Embed Size (px)
Citation preview
1
Earthquake alley: Unconventional oil and gas development, induced seismic activity, and housing price impacts
J. Wesley Burnett,† Christopher Mothorpe,‡ and Steven C. Jaume´*
Tuesday, February 27, 2018
Abstract
This study seeks to identify the economic impacts associated with earthquakes caused by unconventional oil and gas production and wastewater injection activities. Our identification strategy consists of linking geographically proximate housing values to induced earthquake activities within Oklahoma County, Oklahoma. Using both seismographic and self-reported measures, we provide robust evidence that seismic events are negatively impacting property values, which are located in relatively close proximity to an earthquake's epicenter. We argue that induced earthquakes are endogenous to housing values due to the predetermined siting of injection wells into select regions within the metropolitan statistical area. Based on these observations, we conclude any (induced) earthquake impact analysis, which fails to recognize this endogeneity, may yield potentially inconsistent effects estimates. Using an instrumental variables approach, our findings suggest that the cumulative effects of these earthquakes generate, on average, a 2.4 percent reduction in housing values. Based on the reduction in prices, we estimate the total per-household costs, associated with induced earthquakes in Oklahoma County, are approximately $3,628.
Keywords: Induced earthquakes; Natural gas development; Residential housing prices; Wastewater disposal JEL Codes: Q40, Q51, R30 † Department of Economics, 5 Liberty Street, 413 Beatty Center, College of Charleston, Charleston, SC 29401,
843.953.0752, [email protected]. ‡ Department of Economics, 5 Liberty Street, 412 Beatty Center, College of Charleston, Charleston, SC 29401,
843.953.7273, [email protected]. * Department of Geology and Environmental Geosciences, 202 Calhoun Street, 210 School of Science and
Mathematics Building, College of Charleston, Charleston, SC 29401, 843.953.1802, [email protected].
2
1 INTRODUCTION
Unconventional oil and gas development skyrocketed in the state of Oklahoma starting around
the year 2007. Unlike previous plays in the State, unconventional development generates large
volumes of produced wastewater, which is a byproduct of directional drilling and hydraulic
fracturing. First noted in the late 1960's, geologists and seismologists identified a link between
earthquake activities and disposal well injections, which are largely associated with produced
wastewater. The injections arguably decrease the frictional stresses that hold geologic faults
together deep below the earth's surface. Put differently, the wastewater injections can cause a
slippage along a fault and unleash an earthquake (Kuchment, 2016). Keranen et al. (2014) assert
that fluid migration from high-rate disposal wells is potentially responsible for the largest swarm
of seismic activity within the state of Oklahoma. The link between increasing seismicity and
human activities created a category of earthquakes called "induced earthquakes." The
underground focal points of these induced earthquakes have been geographically linked with
disposal formations as far away as 35 kilometers (approximately 22 miles). If the monitoring of
wastewater injections is left unchecked, then increasing seismic activity may pose a significant
future economic risk to the State due to property damage and potential harm to human life.
There are three basic mechanisms by which earthquakes may affect housing prices
(Koster and van Ommeren, 2015; Cheung et al., 2016). One, earthquakes, if strong enough in
magnitude, can cause property damage. For example, in 2011, a 5.7-magnitude earthquake near
Prague, OK destroyed more than a dozen homes and caused nearly $1 million in damages
(Gallucci, April 22, 2015). Two, if the frequency of seismic activity increases, then it potentially
changes the expectations about future earthquake damages. To wit, Liu et al. (2016) estimated
3
that properties sold after the 2011 Prague earthquake and located within two kilometers (km) of
an injection well experienced a 2.2% (or approximately $4.4 thousand) loss to property values.
Three, the resulting earthquakes do not cause monetary damages to properties, but the increase in
small, but disruptive magnitude seismic events make for an unpleasant environment to live in
(Cheung et al., 2016).
The current study seeks to identify the economic impacts associated with seismic activity,
unconventional oil and gas production, and wastewater injections. Our identification strategy
consists of linking geographically proximate housing values to nearby earthquakes in the region
surrounding Oklahoma City - one of the State's hotspots of seismic swarms. Using self-reported
and seismographic measures, we show that induced earthquakes are noticeable to homeowners
who reside relatively close to the epicenters. Further, we identify two potential forms of
estimation bias within this type of study: attenuation and endogeneity bias.
As in Koster and van Ommeren (2015), we address the issue that the induced earthquakes
do not necessarily occur randomly over space. Rather, we demonstrate that the recent increase in
earthquake activities across the Oklahoma City Metropolitan Statistical Area (MSA) has a
tendency to cluster in specific areas. In words, we assume that these human-induced earthquakes
are endogenous to housing prices. Intuitively, this endogeneity problem could stem from
numerous potential causes, including (but not limited to): injection wells were located in poorer
neighborhoods; and, wealthier neighborhoods having the resources to orchestrate collective
action against wastewater injections (for example, “Not In My Backyard” campaign efforts).
To motivate the argument of injections wells being located nearer to lower-priced
housing areas, we ran a series of auxiliary regressions (of housing values on earthquake events
while controlling for other standard covariates found within the hedonic literature) by dividing
4
our sample into quartiles by home sale prices. Our crude initial estimates (results offered in the
Appendix) imply that housing values in the first quartile (equal to or less than approximately
$85,000) were affected by nearly three percentage points (-7.3 percent) more than the average (-
4.4 percent) for the entire sample. Intuitively, this may occur if disposal wells, which mostly
consist of abandoned conventional crude oil wells, were initially drilled (likely occurring 70
years ago or more) on marginal lands or the presence of heavy drilling equipment created a
visual disamenity, which made housing siting and development less appealing. (Looking ahead,
we find that induced earthquakes are generally occurring in close proximity - epicenters are
frequently within about two kilometers - to the injection wells). Our auxiliary regression results
indicate that the third and fourth quartiles, of housing values, were affected the least - the
impacts were nearly 6 percentage points less than the impacts to the first quartile - by the
increase in seismicity. The uneven impact of seismicity across quartiles is to simply illustrate
that the occurrence of induced earthquakes (across geographic space) is not random. Therefore,
unless corrected for, any impact analysis may yield inconsistent, average effects estimates.
To see this, consider the following abstract representation of the relationship between
property values and earthquake activities.
Housing prices = f[neighborhood characteristics, home characteristics, year built,
earthquakes, injection wells, and oil and gas development],
Earthquakes = f[distance to fault zone, underlying geology, injection
wells(housing prices), soil type].
5
The first equation shows that, when modeling housing prices, we may control for neighborhood
and home (or property) characteristics, as well as nearby oil (and gas) and injection wells.
Further, the first equation shows that if a household is located within a historical fault zone (i.e.,
with a known history of seismic activity), then we would expect the house to potentially sell at a
discount. (In other words, any preexisting or natural-occurring earthquake risks (prior to the
swarm in induced activities) should be capitalized into a property's current value). The discount
may be due to perceived seismic risk or an increase in financial outlays for required earthquake
insurance. The second equation shows that, when modeling earthquake activities, we may
control for distance to fault zones, the underlying geological endowments (i.e., sedimentary
basin), and different soil types. Some soil types are more prone to subsidence, which increases
the probability of earthquake damage to a property (Koster and van Ommeren, 2015). Moreover,
the second equation above shows the potential source of endogeneity between property values
and induced earthquakes. As discussed above, less expensive neighborhoods may be nearer to oil
and gas drilling or injection wells (Throupe et al., 2013). Following the onset of the induced
seismic swarms, there is also the possibility that new housing starts occurred away from the
hotspots of induced seismic activity; in which case, housing prices are now highly correlated
with drilling activities. In the presence of such endogeneity bias, one would expect an ordinary
least squares (OLS) regression, of housing prices on earthquakes, to yield upwardly biased (in
absolute terms) coefficient estimate on the impact of earthquake activities.
When analyzing the relationship between oil/gas development and impacts to housing
prices, another potential problem stems from attenuation bias, in which an earthquake's impact
measure itself potentially violates the classical errors-in-variables assumption (Wooldridge,
2012). In other words, earthquake impacts are subject to measurement error, and, unless
6
corrected for, may lead to potential attenuated effects (downwardly biased) estimates. To address
attenuation bias, we match actual seismic data to self-reported "Did you feel it?" data (U.S.
Geological Survey, 2017a). The self-reported seismic data is discussed in further detail below.
After controlling for potential endogeneity and attenuation bias, we find that any earthquakes of
magnitude 3.0 or larger, using an instrumental variables approach, generates an average 2.4
percent reduction in property values (since the swarm of seismic events started in 2007). In order
to generate this estimate, we were careful to control for numerous neighborhood and housing
characteristics including the vintage of the house. Based on the estimated reduction in prices, we
calculate the total average per-household costs, associated with induced earthquakes in
Oklahoma City, are approximately $3,628.
This study offers two major contributions. One, we provide new evidence of the
economic impacts associated with human-induced earthquakes. Recent hedonic studies have
found an indirect, negative link between oil and gas development production and housing prices
- the relationship is indirect in that oil and gas development arguably leads to induced seismic
activity, which in turn leads to a reduction in property values (Koster and van Ommeren, 2015;
Liu et al., 2016; Cheung et al., 2016). The estimated negative impacts, from induced earthquakes
to property values, found within this literature range from a 1.9 - to - 10 percent reduction in
average housing prices. However, this study provides an additional contribution by using an
instrumental variables (IV) approach due to the potential endogeneity of induced earthquakes.
We posit that the IV estimator provides a consistent and unbiased estimate of induced earthquake
impacts. Outside of the oil and gas impact analyses, the IV approach can also be used in studies
that more generally examine the economic impacts associated with natural-occurring earthquakes
and other natural disasters.
7
The rest of the manuscript is organized as follows. Section two offers background
information about oil and gas production, wastewater injections, seismic activities, and trends in
housing markets in Oklahoma City. This section also provides a brief literature review of
hedonic studies that examine the impacts associated with oil and gas production. Section three
discusses the conceptual framework for estimating the relationship between nearby earthquake
activities and the impacts to housing prices - this discussion includes an argument of why the IV
estimator is the superior approach in this particular study. Section four offers an overview of the
data, and Section five provides the estimation results, including the sensitivity analysis. Finally,
in section six we summarize the findings and discuss potential policy implications.
2 BACKGROUND
2.1 History of oil and gas production in Oklahoma City
Oklahoma has a rich history of crude oil and natural gas development. The State’s first
commercially productive crude oil well, the Nellie Johnstone No. 1, was discovered in April
1897; and, a giant oilfield was discovered in Oklahoma City in 1928 (American Oil & Gas
Historical Society, 2017). As of 2015, Oklahoma ranks fifth in the nation in crude oil production
and it is one of the top natural-gas producing states accounting for 7.6 percent of U.S. gross
production (U.S. Energy Information Administration, 2017a). Further, the State is the fifth-
largest shale-gas producing state, with substantial proven reserves. Figure 1 illustrates total
natural gas production in the state of Oklahoma from 1960 to 2014. The dark grey line represents
the total natural gas production whereas the dark link represents the shale gas production alone.
8
The left axis shows the total production while the right axis shows the shale gas production.
Based on the graph, shale production nearly makes up half of the State’s current production,
despite its very short history of development within the State. The dark line clearly shows the
meteoric rise of shale gas production in the State. As such, the industry will likely continue to
produce wastewater, which in turn, will be injected into the State’s Class II injection wells. (We
further discuss wastewater disposals in the next subsection).
Figure 1. Oklahoma’s natural gas production (million cubic feet), 1960-2014
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
800,000
900,000
1,000,000
500,000
700,000
900,000
1,100,000
1,300,000
1,500,000
1,700,000
1,900,000
2,100,000
2,300,000
2,500,000
1960
1964
1968
1972
1976
1980
1984
1988
1992
1996
2000
2004
2008
2012
Shal
e ga
s pro
duct
ion
(Mill
ion
cubi
c fe
et)
Nat
ural
gas
pro
duct
ion
(Mill
ion
cubi
c fe
et)
Year
Natural gas Shale Gas
9
Figure 2. The annual number (count) of producing natural gas and gas condensate wells: 1990-2015
Figure 2 illustrates the total number of producing natural gas wells in Oklahoma from 1990
through 2015 (U.S. Energy Information Administration, 2017b). With the exception of slight
decreases in 1999-to-2000 and 2011-to-2015, the total number of wells increased year-over-year
from 2000 to 2011, and the total number of producing wells increased by a staggering 132
percent between 2000 and 2015. Moreover, the total field production of crude oil has also
increased by 137 percent (not shown) from 2009 to 2015 (U.S. Energy Information
Administration, 2017c). That is, the State produced 66.6 million barrels of crude in 2009 versus
157.8 million barrels in 2015. The State’s large increase in natural gas and oil production, since
the latter part of the 2000s, is largely due to unconventional development – hydraulic fracturing
and directional drilling (Oklahoma Oil and Gas Association, 2017).
20,000
25,000
30,000
35,000
40,000
45,000
50,000
55,000
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011
2013
2015
Num
ber (
coun
t) of
pro
duci
ng w
ells
Year
10
Figure 3. Henry Hub Natural Gas Spot Price: 1997-2016
Figure 4. Cushing, OK Crude Oil (West Texas Intermediate) Spot Prices: 1997-2016
Lastly, Figure 3 and Figure 4 present the U.S. spot prices of natural gas and crude oil for
the years: 1997 - 2016. Henry Hub spot prices and Cushing, OK spot prices are often listed as the
national price of natural gas and crude oil, respectively. Both figures capture the rise of energy
prices through the 2000s. Natural gas spot prices collapsed around 2008, which is consistent with
the onset of the Great Recession, and since then, the prices have remained relatively steady
around US$ 3.5 per million Btu. Crude oil prices, on the other hand, declined slightly around
0
1
2
3
4
5
6
7
8
9
10
1997
19
98
1999
20
00
2001
20
02
2003
20
04
2005
20
06
2007
20
08
2009
20
10
2011
20
12
2013
20
14
2015
20
16
Nat
rual
Gas
Spo
t Pric
e (U
S D
olla
rs p
er M
illio
n B
tu)
Year
0
20
40
60
80
100
120
1997
19
98
1999
20
00
2001
20
02
2003
20
04
2005
20
06
2007
20
08
2009
20
10
2011
20
12
2013
20
14
2015
20
16
WTI
Spo
t Pric
e (U
S D
olla
rs p
er B
arre
l)
Year
11
2008, but then rebounded by 2010 and remained relatively flat until 2013. The price of crude oil
fell after 2013 and settled at near US$40 per barrel by the end of 2016.
2.2 Wastewater Injections
Oklahoma has recently experienced a large increase in the frequency and volume of wastewater
injections, which are disposed of in the State’s Class II injection wells. According to Fractracker
(2016), between 2011 and 2015, a monthly average of 68 million gallons of wastewater was
added to the “New Dominion, LLC Chamber #1 well” alone, which is located in Oklahoma
County. The total volume disposed of, between 2011 and 2015, was 410 billion gallons of
wastewater, which approximately equals 6.8 billion gallons per month (Fractracker, 2016).
Further, the mean monthly injected volume, across the State’s 10,927 total Class II injection
wells, equaled approximately 37 million gallons.
To see how the frequency and volume of wastewater injections have changed through
time, we have provided two diagrams in Figure 5. Both of the diagrams illustrate the State’s total
number of injections (from 1970 to 2016) of what the Oklahoma Geological Survey defines as a
type “2D” well, which are designed to dispose of brine water that was co-produced with oil and
gas (Murray, 2014). The top panel, of Figure 5, shows the frequency of injections as measured
by the pressure (in pounds per square inch) of injections. The bottom panel illustrates the total
volume of injections as measured by the barrels of fluid disposed of in the State’s Class II wells.
