econ. Rev. 2002,92, 304–309. [CrossRef]
7.
Raya, J.M.; Estévez, P.G.; Prado-Román, C.; Pruñonosa, J.T. Living in a Smart City Affects the Value of a Dwelling? In
BT—Sustainable Smart Cities: Creating Spaces for Technological, Social and Business Development; Peris-Ortiz, M., Bennett, D.R., Pérez-
Bustamante Yábar, D., Eds.; Springer International Publishing: Cham, Switzerland, 2017; pp. 193–198, ISBN 978-3-319-40895-8.
8.
Torres-Pruñonosa, J.; García-Estévez, P.; Prado-Román, C. Artificial Neural Network, Quantile and Semi-Log Regression
Modelling of Mass Appraisal in Housing. Mathematics 2021,9, 783. [CrossRef]
9.
Betz, T.; Cook, S.J.; Hollenbach, F.M. Spatial interdependence and instrumental variable models. Political Sci. Res. Methods
2020,8, 646–661. [CrossRef]
10.
Gonzalez, F.; Leipnik, M.; Mazumder, D. How much are urban residents in Mexico willing to pay for cleaner air?
Environ. Dev. Econ. 2013,18, 354–379. [CrossRef]
11.
Ligus, M.; Peternek, P. Impacts of Urban Environmental Attributes on Residential Housing Prices in Warsaw (Poland): Spatial
Hedonic Analysis of City Districts. In BT—Contemporary Trends and Challenges in Finance; Jajuga, K., Orlowski, L.T., Staehr, K.,
Eds.; Springer International Publishing: Cham, Switzerland, 2017; pp. 155–164.
12. Kim, S.G.; Yoon, S. Measuring the value of airborne particulate matter reduction in Seoul. Air Qual. Atmos. Health 2019,12, 549–560.
[CrossRef]
13.
Montero, J.-M.; Mínguez, R.; Fernández-Avilés, G. Housing price prediction: Parametric versus semi-parametric spatial hedonic
models. J. Geogr. Syst. 2018,20, 27–55. [CrossRef]
14.
De, U.K.; Vupru, V. Location and neighbourhood conditions for housing choice and its rental value. Int. J. Hous. Mark. Anal.
2017,10, 519–538. [CrossRef]
15.
Sun, B.; Yang, S. Asymmetric and Spatial Non-Stationary Effects of Particulate Air Pollution on Urban Housing Prices in Chinese
Cities. Int. J. Environ. Res. Public Health 2020,17, 7443. [CrossRef] [PubMed]
16.
Kim, C.W.; Phipps, T.T.; Anselin, L. Measuring the benefits of air quality improvement: A spatial hedonic approach. J. Environ.
Econ. Manag. 2003,45, 24–39. [CrossRef]
17. Chen, S.; Jin, H. Pricing for the clean air: Evidence from Chinese housing market. J. Clean. Prod. 2019,206, 297–306. [CrossRef]
18.
Bayer, P.; Keohane, N.; Timmins, C. Migration and hedonic valuation: The case of air quality. J. Environ. Econ. Manag.
2009,58, 1–14. [CrossRef]
19.
Mei, Y.; Gao, L.; Zhang, J.; Wang, J. Valuing urban air quality: A hedonic price analysis in Beijing, China. Environ. Sci. Pollut. Res.
2020,27, 1373–1385. [CrossRef]
20.
Chay, K.Y.; Greenstone, M. Does Air Quality Matter? Evidence from the Housing Market. J. Political Econ.
2005
,113, 376–424.
[CrossRef]
21. Moran, P.A.P. Notes on Continuous Stochastic Phenomena. Biometrika 1950,37, 17–23. [CrossRef] [PubMed]
22. Liu, R.; Yu, C.; Liu, C.; Jiang, J.; Xu, J. Impacts of Haze on Housing Prices: An Empirical Analysis Based on Data from Chengdu
(China). Int. J. Environ. Res. Public Health 2018,15, 1161. [CrossRef] [PubMed]
23.
Bae, C.-H.C.; Sandlin, G.; Bassok, A.; Kim, S. The exposure of disadvantaged populations in freeway air-pollution sheds: A case
study of the Seattle and Portland regions. Environ. Plan. B Plan. Des. 2007,34, 154–170. [CrossRef]
24.