The grey lines show the temporal variation in disposals, whereas the dark lines show a weekly
moving average of disposals across the entire sample period (1970-2016). Both panels illustrate
relatively stable average levels, of injections, from the 1970 to approximately 2005. However,
both panels show a marked increase in the average levels starting around the year 2007. The
12
panels also show a slight decrease in the frequency and volume of injections around the year
2014 – the period of time in which the State’s injections and seismic activity started to capture
national media attention.
The fluid injections of type 2D have been highly correlated to seismicity (Horton, 2012;
Keranen et al., 2013; Nicholson and Wesson, 1990), so we discuss the increase in seismicity in
the next subsection.
13
Figure 5. Class II wastewater injections in Oklahoma (1970-2016)
(a) Total pressure (psi) of wastewater injection
(b) Total volume (bbls) of wastewater injections
Notes: Panels (a) and (b) illustrate injections into type “2D” wells, which are designed to dispose of brine water that was co-produced with oil and gas.
0
1,000
2,000
3,000
4,000
5,000
1/21/1
970
1/21/1
974
1/21/1
978
1/21/1
982
1/21/1
986
1/21/1
990
1/21/1
994
1/21/1
998
1/21/2
002
1/21/2
006
1/21/2
010
1/21/2
014Po
unds
per
squa
re in
ch o
f was
tew
ater
in
ject
ions
Date
010,00020,00030,00040,00050,00060,00070,00080,00090,000
100,000
1/21/1
970
1/21/1
974
1/21/1
978
1/21/1
982
1/21/1
986
1/21/1
990
1/21/1
994
1/21/1
998
1/21/2
002
1/21/2
006
1/21/2
010
1/21/2
014
Barr
els o
f was
tew
ater
inje
ctio
ns
Date
14
2.3 Earthquake activity
As demonstrated by the dotted, black line in Figure 6, in 2015, Oklahoma experienced
approximately 840 earthquakes measuring at least 3.0 in magnitude. This stands in stark contrast
to the year 2008, in which the State experienced only one earthquake of at least 3.0 in magnitude
(Mw). Marked by the dark grey line in Figure 6, the number of lesser magnitude earthquakes (less
than 3.0 in magnitude) has also increased from one event in 2008 to approximately 1500 total
events in 2015. Further, the sold black line indicates the number of subjective or self-reported
earthquake events (provided by the USGS “Did you feel it?” database). Similar to the other
series in 2015, the total number of self-reported quakes reached nearly 1,000. As a result,
Oklahoma has now been identified as the most seismically active state in the continental U.S.
(Philips, November 7, 2016).
Figure 6. Oklahoma’s annual frequency of earthquake events by year: 2002-2016
Notes: Mw denotes the moment magnitude scale, which is used as a measure of the seismicity of an earthquake event.
0200400600800
1,0001,2001,4001,600
2002
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
Freq
uenc
y of
seis
mic
eve
nt
Year
Mw > 3.0 Mw < 3.0 DYFI
15
2.4 Real estate market trends
Similar to housing market trends in other of metropolitan areas in the Southern U.S., the
Oklahoma City real estate market has fared well over the past several decades. Figure 7 displays
a housing price index for the Oklahoma City MSA from 1980 to 2016 (U.S. Federal Housing
Finance Agency, 2017). With the exception of the 1983-to-1988 period, the figure shows that
real estate prices in Oklahoma City have been steadily appreciating over time. Despite the steady
growth in prices, the Oklahoma City real estate market fell slightly short of the national average
in 2016 – that is, the average three-year appreciation rate (leading up to 2016) was 17.8 percent
for the nation as a whole whereas it was only 3.9 percent for Oklahoma City. (Merrill, November
14, 2016). According to the same source, the area’s 2016 median home price was about
US$154.9 thousand with a population of approximately 610.6 thousand and a median household
income of US$45.7 thousand. Otherwise, the local economy remains fairly strong with relatively
low levels of unemployment; however, the real estate market is forecasted to experience
relatively weaker price growth through 2017 (Merrill, November 14, 2016).
16
Figure 7. All-Transactions House Price Index for Oklahoma City, OK (MSA)
Notes: Shaded areas indicate U.S. recessionary periods. Source: U.S. Federal Housing Finance Agency (2017)
2.5 Literature review
Recent hedonic studies have examined the indirect link between oil and gas development,
induced earthquake activities, and the impact to housing prices (Koster and van Ommeren, 2015;
Liu et al., 2016). For example, Koster and van Ommeren (2015) explored this phenomenon
within the housing market of Groningen, Netherlands (a region with a relatively rich history of
natural gas development). They found that seismic activities, of sufficient magnitude, led to a 1.9
percent decrease in nearby housing prices, which implies a decrease in prices of about €3.5
thousand (euros) (approximately US$4 thousand) per household.
In a somewhat similar vein, Liu et al. (2016) explored the perceived impacts of
wastewater injection and induced seismicity on housing prices in Oklahoma City. Liu et al.
40
60
80
100
120
140
160
180
200
220
1977
1979
1980
1982
1984
1986
1988
1990
1991
1993
1995
1997
1999
2001
2002
2004
2006
2008
2010
2012
2013
2015
Inde
x 19
95:Q
1 =
100
Year
17
(2016) find fairly similar effects (provided the earthquake activity is moderate to large) with an
estimated average loss of 2.2 percent or approximately US$4.4 thousand per household. Our
study differs from Liu et al. (2016) by more closely examining (and estimating) the causal
mechanism among natural gas production, consequent induced seismicity and the impacts to
property values. Unlike Liu et al. (2016), we do not directly analyze the effect of public
perception on housing prices, although public perceptions are implicitly captured within the
census-level and time fixed effects. Instead, we attempt to tease out the exogenous variation,
associated with the induced earthquake activities, to offer a more accurate estimate of the
impacts of induced earthquakes on housing prices.
An additional study was recently offered by Cheung et al. (2016), in which the authors
did not explore the effects of oil and gas development explicitly, but the authors did examine the
general impact of Oklahoma's earthquake activities on property values between 2006 and 2014;
and, they found that prices declined by 3 - 4 percent (on average) in areas where homes
experienced moderate-to-large earthquakes. Our study differs from Cheung et al. (2016) by
offering an unbiased estimate of the induced earthquake impacts. More specifically, the authors
use an attenuation function method to estimate earthquake intensity, which we contend suffers
from measurement error and thus provides for biased impact estimates.
In addition to research that has focused on the economic impacts associated with induced
earthquakes, other past studies have focused on the direct causal pathway from development (or
production) to effects on housing prices (Muehlenbachs et al, 2015; Gopalakrishnan and Klaiber,
2014; Boslett et al., 2016; Weber et al., 2016). For example, Muehlenbachs et al (2015) found
that a residential property, within 1.5 km to an unconventional natural gas well pad, experienced
a 9.9 - 16.5 percent decrease in value if the home was dependent on private well water.
18
Gopalakrishnan and Klaiber (2014) also found negative impacts to home prices, if the property
was dependent on private well water, ranging from a loss of US$4 - US$8 thousand. Further,
Boslett et al. (2016) found that natural gas development led to 10 - 23 percent loss in property
value, which translates to an approximate US$25 thousand loss (based on the 23 percent
estimate), per household, on average. Unlike the previous studies, Weber et al. (2016) found that
shale gas development, in Texas, led to a decline in assessed property tax rates and increasing
revenues to local schools; however, their estimates also suggested that the cumulative density of
nearby gas wells slightly reduced property values.
3 CONCEPTUAL FRAMEWORK
To analyze the effect of earthquake activities on housing prices, we consider numerous potential
intervening factors, including wastewater well injections, oil and gas production, and housing
and neighborhood characteristics, among other covariates. Our primary relationship of interest is
the effect of induced earthquakes on housing prices. In order to obtain credible estimates of the
impacts of induced earthquakes, in an OLS framework, we must ensure that the observed values
of induced earthquakes come from a random (or nearly random) sample. Put differently,
earthquakes activities, conditional on the other observed factors, must be independent of housing
prices.
An experimental or randomized control trial approach, to exploring this particular
phenomenon, would imply that some households should be randomly exposed to earthquakes,
while other households (in the same population) have little or no exposure, and thus satisfying
19
the (conditional) independence assumption (discussed further below). In the absence of
randomly-assigned earthquake exposures, even a quasi-experimental approach, to estimate the
impacts from earthquakes to housing prices, would yield inconsistent effects estimates. For
example, if one were to assign a treatment (exposure to an earthquake) based on some pre-
specified distance band to an earthquake’s epicenter, then biasedness may arise due to a potential
contamination within the experimental design – in other words, some households assigned to the
control group may experience exposure to the earthquake, and vice versa.
Another argument is that wealthier households may self-select away from regions with
existing natural gas and oil drilling wells or wastewater injection wells. In such a case, the
regression analysis would potentially overdraw from a sample of middle to lower income
housing, if we were just to focus on the affected regions within proximity to injection wells. The
conclusion to be drawn here is that the epicenter(s) of earthquake activity, and potentially the
geographic proximity of injection wells, may be correlated with unobserved factors affecting the
housing market, located in proximity to the epicenter, and thus generating biased earthquake
impact estimates.
To make these arguments a little clearer, consider Figure 8. Figure 8 demonstrates the
areal extent of a magnitude 5.8 earthquake (5-kilometer depth) with an epicenter located
approximately nine miles from Pawnee, Oklahoma (Pawnee is located about 72 miles north-by-
northeast of Oklahoma City). The Pawnee earthquake occurred on September 3, 2016. The areal
extent of the Pawnee earthquake was estimated based on the event’s peak ground velocity
(Koster and van Ommeren, 2015, p. 123). It is worth noting here, that seismologists generally
cannot estimate, with a high degree of certainty, the inner variability of ground motion and
subsidence within the areal extent of an earthquake event. Such types of measures would
20
arguably require a sufficient number of evenly-spaced seismographs across the entire areal extent
of an earthquake. However, there are a relatively sparse number of seismographs within the
entire western and central regions of the State. Therefore, if an earthquake occurs in Pawnee,
which is 72 miles north of Oklahoma City, then seismologists are forced to estimate the areal
extent of the event based on randomly located seismographs throughout the State and region.
One such estimate of the areal extent is peak ground velocity, which is demonstrated
below in Figure 8. (There are numerous other seismological estimates of earthquakes impacts;
however, peak ground velocity was used in both Liu et al. (2016) and Cheung et al. (2016)).
Peak ground velocity should provide an estimate of how far, in terms of distance from the
epicenter, a household could potentially feel the ground motion associated with an earthquake.
The darker green shade (which indicates a peak ground velocity of 0.1 - 1.1 centimeters per
second) implies that a majority of households in the Oklahoma City MSA should have
approximately felt the same amount of ground motion associated with that particular earthquake.
The small black dots on the map indicate a “Did you feel it?” or self-reported responses – we
further explore the “Did you feel it?” data below. (The lateral motion, associated with an
earthquake, will depend on the soil and sub-soil profiles underlying a property, so it is possibly
feasible that two properties with differing soil types, in the same region, may perceive the motion
differently. We discuss this further below). The Pawnee earthquake only represents one event;
nevertheless, the peak ground velocity diagram below demonstrates the potential challenge with
assigning a household observation to a treatment or control group based on a pre-specified
proximity band to the earthquake, such as 5 kilometers. Based on this one event, one would have
to assign a treatment proximity indicator based on an approximate 325-kilometer distance band,
and as a result, the price of homes in Oklahoma City (the region of interest) would need to be
21
compared to housing prices in Kansas City or Dallas - Ft. Worth. On the other hand, earthquakes
of smaller magnitude (which are far more frequent) affect a smaller areal extent. Without stable
counterfactual observations, it would be challenging to derive a credible average conditional
treatment effect associated with these earthquake activities. This is sometimes referred to as a
violation of the “stable unit treatment value” assumption – that is, a treatment applied to one unit
does not affect the outcome for another unit (Rubin, 1974, 2006; Abadie and Imbens, 2006).
Figure 8. Areal extent of the impacts associated with the Pawnee, OK Earthquake of 2016
Figure 8 illustrates another potential challenge in estimating the housing price impacts
associated with induced earthquakes. This figure shows the "Did you feel it?" (DYFI) data based
on the same Pawnee earthquake event (U.S. Geological Survey, 2017). The DYFI data contain
self-reported or subjective measures of seismicity; that is, if a household perceives that they have
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!!!
!
!
!
!
!!
!
!!
!!
!!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!
!
!
!
!
!!
!
!
!
!!
!
!
!
!
!!
!
!
!
!!
!
!
!!
!
!
!!!
!
!!!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!!
!
!
!
!!!!!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!!!
!!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!!
!
!
!
!
!!!!
!
!!
!
!
!!!!!
!
!!
!!
!
!!
!!!!!
!!
!!!
!
!
!!
!
!
!
!!!!
!!
!
!
!
!
!
!
!
!
!!!!!!
!
!
!!
!
!!
!
!!
!
!
!!
!!
!
!!
!!!
!
!
!!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!!!
!
!
!
!
!
!
!
!
!!!
!
!
!!
!
!
!
!
!
!
!
!!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!!
!
!!
!
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!!!!
!
!
!
!!
!
!
!
!!!
!
!
!
!!
!
!!
!
!
!
!!!!
!
!
!
!
!
!!!
!
!
!
!!!!!
!
!
!
!
!
!
!!!!
!
!
!
!
!!
!
!!
!
!
!
!!
!!
!!
!!
!
!!
!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!!
!!!
!
!
!
!!!!
!
!
!
!
!!
!
!
!
!!
!!!!
!!!
!
!
!
!
!
!
!
!!!!!
!
!
!
!
!
!
!
!
!
!!
!
!!!
!
!
!
!
!
!
!
!!!
!
!!!!!
!!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!!!
!
!
!
!
!
!!!!!
!!!!
!
!
!
!
!
!
!
!!!!!!!
!!!
!
!
!
!!
!
!
!!!!!
!!
!!!!
!
!
!
!!
!
!
!
!
!!!
!
!
!!!!!!
!
!
!!
!
!
!
!!!!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!!!!!
!
!
!
!
!!
!
!
!
!
!
!
!!!
!
!
!
!
!!
!
!
!
!!
!
!!!!
!
!!
!
!
!
!
!
!
!!!!!!!!
!
!
!
!
!
!!!!!!!
!
!
!
!!!!!!!!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!!!
!!
!
!
!
!
!
!!
!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!!
!
!
!
!!
!
!!
!
!!
!
!
!!
!
!
!
!
!
!
!
!!
!!
!!
!
!
!
!
!!!!!
!!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!!!!
!
!
!
!
!
!!!
!
!
!!!!!!!!
!
!
!
!!!!
!
!
!
!!!
!
!
!
!!
!
!!!!!!!!
!!
!
!
!
!!!!!!!
!
!
!
!!!!
!
!!!!
!
!
!
!!!!
!
!
!
!!!!!!!!!!!
!
!
!
!
!!!!
!!!
!
!
!!!
!
!!!
!
!!!!
!
!!!!
!!
!
!!!!!!
!
!
!
!
!!!
!
!
!
!!!
!
!!
!!!
!
!
!!!!!!
!
!!!!
!
!
!
!
!!
!
!
!!!
!
!
!
!!
!!!!!
!!
!
!
!!!!
!!
!
!
!
!
!!
!
!
!
!
!
!!
!!!!!!!!
!
!!!
!
!
!
!
!
!
!!!!!!
!
!
!
!
!
!!!
!
!
!!
!
!
!
!
!
!!!!!!!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!!
!
!
!
!
!
!!
!
!
!
!
!
!!
!!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!!
!!!!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!!
!
!!!
!
!!!
!!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!!
!
!!!!!!!!
!
!!!!
!
!
!
!
!
!!!
!
!
!!
!
!
!!
!
!!!!!
!!!!!!!!
!
!!!
!
!
!
!
!
!
!
!
!!
!!!
!
!
!
!
!
!
!
!!!!!
!!!!!!!!!!!
!!!
!
!
!
!
!
!!
!!
!
!
!!!
!
!
!!!
!
!
!
!!
!!!!!!!
!!
!!!!!!!!!!!!
!!!
!!!
!
!
!
!
!!
!
!
!
!!!
!
!!!
!
!!!
!
!
!
!
!
!
!
!
!
!!!!!!!!
!!!
!!!!
!!
!
!
!!!
!
!
!
!!!!!!
!!!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!!!
!!
!
!!!!
!!!!!
!
!
!!
!
!
!!
!
!!!!!
!
!
!
!
!
!
!
!
!
!!!!
!
!!!!!!!
!
!
!
!
!
!
!!
!!
!!!
!!!
!
!
!
!
!!
!
!!!
!
!
!
!
!!
!!
!
!
!!!!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!!!
!!!!!!!!
!
!
!
!
!
!
!
!!
!
!
!
!!
!!!
!!!!!!!!!!
!!
!
!!
!
!
!
!
!
!
!
!
!!!!!!!!!
!
!
!
!
!
!
!
!
!
!!
!
!!!!
!
!!
!!!!!!
!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!
!
!
!!!!!!!!!
!
!
!!
!
!
!!
!
!
!
!
!
!!!!
!!!!
!
!
!!
!!
!!!!!!
!!!!!!
!
!
!
!!
!
!
!!!!!!
!
!
!
!
!
!!
!
!!
!
!
!
!!!
!
!!!!!!!!!!!!!!!!!!!
!
!
!
!
!!
!
!
!
!
!
!
!!!!!
!
!
!
!!
!
!
!
!
!
!
!!!
!!!!
!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!
!
!!!
!
!!
!!
!
!
!
!
!!!!!!
!
!!!!
!!!!!!!!!!!!!!!!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!!
!
!!
!
!
!
!!!
!!
!!
!!!!!!!!
!!!!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!!
!
!!!!!!!!!!!!
!!!!
!
!
!
!
!!
!!!
!!
!
!
!
!
!!!!
!
!
!
!
!!
!
!!
!
!!!!!!!!!
!!
!!!!
!
!
!
!
!!
!
!
!
!!
!
!!
!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!
!
!
!!
!!!
!!!
!!!
!
!
!
!
!
!
!!!
!!!
!
!
!
!
!
!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!
!!!
!
!
!
!!!
!
!
!!
!!!
!!
!!!
!
!!
!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!!!
!
!!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!
!
!
!!
!
!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!!
!!
!
!
!!!!!!!!!!!!!!!!
!
!
!
!
!
!
!!!
!
!!
!
!
!!
!!
!!!!!!!!!!
!!!!!!!!!!!!!!
!!!
!!!!!!!!!!!!!!!!
!
!
!!
!!!!!!!!!!!!!!!
!
!
!
!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!
!!!
!!!!!!!!!
!!!
!
!
!
!!
!!!!!!!!!!!!!!!
!!!
!
!
!!
!!!
!
!
!
!
!
!!!
!!!!!!!!!!!!!!!!!!!!!!
!
!!!!
!
!!!
!!!!!!!!!!!!
!
!
!
!!!!!!!!!!!!!!!!
!
!
!!
!
!
!!!!!
!!
!
!
!
!
!!!!!!!
!!!!!!!!
!
!!
!!
!!!!!!!!!!!
!
!
!
!
!
!
!!!!!!!!!!!!!!!
!
!
!!
!!
!
!
!
!
!
!
!!!
!
!
!!
!
!
!
!!
!!!!!!!
!!!!!!!!
!!
!!!!!!!!!!!!!
!
!
!
!!!!!!!!!!!!
!!
!
!!!!
!
!!!
!
!!
!
!
!!!
!
!
!!!!!!!!
!!!!!!!!
!
!!
!
!!!!!!!!!!!!
!!!
!
!
!
!
!
!!
!!!!!!!!!!
!
!!
!
!!
!!!!
!
!
!
!!
!
!!!!!!!!!!!!
!!!!!!!!
!!!!!!!
!!!
!!!!!!!!!!!!!!
!
!
!!!
!
!!
!
!!
!!
!!!!
!
!
!
!
!
!!
!!
!
!
!!!
!
!
!!
!
!
!
!
!
!!!!!!!!!!
!!!
!!!
!!!!!!!
!
!
!!!!!!!!!!!!!!!
!
!!
!
!
!!!!
!
!
!!!!!!!!
!!!
!
!
!
!
!
!
!!!
!
!
!!!!!!!!!!!!!!!!
!!!!!!!!!!!!
!
!!!!!!!
!!
!
!!!
!!!
!
!
!!!!!
!
!
!!!!
!!!!!!!
!!!!
!!!
!
!
!!!!
!!
!
!
!
!
!
!
!!!!!!!!!!!!!
!!!!
!!!!!!!!!!!!!!
!!!!
!!
!
!!
!
!
!!
!
!!!!
!
!
!!!
!!!!!!!!!!!!!!!!!!!!!
!!!!!
!!!
!
!
!
!!!!!
!
!!
!
!
!!
!
!
!!
!!!!!!!!!!!
!!!!
!!!!!!!!!
!!!!!
!
!!
!!!
!
!
!!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!
!
!!!!!!!!!!
!!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!!!!!!!!!!!!
!!
!!!!!!!
!
!!!!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!
!!!!!!!!!
!
!
!
!
!
!
!!!!
!!!
!!!!
!
!
!!
!!!!!
!!!!
!!
!
!
!!!!!!!
!
!!
!
!!!!!
!
!
!
!!!!!!!!!!!!!!!!!!!!
!
!
!!
!!!!
!!
!!
!!
!
!
!
!
!
!
!
!!!!
!!!!
!
!
!
!!!!!!
!
!!
!
!
!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!
!!!!
!!!
!
!
!
!!
!!!!!!!
!
!
!
!!
!
!
!!
!!!!
!
!
!!!!!!!!
!
!
!!!!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!
!!!!
!!!!
!!
!
!
!
!!
!
!!!!
!!!!
!
!
!
!
!
!!!
!
!!!!
!!!!!!
!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!!
!!!!!
!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!
!
!
!
!!!
!!
!
!
!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!!
!!!
!!!!!
!!
!
!!
!!
!
!
!!!
!
!!!!!!
!!
!!!!
!!
!!!!!!!!!!
!
!
!
!
!
!!!
!!
!
!!
!!
!!!
!!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!!!!!!!!
!!
!
!
!
!
!
!
!
!
!
!!
!
!!
!!!!
!
!!
!
!
!
!!!
!!
!!!
!!!!!!
!!!!!!!!!!!!!!!!!
!!
!
!
!
!!!!
!!
!!!!!!!
!!!!
!!!!
!
!
!
!
!!
!!!
!!!!!
!
!
!
!
!
!
!
!!!!!
!
!!!!!!!!!!!!!
!
!!!!!!!!!!!!!
!
!
!
!
!!
!!!!!
!!!!!!
!!!!!
!!
!
!
!!
!
!
!
!
!!
!!!
!!
!!
!
!!
!
!!
!
!
!
!
!!!
!!!!!!!!
!
!!!!!!!!!!!!!
!
!
!
!!!!!
!
!
!!!!
!!!!
!!!!!!
!!
!
!!
!
!
!
!
!!!!
!
!
!!!!!!
!!
!!
!!
!
!
!
!!!!
!!!!!
!!!!!!!!!!!!!!!!!!!!
!
!!
!
!!
!
!!
!!!!!!!!!
!!!!!!!!
!
!
!
!
!
!
!
!!!!!
!!
!
!
!!!!!!!
!
!!
!!!
!!!
!
!!!!!!!!!!!!!!!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!!!
!
!
!
!!
!!
!
!
!
!
!
!
!!
!
!
!!!!!!!!!!!
!
!
!
!
!!
!
!!!!!!!!!!
!
!
!
!
!!
!
!
!
!
!!
!
!
!!!!!!!!!!!!
!!!
!
!
!!
!
!
!
!
!
!
!
!!
!!!
!!!!
!
!
!
!
!!
!
!
!!!!!!!!!!
!
!
!!
!!
!
!!
!!
!!!!!!!!!!!!!!
!
!
!
!
!
!
!!
!
!
!
!!!
!!!!!!
!
!
!
!!
!
!
!
!
!
!
!!!!!!!!!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!
!!
!
!
!
!
!!
!
!
!
!
!
!
!!!!!!
!
!
!!
!!
!
!!!!!!!!!!!!
!
!
!
!
!
!!!
!!!!!!!!!!!!
!
!
!
!!!
!
!
!!
!!!!!!
!
!
!
!
!
!
!
!
!!!!!!!!!
!
!
!!!
!!
!
!
!!!!!!!
!!!!!!
!
!
!
!
!
!
!!!
!
!!
!
!!
!!!!
!!
!
!
!
!
!
!
!
!
!!!
!
!!
!!
!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!
!
!
!
!
!
!!!
!
!
!!!
!
!
!
!
!
!!
!!
!
!!
!
!
!
!!!!
!!!!!!!
!
!
!
!
!
!
!!!!!!!!!!!!!!!
!
!
!
!!!
!!
!
!
!
!
!!
!
!
!!
!!
!!
!!
!
!
!
!!!!!!!
!!!!!!!!
!
!
!
!
!
!
!
!
!
!!!
!!!!!!!!!!!!!
!!
!!!
!
!
!
!
!
!
!
!
!
!
!!
!!!
!!!
!
!
!
!
!
!
!
!
!!!
!!!!!!!!
!!!!
!!
!
!
!
!
!!
!!!
!
!!!!!!!!!!!!!!!
!!
!
!
!
!
!
!
!
!
!
!!
!!!!!
!
!!
!
!!
!
!
!
!
!
!!!
!
!
!!
!!!
!
!
!
!!
!
!
!
!
!
!
!!
!!!!!!!!!!!!!!!
!!
!
!
!
!!
!
!
!
!!!
!!!!!
!
!
!!!!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!!!!!!!!!!!!!!!!!!
!!!
!
!
!
!
!!
!
!
!
!!!!!!!!
!
!!
!
!
!
!
!
!
!
!
!!!!!!!!
!!
!
!!!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!!
!!!
!!!
!
!
!!!!
!
!
!!!
!
!
!!!!
!!
!
!
!
!
!
!!!!!!!
!
!!!!!
!!!
!!!!!!!
!
!!!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!!
!!!!!
!
!!!!!
!
!
!
!!
!
!
!
!
!
!!
!!!!!!!!!
!!!
!!!!!
!!
!!!!!!!!!!!!
!
!!
!!!
!
!!!
!
!
!
!
!
!!!!!!!
!
!
!
!!
!
!
!!
!!
!!!!!
!
!
!!!!!!!
!
!!!
!
!
!!
!
!
!
!!
!
!
!!!!!!
!!!!!!
!!!!
!!!
!
!!
!!
!
!!!!!!!
!!!!
!
!
!
!!
!
!
!
!
!!
!
!
!
!!!!!!!!!
!!!!!!!
!
!
!!
!
!
!!!!
!
!
!!
!
!
!
!
!
!!!
!!
!!!!!!!!
!!
!!!
!!!
!!!
!
!!!
!!!!!!!
!
!
!
!
!!!
!!
!
!!!
!
!
!
!
!!!!!
!
!
!
!
!!!!!!!!!!!
!
!!!!!!!!
!!
!!!!!!!!!
!
!!
!
!!
!
!
!!
!
!!!!!!!!!!!!
!!!
!!!!!
!
!!!
!!!!!!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!!
!!!!!!!!!!!!
!
!
!
!!!!
!
!
!!!!!!
!
!
!
!
!
!
!
!
!
!
!!!!!
!!!
!
!!
!!
!!!
!!!!!
!!
!
!!!
!!!!!
!!!
!
!
!
!
!
!
!
!
!!!!!!!!!!
!!!
!
!
!!!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!!!
!!!!!!!!!!!!!
!!!!!!!!
!!!
!
!!
!!!!!!!!!!!!
!
!
!
!
!
!
!
!!!
!!!!!!!
!!
!
!
!
!
!!
!
!!
!
!
!!
!
!!
!
!
!
!
!!
!
!!!!!!!!
!!!!!!!!
!
!!
!!
!!!!
!!!
!!!!!
!!!!
!
!
!
!!
!
!
!
!!!
!!!!!!
!
!
!
!
!!!
!!!!!!
!!
!!
!!
!
!
!
!
!
!
!!!!!!!!
!!!!!!
!!!
!
!!!!!
!
!
!!!
!!!!
!!!!!
!
!
!
!
!
!
!
!
!!
!
!!!
!
!
!
!
!
!
!!!!!!
!
!
!
!!!!!!
!
!
!
!
!
!!!!!!!!!!!!!!!
!!!!!
!!!!!!
!!!
!!!!!!!!!
!!!
!
!!
!
!
!
!
!
!
!
!
!
!!!
!!!
!
!
!
!!!!
!
!
!!!!!
!!!
!
!!!!
!!!!!
!!!!!!
!!!
!!!!!!!
!!!
!!!!!!!!!!
!!!!!!
!!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!!
!
!
!
!
!
!
!!!
!!!!!
!
!!
!!!!!!!!
!
!
!
!!!!!!!!
!!!!
!
!
!
!!
!!
!
!!!!
!
!
!
!
!
!
!
!
!!
!
!
!!!
!!!!!!!!
!
!!
!!!!
!!
!!!
!!!!
!!!!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!!
!
!
!!!
!
!!!!
!!!
!!!
!!!
!
!!!
!!!!!!!
!!!!!
!
!!!
!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!!
!!
!
!!!!!!!
!!!!!!!!
!
!!!
!!
!!!
!
!!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!!
!
!
!
!!
!!
!
!
!
!
!
!!!
!!!!!!
!!
!
!
!!
!!!
!!!!!!!!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!!!
!!!!!!
!!!!
!
!!!!!!
!!!!!!!
!
!!
!!!
!
!!
!
!!
!
!
!
!!
!
!!
!
!
!
!
!
!
!!
!!!!!!!!!
!
!!
!
!
!
!!!!!!!!!!!
!!!!!!
!!
!!!!!!!!!!!
!!
!!
!
!!
!
!
!
!!
!
!
!
!!!
!!!!!!
!
!
!
!
!
!
!
!!!
!
!
!
!!!!!!
!
!
!
!!!!
!!!!!!!!
!!!
!!!!!!!
!
!!!!!!
!
!
!
!!
!
!!!!!
!!!!!!!!
!
!
!
!
!
!
!
!
!!!
!
!!!!!!!
!
!
!
!
!!!
!
!!!!!!!!!!!!!!!
!!!!!!
!
!!
!
!!
!
!
!
!
!!!!!!!!
!
!!
!
!!
!
!!!!!
!
!
!!
!
!
!
!!!
!
!!
!!!!!!!!!!!!!!!!!!!
!!
!!
!!!
!
!
!!!!!
!!!
!
!
!
!
!
!
!!!!
!
!
!!
!!
!
!
!
!!
!
!
!!
!!!!!!!!!
!!!!!!!
!
!!
!!!!
!
!
!
!
!!!!!!!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!!
!
!!
!!!!!!!!!!!!!!!
!!!!!
!!
!!!
!!!
!
!
!
!!!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!!!!
!
!!
!!!!!!
!
!!!!
!!
!!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!!
!
!!!!!!!!!!!!
!!!!!
!!!
!!
!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!
!!!!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!
!
!!!!!!!!!!
!!!!!!!
!!!!!!!!!!!!!!!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!!
!
!!!!!!!!!!!!
!!!!!!!!!!!!!!!
!!!!!!!
!!
!
!
!
!
!!
!
!!
!
!
!
!!
!
!
!
!!!
!
!!!
!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!!
!!!!!!!
!!!!!!!!!!!!!!!!!
!!!!!!!
!
!
!
!
!
!
!!!!
!
!
!!
!
!
!
!!!!
!!!
!
!!!!!!!!
!
!!!!!!
!
!
!
!
!
!!!!
!!