Point2. Central East Demographics. Available online: https://www.point2homes.com/US/Neighborhood/CA/Oakland/
Central-East-Demographics.html (accessed on 25 July 2021).
25.
Apte, J.S.; Messier, K.P.; Gani, S.; Brauer, M.; Kirchstetter, T.W.; Lunden, M.M.; Marshall, J.D.; Portier, C.J.; Vermeulen, R.C.H.;
Hamburg, S.P. High-Resolution Air Pollution Mapping with Google Street View Cars: Exploiting Big Data. Environ. Sci. Technol.
2017,51, 6999–7008. [CrossRef] [PubMed]
26.
Google. Oakland_201506-201605_GoogleAclimaAQ. Available online: https://www.edf.org/airqualitymaps/oakland/
methodology-how-we-collected- data (accessed on 1 November 2017).
27.
Karner, A.A.; Eisinger, D.S.; Niemeier, D.A. Near-Roadway Air Quality: Synthesizing the Findings from Real-World Data.
Environ. Sci. Technol. 2010,44, 5334–5344. [CrossRef] [PubMed]
28.
Zhou, Y.; Levy, J.I. Factors influencing the spatial extent of mobile source air pollution impacts: A meta-analysis.
BMC Public Health
2007,7, 89. [CrossRef]
29.
Arraiz, I.; Drukker, D.M.; Kelejian, H.H.; Prucha, I.R. A Spatial Cliff-Ord-Type Model with Heteroskedastic Innovations: Small
and Large Sample Results. J. Reg. Sci. 2009,50, 592–614. [CrossRef]
30.
Drukker, D.M.; Egger, P.; Prucha, I.R. On Two-Step Estimation of a Spatial Autoregressive Model with Autoregressive Disturbances
and Endogenous Regressors. Econ. Rev. 2013,32, 686–733. [CrossRef]
31.
Kelejian, H.H.; Prucha, I.R. Specification and estimation of spatial autoregressive models with autoregressive and heteroskedastic
disturbances. J. Econ. 2010,157, 53–67. [CrossRef] [PubMed]
32.
Kelejian, H.H.; Prucha, I.R. A Generalized Moments Estimator for the Autoregressive Parameter in a Spatial Model.
Int. Econ. Rev.
1999,40, 509–533. [CrossRef]
33.
Kelejian, H.H.; Prucha, I.R. A Generalized Spatial Two-Stage Least Squares Procedure for Estimating a Spatial Autoregressive
Model with Autoregressive Disturbances. J. Real Estate Financ. Econ. 1998,17, 99–121. [CrossRef]
34. R Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2021.
35. Piras, G. sphet: Spatial Models with Heteroskedastic Innovations in R. J. Stat. Softw. 2010,35, 1–21. [CrossRef]
Int. J. Environ. Res. Public Health 2021,18, 12195 13 of 13
36.
Bivand, R.; Piras, G. Comparing Implementations of Estimation Methods for Spatial Econometrics. J. Stat. Softw.
2015
,63, 1–36.
[CrossRef]
37. EPA. NAAQS Table. Available online: https://www.epa.gov/criteria-air-pollutants/naaqs-table (accessed on 2 August 2021).
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
We used a large sample of 188,652 properties, which represented 4.88% of the total housing stock in Catalonia from 1994 to 2013, to make a comparison between different real estate valuation methods based on artificial neural networks (ANNs), quantile regressions (QRs) and semi-log regressions (SLRs). A literature gap in regard to the comparison between ANN and QR modelling of hedonic prices in housing was identified, with this article being the first paper to include this comparison. Therefore, this study aimed to answer (1) whether QR valuation modelling of hedonic prices in the housing market is an alternative to ANNs, (2) whether it is confirmed that ANNs produce better results than SLRs when assessing housing in Catalonia, and (3) which of the three mass appraisal models should be used by Spanish banks to assess real estate. The results suggested that the ANNs and SLRs obtained similar and better performances than the QRs and that the SLRs performed better when the datasets were smaller. Therefore, (1) QRs were not found to be an alternative to ANNs, (2) it could not be confirmed whether ANNs performed better than SLRs when assessing properties in Catalonia and (3) whereas small and medium banks should use SLRs, large banks should use either SLRs or ANNs in real estate mass appraisal.