!
!
!
!
!
!
!
!
!
!!!!!!!!!
!!!!!!!
!!
!!!!!!!!!!!
!!
!
!!
!!
!
!
!
!
!
!
!!!!
!
!
!
!!!!!!!!!!
!!!
!!!!!!
!
!
!
!!!
!!!!!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!!!!!
!
!!
!
!
!!
!
!
!
!!!!!!!!!!!
!
!!!!
!!!
!!!
!!!
!!!!
!!
!
!!
!
!!
!
!
!!
!
!
!!!
!
!
!
!!!!!
!
!!
!
!
!
!!!!!!!!!!!!!!
!!!!!
!!
!!!!!
!!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!!!!!!!!
!
!!!!!!!!!!!
!!
!
!!
!!!
!!!
!
!
!
!
!!!
!
!
!
!
!!!!!!!
!
!!!!!!!!
!!!!!!!!
!!!!!!!
!
!!
!!!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!!!!
!!!!
!
!!
!!!!!!!!!!
!
!!!!!
!!!!
!
!
!
!
!!
!
!
!
!
!!!!!!!!!
!
!
!
!
!
!!!!
!!!
!!!!
!
!!
!!
!
!
!
!
!
!!!
!
!
!
!!
!!
!
!
!!!!!!
!
!!!!!
!!
!!!!
!!!
!
!
!
!
!
!!
!
!!!
!!
!
!
!
!
!
!
!!!!!
!
!
!!!
!
!!!
!!
!
!!
!
!
!!!!!
!
!
!
!
!
!!!!!
!!
!
!!!!!
!
!!
!
!
!
!!
!
!
!
!
!
!
!!
!
!!!!
!!!!!!!
!
!!
!
!
!
!!
!
!!!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!!
!
!!!
!
!!!!!
!!!!!!
!
!
!
!
!!
!
!
!
!
!
!!
!!!!
!
!!!
!
!!
!
!!!
!
!!
!!
!!
!!
!!
!!
!
!
!!
!
!
!!!
!
!
!
!
!
!
!!!
!
!
!
!!
!
!
!!
!
!
!!!!
!
!
!!
!!
!
!
!!!
!
!
!
!!!
!
!
!
!!
!!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!!!!
!!!!!
!
!!!!
!
!
!
!
!
!
!!!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!!
!!!!!!!
!!!!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!
!!!
!
!!!
!!
!!
!!!!!
!
!
!!!!!!!
!
!
!!
!
!
!
!
!
!
!
!!!!
!!!!!!!!!!!
!
!!!
!
!!
!!!!!
!!!!
!!
!!
!!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!!!!
!!!!!!
!!!!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!!!!
!!
!
!!
!
!!!!!!!!!!!!!
!!
!!
!
!!!!!
!
!
!
!
!
!
!
!
!
!!!!!
!
!!!!!!!!!!!!!!!!!!!!!!
!!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!!!!
!!!!!
!!
!!!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!!!
!!!!!!!
!!
!
!!!!!
!
!
!
!
!
!
!
!
!!!
!
!!
!!
!!!!!!!!!!!!!!!!
!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!
!!!!
!
!
!
!
!!
!!!
!
!
!
!
!
!!
!
!
!
!
!!
!
!!!!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!!!!!!!!!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!
!
!
!!
!
!
!
!
!
!
!
!
!!!!!
!
!!!!
!!!!!!!!
!
!!
!
!
!
!
!!
!!
!
!
!!!
!
!
!
!!
!!!
!
!!!
!
!
!
!
!
!
!
!
!!
!!!
!
!!
!
!
!
!
!
!
!
!!!
!!
!!
!
!!
!
!
!
!
!!
!!!
!
!
!
!
!
!
!!
!!
!
!!
!
!
!
!!!
!
!!!
!!
!
!!!
!!
!
!
!
!
!
!!
!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!!!
!!
!
!!
!
!
!
!
!
!!
!!
!
!!!
!
!
!
!
!
!!!!
!!!!
!
!
!!!
!
!!
!!
!
!
!
!
!
!!
!
!
!
!!
!
!
!!
!
!
!!
!
!!!
!
!
!!
!
!!
!!!!
!!
!!
!
!
!!
!
!
!
!
!
!!!!!!!!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!!
!
!
!
!
!
!!!
!
!
!
!!!
!
!
!!
!
!
!
!!
!
!
!
!
!!
!
!!
!
!
!!!!!
!
!
!!
!
!
!
!
!
!
!
!!
!!!!!
!!
!
!
!
!
!!!!
!
!!!
!
!!!
!
!
!
!
!
!
!
!
!
!
!!!!!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!!!
!
!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!
!!!!
!
!
!!!
!
!!
!
!
!
!
!!
!!
!
!
!
!
!!!
!
!
!
!!!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!!
!!
!
!!
!!!
!
!!!
!
!
!!
!
!
!!!
!!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!!
!
!
!
!
!
!!
!!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!!!!!!
!
!
!
!
!
!!!
!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!!!!!!!
!
!
!
!
!!!
!
!
!
!
!!!!!!!
!
!
!!!!!
!!
!!
!
!
!
!!!
!
!
!
!!!
!
!
!
!!!!
!
!
!
!
!!
!
!
!
!!!!!!
!!!
!!!!
!
!
!
!
!!
!
!!
!
!
!
!!
!
!
!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!!
!!!!!
!!
!
!
!
!
!
!
!
!
!!
!!!
!!
!!
!
!!!
!
!
!
!
!
!
!
!
!!!
!!!!!!!!!
!!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!
!!!
!
!!!
!
!
!
!
!!
!
!
!
!
!
!!!!!!!!!!!!!!!
!!!!!!
!!!
!
!
!!!
!
!
!
!
!
!
!
!
!!
!!!!!!!
!!!!!!!!!!!!
!
!
!!
!
!
!
!
!
!
!
!!!
!
!!!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!! !
!
!
!
!
!
!
!
!!!!
!
!
!
! !
!
!
!!!!!! !
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!!
!!!!
!
!!
!!
!!
!!!!!!!
!!
!
!
!!!!
!
!!!!!!!
!
!!!!!!!!!!!
!!
!
!
!
!!!!!!!!
!!!!!!!!!!!!!
!
!!!!
!!
!
!!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!
!
!!!
!!!!
!
!
!
!!!
!
!
!!!!!!!!!!!!!!!!!!!!!!
!!
!
!
!
!!!!!
!!
!!
!
!!!
!!!!!!!!!!!!!!!!!!!!!
!
!
!!!!!
!
!
!
!
!
!
!!
!!!!!!!!!!!!!!!!!!!!!!!
!
!!!
!!
!
!
!!
!!!!!!!!!!!!!!!!!!!!!!!!
!
!!
!!!!!!!!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!
!
!!
!!!
!
!!!!!!!!!!!!
!
!
!!
!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!!!
!
!!
!!!
!!
!!!!!!!!!
!!
!
!!!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!!!
!
!
!!!!!!!
!
!
!!!!
!
!
!
!
!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!
!
!
!!!!!!
!
!
!!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!
!
!
!!
!!!
!
!
!!
!
!
!
!
!
!!
!!!!
!!!!!!!!!!!!!!
!
!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!
!!!!!!!!
!
!
!
!
!
!
!
!
!
!!!!!!!!!
!!!!!!!
!!!!!!!!!!!!!!
!
!!!
!
!
!
!
!
!
!
!
!!!!!!!!!!
!!!!!!!
!!!!!!!!
!!!!
!
!
!
!!!
!
!!
!
!
!!!
!
!
!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!
!
!
!
!
!
!
!!
!
!
!!!!!!!
!
!!!!!!!!!!!!!!!!
!
!!
!!!!!!!!!!!!!
!
!
!
!
!
!
!
!!
!!!!!!!!!!!!!!!!
!!!!
!
!!!
!!
!!
!
!
!
!
!
!
!
!
!
!
!!
!
!!!!!!!!!!!!!!!!
!!
!!
!!!
!!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!!!!
!!!!
!!!!
!!
!!!
!
!!!
!!
!
!
!
!
!!
!
!
!!
!!!!
!!!!!!!!
!!!
!!!!!!!!!
!
!!
!
!
!
!
!
!
!
!
!!!!!
!!!!!
!!!!!!
!!!
!!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!!!!!
!!!!!
!
!
!
!
!
!!
!!
!
!
!
!
!
!!
!!
!
!!
!
!
!
!!
!
!
!!!!!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!!!!!!
!
!
!!
!!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!!
!
!!!
!
!
!
!
!
!
!
!
!
!!!!!!!
!
!!!
!
!
!
!
!
!
!
!
!!!!!!
!
!
!
!
!!
!!!!!!!!
!
!!!
!!
!
!!
!
!!!!!
!
!
!
!!!!
!
!!!!!
!
!
!
!
!
!
!
!
!
!
!!!
!
!!
!!!
!!
!
!!!!!
!
!!!
!!!
!
!
!!
!!!!!!!
!
!!
!
!
!
!!!!!
!!
!
!!
!!!
!!!
!!!
!
!
!!!!!
!
!
!!
!
!!!
!!
!!!!!!!!
!
!!!
!
!
!
!
!!
!
!
!!!
!
!
!
!
!!!
!!
!!!!!
!
!
!
!
!
!
!
!!!!!!!!!
!
!!
!!
!
!
!
!
!
!!
!
!!
!!!!!!!
!
!
!
!
!
!!!
!!
!
!
!!
!!!!!!!
!
!
!
!
!
!
!
!
!!
!!
!
!!
!!!!!!!
!
!
!
!
!!!
!
!!!
!
!
!
!
!!
!
!
!
!!!!!!!!!
!!!!
!
!
!
!
!
!
!!!!!!!!
!!
!
!
!!
!
!
!
!
!!
!!!!!!!!!!
!!!!
!
!
!!
!
!
!
!!
!!!!
!
!
!
!!
!!!
!!
!
!
!!
!
!
!!
!!
!!!!!!!!!!!!
!!!!!!
!
!
!
!
!
!
!
!!!!!
!
!
!
!!!
!
!
!
!
!!!!!
!
!!
!
!!!
!
!!!!!!!!!!
!
!!!!
!
!!
!
!
!!
!
!
!
!
!
!
!
!!
!!!!!
!!
!!!
!
!
!!!!
!!!!!!!!
!!
!
!
!!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!
!
!!!
!
!!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!
!!!
!!
!!
!
!!
!!
!!!
!
!
!!
!!
!
!!
!
!!!
!!!
!!
!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!!
!!
!!!
!!!
!
!
!
!
!
!
!
!
!
!!
!!!
!
!
!
!
!
!
!
!!
!
!
!!
!!!!!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!!!
!
!
!
!
!
!
!
!!!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!!!!!!
!
!
!
!!
!!
!
!
!
!
!!!
!
!
!
!
!!
!
!
!
!
!!!!
!
!
!!
!
!!
!!!!
!
!
!
!
!
!!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!!!!!
!!!!!
!
!
!
!
!
!!
!
!
!
!!!
!
!
!
!
!
!
!!!
!!
!
!
!!
!!!
!
!!!
!
!
!
!
!
!!
!
!!!
!!!
!
!
!
!
!
!
!!
!
!!!!
!
!
!
!
!!
!
!
!
!!
!!!
!!
!
!
!
!
!!
!
!!!!
!
!
!
!!
!
!!!!!
!
!
!
!
!
!
!!!
!
!!
!
!
!
!
!!
!!!
!
!!
!!
!!
!
!
!!!!
!
!
!
!
!
!
!
!
!!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!!
!
!
!
!
!
!
!!
!
!
!!
!!
!
!!!
!
!!
!
!!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!!!
!
!
!!!
!
!!
!
!!
!!!
!
!
!
!
!!
!
!
!
!!!
!!
!!!!!
!
!
!
!
!
!
!
!
!
!!!
!
!
!!!!!!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!!!!!!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!!
!
!!!!!!!!
!
!
!
!
!
!!!!
!
!!
!!!!!!!!
!!
!
!
!
!
!
!
!
!
!!
!!
!!!!!
!!
!!!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!
!
!!
!!
!
!!
!
!
!
!
!
!
!!
!
!!!!!!!
!
!
!
!!
!
!
!
!
!
!!!
!
!
!
!!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!!!!!!!
!
!
!
!
!
!
!!
!!
!
!!
!
!
!
!
!
!!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!!!
!!
!
!
!!
!
!
!
!
!!
!
!
!
!
!!!
!
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!!
!
!
!
!
!!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!!!
!!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!!!!!
!
!!!
!!
!
!!
!!
!
!!
!
!
!!
!
!!!!
!
!
!
!!
!
!!
!
!!!
!
!
!!
!!
!!
!!!!!!!!!!
!
!
!
!!
!!
!
!
!
!
!!!
!
!
!
!!!!!
!
!!!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!!!!!
!
!
!
!
!!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!
!!
!
!
!
!
!!
!!!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!!!!
!
!
!
!!
!
!
!!!
!!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!!!
!
!!!!!!!!
!!
!!!!!!
!
!!!!!!
!!!!!!!
!!
!
!
!!
!!!!!
!!
!
!
!
!
!!!!
!
!!
!!
!!
!!
!!!!!!!!!
!
!!
!
!
!
!
!
!
!!
!
!
!!!
!
!
!
!
!!!!
!!!
!
!!
!!!!!!!!!!!!
!
!
!
!
!!
!
!!
!
!
!!
!
!
!
!
!
!
!!
!
!!
!!!!!
!!!
!
!
!
!
!
!
!
!
!
!!
!!!!
!!
!
!!
!
!!!
!!
!!!!!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!!
!
!
!!
!
!
!!!!!
!
!!!
!!
!!
!
!
!
!
!
!!!
!
!
!
!!
!
!
!
!!
!!
!
!
!!
!
!
!
!!
!!
!!
!!!!!!!!!!!!!!!!
!
!!!!!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!!!!!!
!
!!!!!!!
!!!!
!
!!!!!
!
!
!
!
!
!
!
!!!!!
!
!!!
!
!
!
!
!
!
!
!
!
!!!!!!!
!!
!!!!!!!!!
!!!
!!!!!!
!!!!!
!
!
!
!
!
!!!
!!
!
!
!
!!
!
!!!!
!!!!!!!!!!!!!!!!!
!!
!!!!
!!
!
!!!!!!!!
!!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!
!!!!!!!
!!!!!!
!
!!!
!!
!!!
!!
!
!!!!!!!!!!
!!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!!
!
!
!
!!
!!
!
!!!!!!!!!
!!!
!!!!!
!
!
!
!
!!
!
!
!
!
!
!!!!!!!!!!!!!
!!
!!!
!
!!!
!!!!!!!!!!!!!!!
!!!
!
!
!
!
!
!!
!!!
!
!
!
!!!
!
!
!
!
!
!!!!!!!!!!
!!!!!!!!
!!!
!
!!
!
!
!!!
!!
!
!
!!!!!!!!
!
!
!
!
!
!
!!
!
!!
!!!
!
!
!!
!
!
!!
!
!
!!
!
!!!!!!!!!!!
!!!!!!!
!!!!
!
!!!!
!!!
!
!
!!!!!!!!!!
!
!
!
!
!
!!
!!
!
!
!
!!
!!
!
!!!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!
!
!!!!!!!!
!!!!!!
!
!
!
!!!!!!!!
!
!
!
!
!
!
!
!!
!
!
!!
!
!!!!!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!
!!!!!!!!!!!
!
!!!!
!
!!
!!!!!
!
!!
!
!
!!!!!
!
!
!
!
!
!
!
!
!
!
!!!
!!!!!!!!!
!!!
!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!
!!
!!!
!!!!!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!
!!!!!!!!!!!
!!
!!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!
!
!!!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!
!!!!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!!!!!!!!!!!!!!!!!!!!!!!!
!
!!
!!!!!!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!!!!!!!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!!!
!!
!
!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!
!!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!!!!!!!!!!!
!
!
!
!
!!!!
!
!!
!
!
!
!
!
!!
!
!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!!!!!!!!!
!
!!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!
!!!!!!!!!!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!
!!!!
!!!!!!!!!!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!!!
!
!!
!!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!
!!!!!
!!!!!!!!!!
!
!
!
!
!
!
!!!!!
!
!
!!!!
!
!!
!
!
!
!
!
!!
!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!
!
!!
!
!
!
!!
!!!!!!!
!
!!!!
!!!!!
!!!
!
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!
!!
!
!
!
!!!!
!
!!
!
!
!
!!!!!
!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!
!!
!
!
!
!!!!
!
!
!
!
!!
!!!!!!!!!!!!
!!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!!
!
!
!
!
!
!!!
!
!
!!
!
!
!
!
!
!!!!!!!!!!
!!
!
!
!
!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
!!!!!!
!
!
!
!!!
!
!
!!!!!
!
!
!
!
!
!!
!
!
!
!
!
!!!!!!!!
!!!!
!
!!
!
!!!!
!!
!!!!!
!
!!!
!!!
!!!!!!!!!!!!!!!!!!
!!
!
!!!
!
!
!
!
!
!
!!
!!!
!
!
!
!!!!!!!!!!!!
!
!!
!
!
!
!!!!!
!!!
!!
!!!!!
!!!!!!!!!!!!!!!!
!
!
!
!!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!!!!!!!!!!!!!
!!
!
!!!!!
!!!!!!!!!
!
!!!
!
!!!!!!!!!!!!!
!!
!
!
!
!
!
!!
!!
!!
!
!
!
!
!!!!!!!!!!!!!!!!!
!
!
!
!
!!
!!!!!!!!!!!!!!
!
!!!!!
!!!!!!!!
!!!
!!!
!
!!
!
!
!
!
!
!
!!
!!
!!!!!!
!
!
!!!
!!!!!!!!!!!!!!
!
!!
!
!!!
!!!
!!!!
!
!
!!
!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!!!!!!!!!!
!
!
!!!!!
!!!!!!
!
!
!
!!
!!!!!!!!
!
!!!!!!!!!!!
!
!
!!!!
!
!
!
!!
!
!
!!
!!!!
!
!!
!!!
!
!
!
!
!
!
!!!!!!!
!
!!!
!!!!
!!
!!!!!!!!!!!!!!!
!!
!!!
!
!!
!!
!
!
!!
!
!
!
!
!
!
!!!
!
!!
!
!!
!!
!
!
!
!
!!!!!!
!
!!!!!
!
!
!!!!!!!!!!!!
!
!
!
!
!
!
!!!!!!!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!!!!
!
!!!!
!!!!!!!
!
!!!!!!!!!
!
!
!!
!!!!!!
!
!!
!
!
!
!
!
!!!
!
!!!!!!
!!
!!
!!
!
!
!!
!
!
!!!!!
!!
!!
!!!!!!!!!
!!
!!!!!!
!!!
!
!
!!
!
!!!!
!
!!!!
!
!
!
!
!
!
!
!
!!!
!
!
!!!
!
!!!!!!
!!
!!!!!!
!
!
!!!!!!!
!
!
!
!
!!
!
!!!!!
!!!!!
!!
!
!
!
!
!
!!!
!
!
!!
!
!
!!!!!!
!
!
!!!!!!!
!
!!!!!!!!!
!!!!!
!
!!
!
!
!
!
!!!!
!
!
!!
!
!
!
!!
!
!!
!!!!!!
!
!
!
!
!!
!!!!!!
!!
!
!
!!!!!!!!
!!!!!
!
!
!
!
!
!
!!!!!
!
!
!
!
!
!
!
!!!!
!
!
!
!!!!!
!!!
!
!
!!!!!!!!
!!!!
!
!!!!!
!
!
!
!!
!
!
!!!!
!
!!
!!
!!
!
!
!
!
!
!!
!
!
!
!!!!!!!!!
!!
!!!!!!!
!
!!!!!
!
!!
!
!
!
!!!
!
!!
!!!!
!!!!
!
!
!!!
!
!!
!
!!
!!!!!!
!
!!!!!!!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!!!!!
!
!!
!
!
!!
!!
!
!
!
!
!!!
!
!!!
!!!!
!!
!!!
!!!!
!
!
!
!!
!!
!
!
!
!!
!
!
!!!!
!
!!
!!!
!
!!!!
!
!
!!!
!!
!!!!
!!
!
!
!!
!
!!!
!!!!!!
!!!!!!!!!!!!
!!!!!
!
!!!!!
!
!
!
!
!
!
!
!!!!!!
!
!
!!!
!
!!!!
!
!!
!
!
!!
!
!!!!!
!
!
!
!
!
!!!!!
!!
!!!!!!
!!!
!!!!
!!!!
!!!
!
!
!!
!
!
!
!
!
!
!!!!!!!
!!!!
!
!
!!
!!!!
!
!
!
!
!
!!!!!!!
!
!!
!!
!!!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!!!!!
!!
!
!!!
!
!
!!!!!!
!!!!!!
!!
!!!!
!
!!
!
!
!
!
!!!
!
!
!
!
!!
!
!
!!
!!
!
!!!
!
!!!
!!
!!
!
!
!
!!!!!!!!
!!!!!!!!
!
!!
!!
!
!
!
!
!!
!!
!
!!!!!!!!
!
!
!
!
!!
!
!
!!
!
!!!
!
!!!!
!!
!!!!!!!!
!!!!!!
!
!!
!!
!
!
!!!
!
!
!!!
!!
!!!!
!
!
!
!
!
!
!
!!!
!
!
!!!
!!
!!!!
!
!
!!!!!!!!!!!!!!
!!!!!!!!
!!!
!!
!
!
!
!
!!!!!!
!
!
!
!
!
!
!
!!
!
!
!!
!!!!!!!!!!!!!!!!!!!!
!
!!!!!!!!
!!!!!!!
!!
!
!
!
!
!!!!!!!!
!
!
!
!
!
!
!
!
!
!!!!!!!!!!!!
!
!!!!!!!!
!!!!!!!!!!!!!!!
!
!
!
!
!
!
!
!!
!!
!
!!!!!!!!!
!
!
!
!!!
!!!
!
!
!
!
!
!!!!!!!!
!
!!!!!!
!!!!!!!!!!!!!!
!!!!!!
!
!
!
!
!
!
!!!
!
!!!
!!
!
!
!
!
!
!
!
!!
!
!!
!
!!
!!!!!!!!
!
!!!!!
!!!!!!!!!!!!!!
!!!!!!
!
!
!
!
!
!
!
!!
!!
!
!!
!
!!
!!
!!
!
!
!
!
!
!!
!!!!!!!
!!!!!!!!!
!!!!!!!!!!!!!!
!!!!!!!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!!
!
!!!
!
!
!
!!
!
!
!!!!!!!!!!!!!!!
!!!!!!!!
!!!!!!!!!!!!!!
!
!
!
!!!!!!
!
!!
!
!!
!
!
!
!!!!!!!!!!!!!!!!
!!!!!!!
!!!!!!!!!!!
!
!
!
!
!
!!!!!!
!
!
!
!
!!
!
!
!
!!
!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!
!
!
!
!
!!
!!!
!
!
!!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!
!!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!!
!
!!!!!!!!!!!!!!!!!!!
!
!!!!!!!!!!!!
!!
!!
!
!!
!
!
!!
!
!
!
!
!!
!!!!!!!
!!!!!!!
!
!!!!!!!!
!!
!!!!!
!
!
!
!
!!
!
!
!
!
!
!!!!!!
!!
!
!
!!!!!
!
!!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!
!!!!
!!!
!
!
!!!!!!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!!!!
!!!!!!!!!
!
!
!
!!
!
!
!
!
!
!!!
!
!
!
!!!!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!!
!
!
!!!!
!
!!
!
!
!
!
!
!
!!!
!!
!!!
!!!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!!!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!!!
!!!
!
!
!
!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!!
!
!!
!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!!
!
!
!
!
!!
!!
!
!
!
!!
!!
!
!
!
!
!
!
!
!!!
!!
!
!
!
!
!!
!
!!
!
!
!!
!
!!
!
!!
!
!
!
!!!
!
!!!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!!
!!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!!!!
!!
!
!
!!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!!
!
!
!!
!
!
!
!
!
!
!!
!
!!
!
!
!
!!
!
!
!
!
!
!!!!!
!
!
!
!
!
!
!!
!
!!!!
!
!
!
!
!
!
!
!
!
!!!
!
!!!
!
!!!
!
!
!!!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!!!!!
!
!
!
!
!
!
!
!
!!!
!!!!!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!!!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!
!
!
!
!!
!!
!
!
!!
!
!
!
!!!!
!!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!
!!
!
!!
!
!
!
!!!
!!
!
!
!
!!!!!
!
!
!
!
!
!
!
!
!
!
!!!!!!!!
!
!
!
!
!
!
!!
!!
!!!!!
!
!
!
!
!
!
!
!!
!
!!!
!
!
!!
!!
!
!
!
!
!!!
!
!!
!!!!!!!!
!
!
!
!!
!!
!
!!!!!
!!!!!!!!!!!!
!
!!
!
!
!
!
!
!
!
!!!!
!!
!!!!!
!!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!!!!!!
!
!
!
!
!
!!
!!
!!
!
!!!!!
!
!
!
!
!
!
!
!!!!!!!!!!
!
!
!
!
!
!
!!
!
!
!
!!!!
!
!
!
!
!
!
!
!!!!!
!
!
!
!!
!
!!
!!
!
!
!
!!!!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!
!
!!
!
!
!
!!!!!
!!!
!
!
!!
!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!!
!
!
!
!
!!!!
!
!
!!!!
!
!
!
!
!
!
!
!
!!!!
!
!!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!!
!
!
!!
!!
!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!!
!
!
!
!
!
!!!
!!
!!
!
!
!!
!
!!!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!!
!!!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!
!
!!!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!!
!
!!!!!
!
!!
!!
!
!
!!!
!!!
!
!
!
!
!
!
!
!
!!
!!!!
!
!!!!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!
!!!
!
!!
!
!
!
!
!!
!!!
!
!
!
!!
!
!
!
!!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!!
!
!
!
!
!!!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!
!!
!
!!!!!
!
!!
!
!
!
!
!
!
!
!
!
!!
!
!!
!
!
!!
!
!!
!
!
!
!!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!!
!!
!
!
!
!
!
!!
!
!!
!
!!
!
!
!!
!
!
!
!!!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!!!!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!!!!
!!
!
!!
!
!
!!
!
!
!!
!!
!
!
!
!
!
!
!
!
!
!!
!
!
!!
!
!
!
!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
! !
!
!
!
!
!!!
!!!
!
!
!
! !
!!!
!!!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
! !
!!!
!! !
!
!
!!!
!!
!
!
!!!
!
!
! ! !
!!!
Peak Ground VelocityLess than 0.1 cm/sec
0.1 - 1.1 cm/sec
1.1 - 3.4 cm/sec
3.4 - 8.1 cm/sec
8.1 - 16 cm/sec
16 - 31 cm/sec
31 - 60 cm/sec
State Boundaries! DYFI Response
Impacts of Pawnee, OK Earthquake 2016
Texas
Kansas
Oklahoma
Oklahoma City
Tulsa
Wichita
22
been exposed to ground motion, then the household is able to self-report the incident to the DYFI
database. The DYFI data points are represented by the tiny black dots in the figure. The dots
demonstrate the common misconception of earthquake event - that is, an earthquake is not like a
pebble that hits the water when dropped in a pond, which creates a uniform ripple effect that gets
weaker as it travels from the center. The earth's surface is not uniform, so during an earthquake
event, one area can experience over ten times the effects as a neighboring area that is the same
distance from the epicenter (National Earthquake Hazards Reduction Program, 2013).
As such, the self-reported measures demonstrate a tremendous amount of heterogeneity of
perceived responses to ground motion, despite the fact that the measured areal extent (based on
the peak ground velocity estimates) offered within the same figure suggest that the ground
motion should be approximately the same for all residents of the Oklahoma City MSA.
Household subjectivity offers part of the explanation; however, the difference in perceived
responses can also be driven by proximity to (other) known faults, property soil type, and general
height (or elevation) of residency. In other words, a person living in a home, built on loose soil,
may perceive much more lateral ground movement than a person living in a similar home built
on clay soil. As household perceptions of earthquake activities are subjective and vary by a large
degree, we posit that creating an experimental design, in this particular context, would be
challenging to implement an estimate.
As an alternative to a quasi-experimental framework, we take a slightly different
approach. To wit, we consider numerous intervening factors in order to empirically link housing
prices with induced earthquake activities. These intervening or conditional factors consist of
various measures of wastewater well injections, oil and gas production indicators, and housing
23
and neighborhood characteristics, among other covariates. To see this, consider the directed
acyclic graph in Figure 9.
Figure 9. Potential causal pathways from unconventional gas development to residential real estate impacts
Notes: The variables within the figure represent the following: X1 denotes oil and gas development; X2 is produced wastewater; X3 is wastewater disposal; X4 denotes induced earthquakes; X5 are household or neighborhood characteristics; Y represents housing prices; and, Ui denotes any unobserved or random factor acting on the ith observed variable.
The variables in Figure 9 represent the following: X1 denotes oil and gas development; X2
is produced wastewater; X3 is wastewater disposal; X4 denotes induced earthquakes; X5 is
household or neighborhood characteristics; Y represents housing prices; and, Ui denotes any
unobserved or random factor acting on the ith observed variable. The pointed arrows demonstrate
the posited direction of causality. The astute reader will discern two potential causal pathways:
one goes directly from oil and gas development (X1) to housing prices (Y); the other pathway
goes from development (X1) through a series of intervening effects (X2 – X4) and finally to
housing prices (Y). We describe the former pathway as the direct link and latter pathway as the
indirect link between oil and gas development and housing price impacts. (Another potential
pathway goes directly from wastewater disposal (X3) to housing prices (Y)). These alternative
X1: Unconventional oil and gas development X2: Produced wastewaterX3: Wastewater disposal X4: Induced earthquake activityX5: Neighborhood and household characteristics Y: Housing pricesUi: Unobserved factors
Causal Pathways:Direct link: from X1 to Y, Indirect link: from X1 to X2 to X3 to X4 to Y.
X1 X2 X3 X4
Y
U1U2 U3 U4
UY
X5
U5
24
pathways are there to illustrate the complexity of attributing causality from oil and gas
development to housing prices.
In order to obtain credible estimates of the impacts of induced earthquakes on housing
prices, in an OLS framework, then any regression of the former on the latter would have to
satisfy the conditional independence assumption. That is, earthquakes activities, conditional on
the other observed factors, must be independent of housing prices or {Y ∐ X4 | X1, X2, X3, X5}.
The conditional independence assumption is sometimes referred to as “unconfoundedness”
(Imbens, 2014).
Our specified regression is similar to that of Koster and van Ommeren (2015). That is, let
𝑝#$ denote the natural log of housing price i at time t. The basic equation to be estimated is
represented as:
𝑝#$ = 𝛼𝑒#$ + 𝒙𝒊𝒕, 𝜷 + 𝜇# + 𝜃$ +𝑢#$, (1)
where 𝑒#$denotes a measure of earthquake activity; 𝑥#$ denotes a vector of explanatory variables
(including a constant term) and 𝜷 is the corresponding vector of estimated coefficients; 𝜇#
denotes a Census-tract-level fixed effect; and, 𝜃$ indicates a trend term. The Census tract fixed
effects help control for shocks that would affect an entire Census tract. We include the trend
term to de-trend home sale prices due to fluctuations in the real estate market. Cameron and
Miller (2015) argue that is generally best practice to cluster at the highest level of potential
correlation in the data. In our case, the highest level of potential correlation is the county;
however, since our analysis focuses on one county, this would result in the situation with only
one cluster in which case clustering is not necessary. The next highest level of potential
25
correlation in our data is the city. Clustering at the city level leads to two potential issues which
may lead to incorrect inference: 1) there are only 18 clusters; and 2) the clusters are heavily
unbalanced; therefore, we decide to eschew the use of cluster-robust standard errors and use
White-Huber standard errors.