Article
Full-text available
Fine particulate matter(PM2.5) pollution will affect people’s well-being and cause economic losses. It is of great value to study the impact of PM2.5 on the real estate market. While previous studies have examined the effects of PM2.5 pollution on urban housing prices, there has been little in-depth research on these effects, which are spatially heterogeneous at different conditional quantiles. To address this issue, this study employs quantile regression (QR) and geographically weighted quantile regression (GWQR) models to obtain a full account of asymmetric and spatial non-stationary effects of PM2.5 pollution on urban housing prices through 286 Chinese prefecture-level cities for 2005–2013. Considerable differences in the data distributions and spatial characteristics of PM2.5 pollution and urban housing prices are found, indicating the presence of asymmetric and spatial non-stationary effects. The quantile regression results show that the negative influences of PM2.5 pollution on urban housing prices are stronger at higher quantiles and become more pronounced with time. Furthermore, the spatial relationship between PM2.5 pollution and urban housing prices is spatial non-stationary at most quantiles for the study period. A negative correlation gradually dominates in most of the study areas. At higher quantiles, PM2.5 pollution is always negatively correlated with urban housing prices in eastern coastal areas and is stable over time. Based on these findings, we call for more targeted approaches to regional real estate development and environmental protection policies.
Article
Full-text available
Although Chinese government has implemented a series of laws, regulations, and policies to deal with air pollutants, air pollution is still one of the biggest concerns in China. Most of the cities in China suffered from extremely high air pollution levels and cannot meet the national air quality standards. In this study, we attempt to measure individuals’ average marginal willingness to pay (MWTP) in Beijing with the hedonic price model. We combine an extensive housing transaction dataset with emission data of six air pollutants from 2013 to 2016 in Beijing, China. When estimating the hedonic price function, we apply both ordinary least squares (OLS) and panel model with various fixed effects to better control for unobservables. The empirical results reveal that the concentrations of CO, NO2, PM2.5, and PM10 are significantly negatively correlated with housing prices. However, we found an insignificant relationship with the concentration of SO2 and the concentration of O3 appears to positively increase the housing values. Policy implications based on these results were also discussed.
Article
Full-text available
Airborne suspended particles, also generally referred to as fine dust, are one of Korea’s most serious environmental and economic concerns. This paper estimates the perceived cost of damage from fine dust in the housing market in Seoul. To do this, we collected data on approximately 120,000 apartment transactions in 2016 and created variables explaining the apartment price, including the concentration of fine dust at the district level. Our results indicate that price premiums exist for apartments in areas with fine dust concentration decrease. This finding reflects the willingness to pay for reductions in fine dust levels, which can be viewed as the monetary value of the benefit with respect to a cost-benefit analysis for fine dust reduction policies. This finding might also provide valuable information when making decisions on the degree of effort that our society must make, and the resources needed, to mitigate this problem.
Article
Full-text available
Instrumental variable (IV) methods are widely used to address endogeneity concerns. Yet, a specific kind of endogeneity – spatial interdependence – is regularly ignored. We show that ignoring spatial interdependence in the outcome results in asymptotically biased estimates even when instruments are randomly assigned. The extent of this bias increases when the instrument is also spatially clustered, as is the case for many widely used instruments: rainfall, natural disasters, economic shocks, and regionally- or globally-weighted averages. Because the biases due to spatial interdependence and predictor endogeneity can offset, addressing only one can increase the bias relative to ordinary least squares. We demonstrate the extent of these biases both analytically and via Monte Carlo simulation. Finally, we discuss a general estimation strategy – S-2SLS – that accounts for both outcome interdependence and predictor endogeneity, thereby recovering consistent estimates of predictor effects.
Article
Full-text available
Based on cross-section data of 20 districts in Chengdu, this article reviews the relationships between haze and housing prices with the combined application of Spatial Error Model (SEM) and Spatial Lag Model (SLM). The results illustrate that haze significantly have negative impacts on both the selling and rental prices of houses. Controlling other variables, if the air quality index rises by 0.1, the housing selling prices and rental prices will drop by 3.97% and 4.01%, respectively. Interestingly, housing rental prices have a more significant response to the air quality than housing sale prices. Residents are willing to pay a premium for better air quality and the influence of air quality is partially reflected in housing prices, which indicates that better air quality has been becoming a scarce resource with the improvement of people's living standard. Furthermore, the impacts of haze on housing prices are also expected to lead to a "crowding out effect" in different regions. This would be detrimental for human capital accumulation and will accelerate the regional divergence in the internal economy and population structure, thus forming a region "fence" within cities.