Primary interest lies within the estimation and interpretation of 𝛼, the partial effect of an
induced earthquake on (the log of) housing prices. Consistent with the conditional independence
assumption discussed above, equation (1) is specified such that the partial effect, of earthquakes
on housing prices, is conditional on the fixed effects terms and the vector of other explanatory
variables, which include oil and natural gas production indicators, the amount of wastewater
injection, and housing and neighborhood characteristics. A violation of conditional independence
assumption is simply another way to express, in terms of an OLS regression approach, a
violation of the zero-conditional mean assumption for equation (1):
𝐸(𝑢#$|𝑒#$, 𝒙𝒊𝒕) ≠ 0. (2)
On the other hand, if the zero-conditional mean assumption were to hold, then it would imply an
earthquake event should exogenously determine an effect on housing prices. However, as
discussed above, larger magnitude (induced) earthquake activities, 𝑒#$, are very likely to be
endogenous to housing prices. In other words, an experienced (moderate-to-larger scale)
earthquake event is likely to be correlated with the error term (𝑢#$), thus generating biased effects
estimates.
Somewhat similar to Koster and van Ommeren (2015), we use instruments to estimate the
impacts associated with the moderate sized to larger magnitude seismic events (i.e., earthquakes
26
of magnitude 3.0 or larger), which are more likely to create ground motion and be felt by
humans. That is, we generate unbiased earthquake effects estimates by exploiting random
variation in the underlying geological endowment (i.e., the sedimentary basin) interacted with
natural gas prices (i.e., the average annual Henry Hub spot price). Generally speaking, any
earthquakes of magnitude less than 3.0 are not detected by humans without specialized
equipment (U.S. Geological Survey, 2017b). Keranen et al. (2014) explain how to distance to a
known fault and soil type help explain the induced earthquake activity. The geological
endowment (interacted with natural gas prices) instrument is derived from Weber et al. (2016),
which is used to determine if the underlying geology, below a specific area, contains any known
conventional or unconventional oil or gas (in situ) formations. The latter instrument is used to
assess whether an injection well or drilling activities are likely to occur in close proximity to a
particular property.
Using identified instruments, we conduct a two-stage regression in which we run a
separate auxiliary regression of the instruments on the instrumented variable as:
𝑒#$ = 𝜋< + 𝜋=𝑧#$ + 𝑣#$, (3)
where 𝑧#$ denotes a vector of instruments; and as before 𝑒#$ denotes an earthquake measure as
reflected in a self-reported DYFI observation.
To serve as a valid instrument, the exogenous variables, 𝑧#$, must be statistically related
to the endogenous variable, 𝑒#$, they are instrumented for: Cov(𝑒#$, 𝑧#$) ≠ 0. This condition is
often referred to as instrument relevance (Imbens, 2014). Our proposed instruments should be
relevant in that the increased volume in wastewater injections has led to a swarm of seismic
27
activity, including earthquakes occurring close to known faulty (Keranen et al., 2014). As found
within the past literature, the increase in seismicity is expected to lead to a decrease in the value
of a proximate residential property. However, in order for the instrument to generate unbiased
effects estimates, the instrument must also be exogenous. The exogeneity condition implies that
the covariance between the instruments, 𝑧#$, and the error term, 𝑢#$, should equal zero; i.e.,
Cov(𝑧#$, 𝑢#$) = 0.
The exogeneity condition can be further broken down into two assumptions. One, the
instrument is randomly defined; and, two, the instrument only affects housing prices through the
stated channel. The second assumption is sometimes referred to as the exclusion restriction
(Imbens, 2014).
To demonstrate why an instrumental variables approach is needed to estimate the
earthquake impacts in the context of Oklahoma City, we provide maps in Figure 10, which
demonstrate the areal extent of the earthquake activities in the MSA. The top panel shows the
cumulative earthquake activities from 1980 to 1999, whereas the bottom panel illustrates the
cumulative activities from 2000 to 2016. The small circular symbols (darker color with black
border) represent disposal wells, whereas the larger circular symbols (lighter color with white
border) represent an individual earthquake event. The larger in circumference the earthquake
symbol, the higher the magnitude of the particular event. Upon examination of Figure 10, one
will see few earthquake events, which appear randomly distributed over space, for the 1980-to-
1999 period. However, the 2000-to-2016 period demonstrates far more earthquake events, and
the events appear to be concentrated within the eastern and northern portions of the MSA. As
expected, the earthquake activity appears to coincide with where the disposal wells are
concentrated within the MSA.
28
Figure 10. Oklahoma Earthquake Activities 1980-2009 and 2010-2016
(a)1980-2009
(b) 2010-2016 Notes: Smaller circles (with black border) denote injection wells. The larger circles (with
white border) denote an individual earthquake event. The larger in circumference (and darker hue) the earthquake symbol, the higher in magnitude the earthquake event.
Source: Oklahoma Office of the Secretary of Energy and Environment (2017).
29
Based on the concentration or clustering of earthquake activities, examined in Figure 10,
an OLS approach would arguably yield biased estimates of the earthquake impacts, as the
earthquake events are not randomly distributed over space. Moreover, an OLS approach with
cluster-robust standard errors would potentially alleviate the bias within the standard error
estimates; however, cluster-robust standard errors would not necessarily address the potential
endogeneity bias associated with location-specific earthquake events and property values. Thus,
we argue that an instrumental variables approach, in which we instrument on the induced
earthquake events, will provide more accurate impact estimates.
It seems plausible that these three instruments would offer exogenous variation in
explaining the impacts associated with earthquakes; however, these instruments may violate the
exclusion restriction. It is for this reason that Imbens (2014) describes the exclusion restriction
as the most critical, but also the most controversial assumption underlying instrumental variables
methods. An example of an alternative channel, other than the effect of earthquakes directly, is
that higher natural gas (or crude oil) prices may induce more drilling, which in turn stimulates
the local economy and adds value to local real estate properties. Weber et al. (2016) found that
unconventional gas production impacted housing prices through an effect on the local tax base.
Another potential channel could stem from the manner in which drilling activities, and the
consequent earthquake activities, are covered through local and national media outlets, which in
turn could affect households’ perceptions of property values. Although we do not empirically
address media coverage, Liu et al. (2016) explore how household perceptions, of current and
future earthquake activities, may affect housing prices in the Oklahoma City region. They argue
that the perceptions of local households, following the 5.7-magnitude Prague earthquake in 2011,
30
led to an average 2.2 percent (or about $4.4 thousand loss per property) decrease in property
values.
As household perceptions have already been explored in Liu et al. (2016), we focus on a
slightly different channel, which we describe as the disamenity effect. That is, perhaps it is not
earthquakes but rather local drilling and disposal disamenities that are driving down home
values. In other words, drilling rigs and injection wells create (visual or noise) disamenities that
reduce people’s willingness to live in a particular neighborhood. This motivates us to control for
the number of drilling wells and injection wells per square kilometer within a local Census tract.
To capture the effect of persistent disamenities, we also control for the number of wells drilled
(and injection wells) per square kilometer in all years prior to home’s assessed date.
However, in controlling for the number of drilling and injection wells, we must
acknowledge that we may have created another problem in that the number of wells may be
endogenous to housing values. We therefore also estimate equation (3), the first stage IV
regression, both with and without the drilling and injection well density variables. The predicted
values from the first stage regression are then used to estimate equation (1) in order to produce
unbiased estimates of the effect of earthquakes on housing prices.
31
4 DATA
Table 1. Summary statistics
4.1 Home sales data
We acquired real transactions and home characteristic data from the Oklahoma County Tax
Assessor’s office. The data includes all real estate parcels and characteristics (e.g. liveable area,
bedrooms, baths, year built, lot size, etc.) as of February 2017 and each transaction (date, price,
sale validity, grantor, grantee) for every parcel dating back to 2000. We cleaned the data to
represent fair market value, arms lengths transactions of single-family residential homes. Our
Variable Mean Std. Dev. Min Max UnitsSale Price 150,517 94,657 23,000 670,000 Nominal Sale PriceSquare Feet 1.81 0.74 0.72 10.08 Thousands of Square FeetBasement Square Feet 0.01 0.11 0 3.15 Thousands of Square FeetGarage Square Feet 0.47 0.23 0 5.77 Thousands of Square FeetCarport Square Feet 0.00 0.04 0 1.67 Thousands of Square FeetBedrooms 3.11 0.63 1 5 CountBaths 1.97 0.66 0.75 4 CountAge 30.57 24.30 0 122 AgeRemodeled 0.39 0.49 0 1 Indicates if the home is remodeled at time of saleLot Size 0.32 0.53 0.05 7.00 AcresPost 2008 0.40 0.49 0 1 Indicates if the sale offer after 2008Wells, 1.5 km 1.07 1.62 0 11 Number of wells within 1.5 kmWells 20 km 299.71 138.90 20 755 Number of wells within 20 kmGround water x Wells, 1.5 km 0.11 0.57 0 8 Ground water × wells within 1.5 kmGround water x Wells, 20 km 24.61 83.16 0 587 Ground water × wells within 20 kmDistance to Okahoma City 9.12 4.01 1.03 25.30 Distance to downtown Oklahoma city, miles
Variable Mean Std. Dev. Min Max Unitseit 0.35 1.01 0 11 Number of magnitude 3.0 or greater earthquakes at time of sale with PGV > 0.5 cm/szit 0.04 0.49 0 49 Number of earthquakes within 1 km, at time of sale with a PGV less than 0.5 cm/secCumulative peak ground velocity 0.78 2.03 0 28.52 Centimeters per second (cm/sec)Earthquakes within 5 km 1.04 6.69 0 203 Number of earthquakes that occurred within 5 km at time of saleEarthquakes within 10 km 5.63 24.90 0 379 Number of earthquakes that occurred within 10 km at time of saleCumulative CDI Value 150.22 319.66 0 2,094.45 Cumulative CDI within 10 km at time of saleNatural Gas Prices 5.23 1.44 3.11 7.01 Dollars, 10 year moving averageWithin 5 KM 0.49 0 1 Within 5 km of any earthquake epicenter regardless of time of saleWithin 5 KM × Post 2008 0.19 0 1 Within 5 km of any earthquake, interacted with post 2008 periodSedmintary basin one 0.018 0 7.01 Indicates Anadarko sedimentary basinSedmintary basin two 0.042 0 7.01 Indicates Cherokee platform sedimentary basinSedmintary basin three 0.056 0 1 Indicates no sedimentary basin
Panel A: Home Sale Characteristics
Note: There are 152,522 home sale transactions within the sample.
Panel B: Earthquake Characteristics
32
final data set consists of 152,522 transactions spread across 99,305 unique parcels. As displayed
in Table 1, the average property sales price was approximately US$150,500.
We geocoded the residential parcels using data from the U.S. Census Bureau and the
Oklahoma City GIS Department. Using the geocoded data, we assigned each parcel to a 2010
census tract. We also calculated the straight-line distance between each parcel’s geocoded
location and downtown Oklahoma City. To control for nearby oil and gas production and
underground injection control (UIC) (dis)amenities, we followed Muehlenbachs et al. (2015).
More specifically, we calculated the number of operating oil, gas, and injection wells within 1.5
kilometers (km) and 20 km at the time of sale. Further, we interacted each wells variables with a
groundwater indicator variable.
4.2 Earthquake Measures
We collected earthquake data from the United States Geological Survey (USGS) for every
earthquake that occurred within 50 miles of Central Oklahoma or an earthquake that had Did
You Feel It (DYFI) Data reported in Central Oklahoma since 2000. These search parameters
yielded 5,715 earthquakes. For each earthquake, we received the following information: (1) the
time; (2) the location (longitude and latitude) of the epicenter; (3) the magnitude of the event; (4)
the depth; and, (5) the number of Did You Feel It Responses. For each home sale, we calculated
the number of earthquakes within a distance range (5 km and 10 km) of the geocoded address
that occurred between January 1, 2000, and the time of sale. We also calculated the peak ground
velocity (PGV) by using the attenuation function offered by Atkinson (2015). We discarded
observations where the peak ground velocity is less than 0.1 centimeters per second. Using the
estimated peak ground velocity, we calculated three variables. First, we calculated the
33
cumulative peak ground velocity (CPGV) at the time of sale. The CPGV variable, with PGV
larger than 0.1 centimeters per second, represents the summation of all peak ground velocities
generated by earthquakes that occurred between January 1, 2000 and the time of sale. Second,
we calculated the number of earthquakes (zit) that occurred within 1 km, before the time of sale,
and have a PGV of fewer than 0.5 centimeters per second. Finally, we calculated the number of
3.0-or-higher magnitude earthquakes that occurred before the time of sale and have a peak
ground velocity greater than or equal to 0.5 centimeters per second (eit).
Our primary measure of earthquake intensity is the USGS Did You Feel It data. The
DYFI data is collected through a self-reported survey after earthquake events. The survey
responses are aggregated by the U.S. Geological Survey (2017c), which in turn creates an
earthquake intensity index referred to as the Community Decimal Intensity (CDI). (A brief
outline of the CDI index algorithm is offered in the appendix). We obtained 2,496 unique
earthquake events within the DYFI data. From the USGS website, we downloaded the DYFI data
average over square km rectangles. To assign the DYFI data to some sales, we used the
following procedure. First, we divided Oklahoma County into square km blocks (UTM zones).
Second, we calculated the average CDI value for each earthquake and for each UTM zone using
all DYFI responses within 10 km. Finally, we calculated the aggregate CDI value at the time of
sale for each home sale by using all earthquakes occurring prior to the sale that was recorded for
the location’s UTM zone.
34
5 RESULTS
5.1 Induced Earthquake Impact Estimates
The regressions for the (induced) earthquake impacts to housing prices are provided in Table 2.
The base regression results are provided in the columns labeled as (1) - (3) in the table. The
estimates in columns (1) and (2) include a measure of induced earthquake impacts simply based
on the number of earthquake events (greater than or equal to 3 magnitude event) within a 5-
kilometer and 10-kilometer radius of the epicenter of the event. Based on 5-kilometer count
measure of seismicity, column (1) suggests that for every additional (annual) earthquake event,
property values decreased, on average, by 0.12 percent. The 10-kilometer county measure
implies that for every additional (annual) earthquake event, property valued decreased by an
average 0.04 percent. As we expected, the magnitude of the negative impacts declined (in
absolute terms) as the seismic force moves out in distance away from the epicenter. Put different,
these measures are consistent with tectonic models of seismic activity, in which the lateral force
is assumed to be strongest at or near the epicenter, but the force becomes attenuated as we move
out in distance away from the epicenter.
These particular measures of (induced) earthquake activities, offered in columns (1) and
(2), are potentially problematic for a host of reasons, but most importantly this measure ignores
inner variability in the magnitude of earthquake events. In other words, this measure assumes the
impacts will be identical whether a household experiences numerous magnitude 5 events versus
a household that experiences numerous magnitude 3 events. As such, we expected the base
impact estimates to suffer from measurement error and attenuation bias, and thus to be
downwardly biased.
35
The estimates in the column labeled as (3) provide a measure of earthquakes based on the
attenuation function, which was used to calculate the peak ground velocity of a particular
earthquake event. Similar to the results in column (2), the impact estimates (associated with the
calculated PGV of an earthquake event) suggest that an additional earthquake event led to an
average 0.82 percent decrease in property values. As outlined in Koster and van Ommeren
(2015), the attenuation function measurement approach should provide a more accurate
assessment of the areal extent of a particular earthquake. The magnitude of the impact estimate,
based on PGV measure, is greater than the former two equations as the column (3) estimates are
based on the cumulative effects of the earthquake events. In other words, column (3) examines
the aggregate number of earthquake events through time, whereas columns (1) and (2) only
examines the impacts based on the number of (induced) earthquake events that occur within a
particular year.
We provide additional impact assessments in columns (4) – (7) of Table 2. Our estimated
negative impacts, associated with earthquakes, are all consistent with the existing literature’s
range of approximately 2-10 percent. Column (4) offers a measure of induced seismicity by
examining the aggregate number of earthquake events based on the “Did You Feel It” (or DYFI)
data and estimated intensity index (i.e., the Community Decimal Index or CDI). The impact of
induced earthquakes, in column (4), suggests that every additional (cumulative) earthquake
exerts an average 4.4 decrease in property values. As noted in the preceding sections above, we
expected the estimate in column (4) to be upwardly biased (in absolute terms) due to the
(induced) earthquakes not occurring randomly through space. Column (5) replicates the IV
estimation method as outlined in Koster and van Ommeren (2015). According to the Koster and
van Ommeren IV estimation procedure, an additional earthquake leads to a 2.9 percent decrease,
36
on average, in property values. The results for our preferred specification, the instrumental
variables approach based on the DYFI observations, are provided in the column labeled as (6).
Similar to the preceding column, the column (6) estimate implies an average 2.4 decrease in
property values associated with an additional induced earthquake event.
Column (7) offers the difference-in-differences estimation approach of earthquake
impacts based once again on the DYFI observations. The column (7) estimate suggests that an
additional (induced) earthquake event led to a 3.95 decrease in property values on average. We
expected the difference-in-differences impact estimate, in column (7), to be biased due to the
violation of the SUTVA assumption. Put differently, given the geographically dispersed nature
of earthquake events, coupled with the variance in magnitudes of said events, it would be
challenging to define a stable treatment and control group. Further, the mismeasurement of
seismic impacts could potentially lead to biased average treatment effects estimates due to
contamination in treatment and control assignment. We demonstrated, with the set of two
equations in the Introduction, that seismic activity is likely to be endogenous with housing prices
due to the proximity of injection well site locations. Therefore, we conclude that the difference-
in-differences estimates are upwardly biased (in absolute terms) due to omitted variable bias.
As provided in Table 1, the mean sales price (during our period of observation) for a
property in Oklahoma City was US$150,517. Therefore, our preferred negative impact estimate
of 2.41 percent, provided in column (6) of Table 2, suggests that property values, on average,
would decline by approximately US$3,635 as a result of the swarm in seismic activities. As we
observed approximately 152.5K housing transactions, this average estimated impact suggests
that the MSA suffered approximately US$554 million (152,522 × $3,635) in aggregate short- to
medium-term damages.
37
Table 2. Induced Earthquake Impact Estimates
An approximate 4 percent decline, found within the difference-in-differences method of
column (7) of Table 2, would imply a short-run decline of US$5,950 in property value since the
swarm of activity began. This 4 percent estimated coefficient estimate is nearly twice as large as
our preferred estimate of 2.41 percent or Koster and van Ommeren’s (2015) estimated 1.9
(1) (2) (3) (4) (5) (6) (7)
Earthquake measureEarthquake Count
(5 km)Earthquake Count
(10 km)
Cumulative peak ground velocity
(PGV ≥ 0.5 cm/sec)Cumulative CDI
IV – Number of earthquakes (PGV ≥
0.5 cm/sec)IV - Cumulative CDI DID
Dependent variable Log of sale price Log of sale price Log of sale price Log of sale price Log of sale price Log of sale price Log of sale price
Earthquake impact -0.0012*** -0.0004*** -0.0082*** -0.0436*** -0.0290*** -0.0241*** -0.0395***
(0.0001) (0.0000) (0.0004) (0.0009) (0.0039) (0.0086) (0.0027)
Within 5 km of earthquake epicenter 0.0292***
(0.0052)
Square feet 0.3460*** 0.3460*** 0.3460*** 0.3460*** 0.3470*** 0.3460*** 0.3470***
(0.0029) (0.0029) (0.0029) (0.0029) (0.0029) (0.0029) (0.0029)
Basement square feet 0.2460*** 0.2460*** 0.2460*** 0.2450*** 0.2450*** 0.2450*** 0.2460***
(0.0116) (0.0117) (0.0117) (0.0116) (0.0117) (0.0116) (0.0117)
Garage square feet 0.1460*** 0.1460*** 0.1460*** 0.1460*** 0.1460*** 0.1460*** 0.1460***
(0.0051) (0.0051) (0.0051) (0.0051) (0.0051) (0.0051) (0.0051)
Carport square feet 0.0227 0.023 0.023 0.0245 0.0225 0.024 0.023
(0.0211) (0.0211) (0.0211) (0.0210) (0.0211) (0.0210) (0.0211)
Bedrooms 0.0019 0.0018 0.0018 0.0023 0.0017 0.0021 0.0017
(0.0017) (0.0017) (0.0017) (0.0017) (0.0017) (0.0017) (0.0017)
Baths 0.0718*** 0.0717*** 0.0715*** 0.0719*** 0.0716*** 0.0719*** 0.0716***
(0.0024) (0.0024) (0.0024) (0.0024) (0.0024) (0.0024) (0.0024)
Age -0.0100*** -0.0010*** -0.0099*** -0.0099*** -0.0097*** -0.0100*** -0.0100***
(0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002)
Square of age 4.3e-05*** 4.3e-05*** 4.1e-05*** 4.3e-05*** 3.9e-05*** 4.3e-05*** 4.2e-05***
-1.9E-06 -1.9E-06 -1.9E-06 -1.9E-06 -2.0E-06 -1.9E-06 -1.9E-06
Remodeled 0.0659*** 0.0658*** 0.0656*** 0.0657*** 0.0650*** 0.0659*** 0.0664***
(0.0014) (0.0014) (0.0014) (0.0014) (0.0014) (0.0014) (0.0014)
Log of lot size 0.0644*** 0.0644*** 0.0645*** 0.0649*** 0.0643*** 0.0646*** 0.0643***
(0.0022) (0.0022) (0.0022) (0.0022) (0.0022) (0.0022) (0.0022)
Post 2008 -0.1160*** -0.1170*** -0.1190*** 0.0164*** -0.1190*** -0.0425 -0.0949***
(0.0023) (0.0023) (0.0023) (0.0031) (0.0023) (0.0261) (0.0026)
Wells (1.5 km) -0.0055*** -0.0058*** -0.0065*** -0.0079*** -0.0080*** -0.0068*** -0.0061***
(0.0008) (0.0008) (0.0008) (0.0007) (0.0008) (0.0009) (0.0008)
Wells (20 km) 0.0005*** 0.0005*** 0.0004*** 0.0002*** 0.0004*** 0.0003*** 0.0005***
-7.8E-06 -7.7E-06 -7.8E-06 -9.5E-06 -7.7E-06 -5.4E-05 -8.3E-06
Ground water × wells (1.5 km) 0.0005 0.0006 0.0008 -0.0006 0.0007 -0.0002 -0.0009(0.0021) (0.0021) (0.0021) (0.0021) (0.0021) (0.0021) (0.0021)
Ground water × wells (20 km) -2.9e-05** -2.8e-05** -3.4e-05*** -3.5e-05*** -3.7e-05*** -2.9e-05** -2.4e-05*(1.3E-05) (1.3E-05) (1.3E-05) (1.3E-05) (1.3E-05) (1.3E-05) (1.3E-05)
Distance to Oklahoma City -0.0021 -0.002 -0.0019 -0.0031 -0.0023 -0.0027 -0.0024
(0.0020) (0.0020) (0.0020) (0.0020) (0.0020) (0.0020) (0.0020)
Constant 10.2400 10.2400 10.2400 10.3000 9.8160*** 9.8260*** 10.23
(35.1900) (37.9100) (23.0200) (0.0663) (0.0668)
Observations 152,522 152,522 152,522 152,522 152,522 152,522 152,522
R-squared 0.827 0.827 0.828 0.829 0.827 0.829 0.827
Adjusted R-squared 0.827 0.827 0.827 0.829 0.827 0.829 0.827
Notes: Each regression also includes a time trend and 238 individual census tract fixed effect estimates, which have been omitted for the sake of exposition. Robust standard errors in parentheses. The asterisk symobls denote the following: *** p < 0.01, ** p < 0.05, * p < 0.1.
38
percent. In terms of total impacts to the MSA, if we scale the estimated 4 percent by the total
number of sales transactions, then we can get an approximation of the total impacts, which is
US$907 million ($5950 × 152,522 sales transactions). This estimated total impact is nearly
double the approximation based upon our preferred, average impact measure of 2.41, which
translates into an approximate total impact of US$554 million ($3635 × 152,522 sales
transactions). Therefore, we do not lend the difference-in-differences estimate credence and
instead simply provide it here to help frame (as a credible upper bound) our entire scope of
impact estimates.
As a robustness check for the IV estimates (offered in columns (5) and (6) of Table 2),
we offer the first-stage regressions (of the two-stage IV estimation procedure) in Table 3. More
specifically, the second column in Table 3 corresponds with the regression results labeled as
column (5) in Table 2, and the third column in Table 3 corresponds with the regression results
labeled as column (6) in Table 2. The results in the second column of Table 3 suggest that the
lower magnitude (induced) earthquakes (i.e., earthquakes with peak ground velocities that are
measured as less than 0.5 centimeters-per-second) are highly correlated to the higher magnitude
(induced) earthquake events (i.e., earthquakes with peak ground velocities that are measured as
greater than or equal to 0.5 centimeters-per-second).1 (This correlation between lower magnitude
and higher magnitude seismic events is the instrument used by Koster and van Ommeren
(2015)). More specifically, the results in the second column imply that an addition lower
magnitude (induced) earthquake event is associated with a 0.33 increase in higher magnitude
(induced) earthquake, and the relationship is highly statistically significant. Otherwise, we
1 The reader will recall that a higher magnitude (PGV ≥ 0.5 cm/sec) seismic event is generally detected by humans as such events create lateral ground movement; whereas, the lower magnitude events (PGV < 0.5 cm/sec) are not detected by humans, but rather through seismic instrumentation.
39
include a host of covariates as instruments for the higher magnitude (induced) quakes, including
the number of drilled wells within a 1.5 and 20-kilometer radius of the epicenter of a particular
earthquake event.
The third column of Table 3 provides the first-stage regression results for our preferred
IV estimation procedure. As discussed above, we used the primary sedimentary basins
(underlying the Oklahoma City MSA) as instruments for the higher magnitude earthquakes. The
sedimentary basins offer information for the proclivity of an earthquake event given the
underlying geological profile within the basin. As illustrated in the table, the cumulative CDI
measures earthquakes were more likely to occur within the Cherokee platform basin and less
likely to occur within the Anadarko basin. Specifically, the instruments indicate that if an
observation occurred within the Anadarko basin, then there was, on average, 0.09 decrease in the
cumulative Community Decimal Index measure of (induced) earthquakes within that basin.
Conversely, if an observation occurred within the Cherokee platform basin, then there was, on
average, 0.06 increase in the cumulative Community Decimal Index measure of (induced)
earthquakes. Both instruments are highly statistically significant, which suggests that the
instruments are relevant (as discussed in Section 3). (We discuss the exclusion restriction
assumption in greater detail below). Otherwise, we also use the same set of covariates (as in the
second column of Table 3) as instruments for the CDI measure of (induced) earthquakes.
40
Table 3. First-stage regression results for two different seismicity measures
Dependent variable
Number of earthquakes (PGV ≥ 0.5 cm/sec) Cumulative CDI
Number of earthquakes (PGV < 0.5 cm/sec) 0.3333***(0.0231)
Sedimentary basin one (Anadarko basin) -0.0894***(0.0046)
Sedimentary basin two (Cherokee platform basin) 0.0611***(0.0018)
Square feet 0.0245*** -0.0181***(0.0068) (0.0050)
Basement square feet -0.0176 -0.0224(0.0248) (0.0147)
Garage square feet 0.0033 0.0109(0.0136) (0.0097)
Carport square feet -0.0187 0.032(0.0479) (0.0375)
Bedrooms -0.0075* 0.0119***(0.0045) (0.0035)
Baths -0.0101 0.0029(0.0068) (0.0051)
Age 0.0136*** 0.0029***(0.0004) (0.0003)
Age squared -0.0002*** -1.5E-05***(5.6E-06) (3.9E-06)
Remodeled -0.0359*** -0.0083**(0.0056) (0.0042)
Log of lot size -0.0017 0.0159***(0.0065) (0.0046)
Post 2008 -0.1300*** 3.0054***(0.0068) (0.0088)
Wells (1.5 km) -0.0897*** -0.0556***(0.0020) (0.0017)
Wells (20 km) -1.8E-05 -0.0065***(3.1E-05) (3.1E-05)
Groundwater × wells (1.5 km) 0.0154*** -0.0236***(0.0048) (0.0043)
Groundwater × wells (20 km) -0.0004*** -0.0031***(4.0E-05) (3.2E-05)
Distance to Oklahoma City -0.0098* -0.0178***(0.0055) (0.0037)
Temporal trend 0.0004*** 0.0005***(3.0E-06) (2.6E-06)
Constant 0.7555*** 0.8432***(0.1969) -0.1111
F-statistic 207.58 770.83Observations 152,522 152,522Notes: Each regression also includes 238 individual census tract fixed effect estimates, which have been omitted for the sake of exposition. Robust standard errors in parentheses. The asterisk symobls denote the following: *** p < 0.01, ** p < 0.05, * p < 0.1.
41
In addition to the two-stage regressions, we also conducted a series of under- and over-
identification tests, which are provided in Table 4. The null hypothesis for the Kleibergen-Paap
rank test is that the specification is under-identified. Intuitively, the rank test ensures (within the
auxiliary regression or equation (3) above) that at least one instrument provides exogenous
variation that is excluded from the main regression (equation (2) above) of interest. The rank test
is strongly rejected implying that our instrumental variables approach is not under-identified.
Further, the barrage of weak identification tests (Cragg-Donald Wald F statistic and the
Kleibergen-Paap rank Wald F statistic) imply that our instruments are not weakly correlated with
the endogenous variable. This result stems from the largely estimated F statistics, reported in the
7th and 8th rows of
Table 4, which greatly exceed the critical values listed in rows 10 through 13. In short, the weak
identification tests suggest that our instruments are valid. Finally, the null hypothesis Hansen J
statistic test is rejected, suggesting that our instruments are over-identified. The over-
identification condition requires that the auxiliary regression (or equation (3) above) contain
more instruments than instrumented variables. A rejection of the Hansen J statistic suggests that
our particular IV procedure is well specified. Put differently, an overly strong rejection of the
Hansen J statistic casts doubt on the validity of the estimates (Burnett and Madariaga, 2017).
42
Table 4. Instrumental variables approach under-identification and over-identification tests
5.2 Alternative Interpretations and Confounding Factors
We interpret the statistical relationship between induced earthquakes and property values as
reflecting the causal effect of oil and gas development, wastewater injections, and induced
seismicity within the MSA. However, there are several potential confounding factors that might
suggest a different interpretation.
First-Stage Regression with Dependent variable
Number of earthquakes (PGV ≥ 0.5 cm/sec) Cumulative CDI
Underidentification testKleibergen-Paap rank Lagrange Multiplier statistic 264.85 1539.99Chi-squared propbability-value 0.00 0.00
Weak identification testCragg-Donald Wald F statistic 5815.20 954.92Kleibergen-Paap rank Wald F statistic 207.58 770.83Stock-Yogo weak ID test critical values
10% maximal IV size 16.38 19.9315% maximal IV size 8.96 11.5920% maximal IV size 6.66 8.7525% maximal IV size 5.53 7.25
Weak-Instrument-Robust inferenceAnderson-Rubin Wald test (F statistic) 33.64 4.61
Probability-value 0.00 0.01Anderson-Rubin Wald test (Chi-squared statistic) 33.69 9.23
Probability-value 0.00 0.01Stock-Wright Lagrange Multiplier S statistic 93.30 11.36
Probability-value 0.00 0.00
Overidentification test of all instrumentsHansen J statistic 0.14Chi-squared probability-value 0.71
43
5.2.1 Correlation Between Housing Prices and Personal Income
Our interpretation of our findings is that wastewater injections created additional seismic
activity, which in turn has led to a decrease in property values in the greater Oklahoma City
region. Additionally, Oklahoma City is home to several oil and gas firms, and roughly one-
quarter of jobs (within the entire State) are tied to oil and natural gas (Wertz, January 19, 2012).