Article
Full-text available
House price prediction is a hot topic in the economic literature. House price prediction has traditionally been approached using a-spatial linear (or intrinsically linear) hedonic models. It has been shown, however, that spatial effects are inherent in house pricing. This article considers parametric and semi-parametric spatial hedonic model variants that account for spatial autocorrelation, spatial heterogeneity and (smooth and nonparametrically specified) nonlinearities using penalized splines methodology. The models are represented as a mixed model that allow for the estimation of the smoothing parameters along with the other parameters of the model. To assess the out-of-sample performance of the models, the paper uses a database containing the price and characteristics of 10,512 homes in Madrid, Spain (Q1 2010). The results obtained suggest that the nonlinear models accounting for spatial heterogeneity and flexible nonlinear relationships between some of the individual or areal characteristics of the houses and their prices are the best strategies for house price prediction.
Article
Full-text available
Purpose The purpose of the study is to understand the relative impacts of structural characteristics of house, its locational attributes and neighbourhood characteristics on the choice of house and the rent paid by the individual tenants. The micro level study helps in understanding the issues of urban housing and help in policy formulation. Design/methodology/approach This paper tried to identify the socio-economic, locational and neighbourhood factors that influence tenant households in determining their residential choices in an urban area of North-east India. Also, the extent of impacts of those characteristics on the monthly rent for housing is analysed. The analysis is based on the primary data collected from the sample residents of Dimapur Town in Nagaland. The sample units are selected by cluster sampling technique from all the wards. Regression technique is used under hedonic pricing technique to examine the impacts of various potential factors on the rent. Findings The analysis reveals that family size, income, education of the head of family, water availability, security, convenience to access workplace, road conditions, etc. have significant positive impacts on the monthly rent. However, the impacts of some locational and neighbourhood variables vary across social and economic groups. Research limitations/implications Time series data on the growth of house price are not available, so that the authors could examine the escalation of house price and rising scarcity of houses in the selected town. Practical implications Rental housing accounts for a significant percentage of housing in many urban centres. It is particularly important for the migrants and urban poor for whom it is the only source of accommodation. Location choice depending on capability, availability, requirement and neighbourhood conditions are integral parts of selection of accommodation in a city. Hence the present study has its relevant practical implications. Social implications Housing choice in a hilly urban area like Dimapur assumes a special significance for the presence of heterogeneous ethnic, social and cultural groups with majority of Tribal inhabitants with varied requirements and lifestyle. Originality/value No study on housing choice or pricing of residential accommodation has been done in North-East India. Here along with the socio-economic characteristics of the individuals, structural characteristics of house, neighbourhood and locational characteristics are used simultaneously to find out the impact of various factors on the total prices through hedonic pricing method.
Article
Emissions of most pollutants that result in fine particulate matter (PM 2.5 ) formation have been decreasing in the United States. However, this trend has not been uniform across all sectors or regions of the economy. We use integrated assessment models (IAMs) to compute marginal damages for PM 2.5 -related emissions for each county in the contiguous United States and match location-specific emissions with these marginal damages to compute economy-wide gross external damage (GED) due to premature mortality. We note 4 key findings: First, economy-wide, GED has decreased by more than 20% from 2008 to 2014. Second, while much of the air pollution policies have focused to date on the electricity sector, damages from farms are now larger than those from utilities. Indeed, farms have become the largest contributor to air pollution damages from PM 2.5 -related emissions. Third, 4 sectors, comprising less than 20% of the national gross domestic product (GDP), are responsible for ∼75% of GED attributable to economic activities. Fourth, uncertainty in GED estimates tends to be high for sectors with predominantly ground-level emissions because these emissions are usually estimated and not measured. These findings suggest that policymakers should target further emissions reductions from such sectors, particularly in transportation and agriculture.
Article
Air pollution is widely believed to cause healthy risks and influence nations’ economic dimensions. This paper investigates the impact of air pollution on housing prices in China's 286 prefectural-level cities during 2005–2013. To address the endogeneity problem of PM2.5 concentrations in our regression, we propose an instrument variable strategy based on ventilation coefficient. The estimation results suggest that a 10% increase in PM2.5 concentrations causes a 2.4% reduction in local housing prices. Air pollution also hinders the progress of urbanization, limits cities’ human capital formation and changes people's expectations of housing prices that we interpret as three potential mechanisms behind our main findings.