Therefore, it is possible that the negative correlation between (induced) earthquakes and housing
values are not due to seismic activity but rather due to the changes in employment levels (or
labor incomes) resulting from the volatility of oil and natural gas prices. Based on this logic, one
could argue that a decline in oil and natural gas prices could lead to local labor shocks, which
induce an out-migration of labor from the MSA and a reduction in housing prices. Alternatively,
a household’s wealth, as reflected in the private ownership of oil and natural gas rights below a
particular property, may have declined as a result of a decrease in oil and natural gas prices.
We probed this potential explanation further by estimating an IV regression including a
covariate for the annual, state-level personal income in the NAICS code for mining. The
additional IV regressions are provided in the Appendix A3. The inclusion of the personal income
covariate did not change the size, significance or magnitude of our preferred measure of
earthquake activity. In both instances, a one percent increase in earthquake activity leads to an
approximately 2.41 percent decline in home sale prices. Based on this evidence we conclude
that negative relationship between home sale price and earthquake activity is not driven by
changes in oil and natural gas industry employment levels.
44
5.2.2 Induced Earthquakes and Housing Construction
Another plausible explanation is that new housing construction anticipated the potential hotspots
of seismic activity, following the onset of induced earthquakes around the year 2008. Following
this logic, observations of housing prices are no longer randomly distributed, which would imply
a type of omitted variable bias in the impact estimates. If this is a plausible explanation, then it
would not obviate the marginal impact estimates we derived above, and our preferred IV
estimation procedure (that used the underlying sedimentary basin and cumulative CDI measures
of earthquakes) should incorporate this type of behavior into impact estimates. Nevertheless, this
type of sorting within housing starts may affect the second-stage regression results.
To probe the potential confounding effects of new housing starts, we use the Oklahoma
County Tax Assessor data to construct an annual measure of new housing construction in each
census tract. The measure represents the number of residential homes constructed in a given
year. We include the measure as an additional variate in our preferred IV estimated procedure.
The results are reported in Appendix Table A3. The estimates indicate that a 1% increase in
earthquake activity leads to a 3.32% decline in home sale prices compared to a 2.41% decline
from our preferred specification. Despite a nearly one percentage point differences in
magnitude, a series of Wald t-test indicate that the results are not statistically different. Based on
this evidence, we conclude that real estate developers did not anticipate potential hotspots of
seismic activity.
5.2.3 Capitalization of Natural Gas Rights into Housing Values
As mentioned briefly above, another alternative explanation is that the negative correlation
between induced earthquake activities and housing prices is caused by a decline in household
45
wealth associated with private ownership of oil or natural gas rights underlying a property. In
other words, housing values capitalize the rights to revenue from natural gas production. We do
not control for the local ownership of oil or natural gas rights, nor do we control for the value of
oil or natural gas production. We do not think this is likely explanation as Weber et al. (2016)
found evidence to suggest that the capitalization of natural gas prices was linked to greater
appreciation in housing values. Following this logic, if oil or natural gas prices reflect the
expected revenues from royalties, then the 72 percent decline in natural gas prices (between 2008
and 2016) would have potentially decreased the present value of the capitalization of the oil or
natural gas resources, and thus, housing values should have declined during that same period.
The 2008-to-2016 decline in natural gas prices is clear from Figure 3, and housing prices in
Oklahoma City declined marginally between 2008 and 2010, but as listed in Figure 7, housing
prices increased by 24 percent, on average, from 2011 through 2017. Likewise, a similar
argument cannot be made about the capitalization of oil rights, as oil price increased from 2009
to 2013, and then declined thereafter. During this same period, housing values (in the Oklahoma
City MSA) continued to rise, with the exception of a short dip in 2009 and 2010.
6 DISCUSSION
Despite the strong evidence of a link between disposal injections and seismic activity, geologists
or geophysicists have not yet identified a precise dose-response. In other words, we do not know
if the increasing seismicity is due to the number of injections or the accumulation of total
injections into the subsurface basin (U.S. Geological Survey, 2016). Unfortunately, we do not
have enough data at this point in time, as the relationship between injections and seismicity is
46
still a relatively new phenomenon, to fully flesh out the exact dose-response. Therefore, once
more data comes available, future research should further explore the sensitivity of seismicity
and price impacts to future wastewater injections.
Our findings have implications for state policies regarding the regulations of wastewater
injections into disposal wells. Based on our impact estimates, we believe the state of Oklahoma
(or a jurisdiction within all or part of the Oklahoma City MSA) has three potential policy
avenues to address this issue: (1) business as usual; (2) treat induced earthquakes as negative
externalities; and (3) institute a seismicity compensation fund. With the first policy option, the
State could do nothing and simply allow for its citizens (or earthquake insurance providers if
applicable) to assume the risks from possible earthquake damages to private property. The policy
option for the business-as-usual case does have one potential caveat though. That is, it is not
clear if a standard earthquake insurance policy would cover damages from human-induced
earthquakes (Summars, October 4, 2015). In other words, several standard earthquake policies
only provide coverage for natural occurring damages. However, some policies are now carving
out coverage specifically for earthquakes not naturally occurring, such as earthquakes attributed
to wastewater injection from hydraulic fracturing activities (National Association of Insurance
Commissioners, 2017). Hence, a government or regulatory agency may need to ensure that its
private citizens, who reside within a seismically prone region, have the proper type of insurance
coverage given the increase in frequency and severity of (induced) earthquake activities (Monies,
September 8, 2016).
With the second policy option, the State could treat the induced earthquakes as a negative
externality and levee a corrective tax (or impact fee) to provide a source of public funds to
compensate individuals for potential future impact damages. According to Richardson et al.
47
(2013), there are currently 26 states in the U.S. that levee a severance tax for crude oil and
natural gas production. Pennsylvania assesses an “impact fee” against operators in counties that
choose to impose the fee. However, these types of taxes generally are not related to
environmental regulations at all (Richardson et al., 2013). To the best of the authors’ knowledge,
no state in the U.S. has legislated a tax or impact fee (based on environmental regulations)
assessed against direct environmental impacts (including induced seismicity) associated with oil
and natural gas development. Instead, most states have elected to simply monitor and regulate (or
outright prohibit as in the case of North Carolina) underground injection of fluids produced in
the extraction of oil and gas (Richardson et al., 2013). In a similar vein, the Oklahoma
Corporation Commission, which has exclusive jurisdiction to regulate class II underground
injection wells, launched a plan (starting in 2015) to heavily monitor and reduce the risk of
induced earthquakes.
A third policy option for the State would require a supplementary industry-level
insurance or compensation fund, collected through production levees against individual shale gas
developers (Konschnik, 2017). The fund would pay out compensation for large environmental
damages that would otherwise be impractical to recover from individual funds (Daniel et al.,
2017). According to Daniel et al. (2017), this would give the industry (within a particular state or
jurisdiction) to self-regulate, by monitoring member firm’s efforts to limit environmental (or
seismic) risks and thereby contain the production fees. A similar type of arrangement has been
established with International Oil Pollution Compensation Funds, which provide financial
compensation for pollution damages caused by oil spills, among other types of accidents (IOPC
Funds, 2017).
48
Our findings also have implications for states located in neighboring regions where the
unconventional oil and gas development is taking place. For example, the state of Pennsylvania
did not adopt the primacy of the U.S. Environmental Protection Agency’s (EPA) underground
injection controls (UIC) program; therefore, the EPA implements the State’s UIC program (U.S.
EPA, 2015). As such, a large volume of wastewater from Pennsylvania is disposed of in the
neighboring states of West Virginia and Ohio. If seismicity is definitively linked to wastewater
injections, then interstate disposal of natural gas and oil production wastewater could potentially
require heavier future federal oversight as production in one state is arguably linked with seismic
activity in a neighboring state. The case of cross-state induced seismicity would seem to clearly
constitute a negative externality, which implies perhaps federal intervention and regulation.
REFERENCES
Abadie, A. and G. Imbens. 2006. Large sample properties of matching estimators for average
treatment effects. Econometrica, 74, 235–267.
American Oil & Gas Historical Society. 2017. The first Oklahoma oil well. Retrieved April
2017, from http://aoghs.org/petroleum-pioneers/first-oklahoma-oil-well/.
Boslett, A., T. Guilfoos, C. Lang. 2016. Valuation of expectations: A hedonic study of shale gas
development and New York’s moratorium. Journal of Environmental Economics and
Management, 77, 14-30.
Burnett, J.W. and J. Madariaga. 2017. The convergence of US state-level energy intensity.
Energy Economics, 62, 357-370.
Cameron, A.C. and D.L. Miller. 2015. A practitioner’s guide to cluster-robust inference. The
Journal of Human Resources, 50, 317-372.
49
Cheung, R., D. Wetherell, and S. Whitaker. 2016. Earthquakes and house prices: Evidence from
Oklahoma. Federal Reserve Bank of Cleveland, Working Paper no. 16-31.
Daniel, P., A. Krupnick, T. Matheson, P. Mullins, I. Parry, and A. Swistak. How should shale
gas extraction be taxed? IMF Working Paper, WP/17/254. Retrieved November 2017,
from www.imf.org/~/media/Files/Publications/WP/2017/wp17254.ashx.
Duranton, G. and H.G. Overman. 2005. Testing for localization using micro-geographic data.
Review of Economic Studies, 72(4), 1077-1106.
Gallucci, M. April 22, 2015. Oklahoma earthquake swarm 2015: In sharp turnaround, Oklahoma
officials confirm the link between fracking wastewater and earthquakes. International
Business Times. Retrieved April 2017, from http://www.ibtimes.com/oklahoma-
earthquake-swarm-2015-sharp-turnaround-oklahoma-officials-confirm-link-1892086.
Gopalakrishnan, S. and H.A. Klaiber. 2014. Is the shale energy boom a bust for nearby
residents? Evidence from housing values in Pennsylvania. American Journal of
Agricultural Economics, 96(1), 43-66.
Horton, S. 2012. Disposal of hydrofracking waste fluid by injection into subsurface aquifers
triggers earthquake swarm in central Arkansas with potential for damaging earthquake.
Seismological Research Letters, 83, 250-260.
Imbens, G. 2014. Instrumental variables: An econometrician’s perspective. Statistical Science,
29(3), 323-358.
IOPC Funds. 2017. The international regime for compensation for oil pollution damage:
Explanatory note. Retrieved December 2017, from
http://www.iopcfunds.org/fileadmin/IOPC_Upload/Downloads/English/explanatory_note
.pdf.
50
Keranen, K.M., H.M. Savage, G.A. Abers, and E.S. Cochran. 2013. Potentially induced
earthquakes in Oklahoma, USA: Links between wastewater injection and the 2011 Mw
5.7 earthquake sequence. Geology, 41, 699-702.
Konschnik, K. 2017. Regulating stability: State compensation funds for induced seismicity. The
Georgetown Environmental Law Review, 29(2), 227-300.
Koster, H.R.A. and J. van Ommeren. 2015. A shaky business: Natural gas extraction,
earthquakes and house prices. European Economic Review, 80, 120-139.
Kuchment, A. March 28, 2016. Drilling for earthquakes: Scientists are increasingly confident the
link between earthquakes and oil and gas production, yet regulators are slow to react.
Scientific American. Retrieved July 2017, from
https://www.scientificamerican.com/article/drilling-for-earthquakes/.
Liu, H., S. Ferreira, and B. Brewer. 2016. The housing market impacts of wastewater injection
induced seismicity risk. Paper presentation, Agricultural and Applied Economics
Association 2016 Annual Meeting, Boston, MA.
Merrill, T. November 14, 2016. Oklahoma City, OK: Real estate market and trends 2016.
FortuneBuilders, Inc. Retrieved March 2017, from
https://www.fortunebuilders.com/oklahoma-city-ok-real-estate-market-trends-2016/.
Monies, P. September 8, 2016. New insurance products may be needed for Oklahoma quakes,
report says. The Oklahoman. Retrieved December 2017, from
http://newsok.com/article/5517311.
Muehlenbachs, L., E. Spiller, and C. Timmins. 2015. The Housing Market Impacts of Shale Gas
Development. American Economic Review, 105(12), 3633-3659.
51
Murray, K.E. 2014. Class II underground injection control well data for 2010-2013 by geological
zones of completion, Oklahoma. Oklahoma Geological Survey. Open-File Report (OF1-
2014). Retrieved April 2017, from
http://www.ogs.ou.edu/pubsscanned/openfile/OF1_2014_Murray.pdf.
National Association of Insurance Commissioners. 2017. Hydraulic Fracturing (Fracking). The
Center for Insurance Policy and Research. Retrieved December 2017, from
http://www.naic.org/cipr_topics/topic_hydraulic_fracturing.htm.
National Earthquake Hazards Reduction Program. 2013. Getting in GEER for New Zealand:
Joint reconnaissance of the 2011 Christchurch earthquake. Seismic Waves, March 2013,
1-2. Retrieved August 2017, from http://www.nehrp.gov/pdf/SeismicWavesMar13.pdf.
Nicholson C. and R.L. Wesson. 1990. Earthquake hazard associated with deep well injection – A
report to the U.S. Environmental Protection Agency. US Geological Survey Bulletin
1951, 74.
Oklahoma Office of Energy and Environment. 2017. Earthquakes in Oklahoma: Earthquake
Map. Retrieved May 2017, from http://earthquakes.ok.gov/what-we-know/earthquake-
map/.
Oklahoma Oil and Gas Association. 2017. Shale plays and unconventional sources. Retrieved
March 2017, from http://okoga.com/sources/.
Philips, M. November 7, 2016. Why Oklahoma can’t turn off its earthquakes. Bloomberg
Businessweek. Retrieved March 2017, from
https://www.bloomberg.com/news/articles/2016-11-08/why-oklahoma-can-t-turn-off-its-
earthquakes.
52
Richardson, N., M. Gottlieb, A. Krupnick, and H. Wiseman. 2013. The State of state shale gas
regulation. Resources for the Future, Washington, D.C. Retrieved November 2017, from
http://www.rff.org/files/sharepoint/WorkImages/Download/RFF-Rpt-
StateofStateRegs_Report.pdf.
Rubin, D.B. 1974. Estimating causal effects of treatments in randomized and nonrandomized
studies. Journal of Educational Psychology, 66(5), 688–701.
Rubin, D.B. 2006. Matched sampling for causal effects. Cambridge, England: Cambridge
University Press.
Summars, E. October 4, 2015. Who’s at fault? Quake calls defined by insurance policies. Enid
News and Eagle. Retrieved November 2017, from
http://www.enidnews.com/news/earthquakes/who-s-at-fault-quake-calls-defined-by-
insurance-policies/article_9f716a82-6a4c-11e5-8d0c-0b66c3623c0e.html.
U.S. Energy Information Administration. 2017a. Oklahoma: State profile and energy estimates.
U.S. Department of Energy, Washington, D.C. Retrieved March 2017, from
https://www.eia.gov/state/?sid=OK.
U.S. Energy Information Administration. 2017b. Oklahoma natural gas number of gas and gas
condensate wells. U.S. Department of Energy, Washington, D.C. Retrieved April 2017,
from https://www.eia.gov/dnav/ng/hist/na1170_sok_8a.htm.
U.S. Energy Information Administration. 2017c. Oklahoma field production of crude oil. U.S.
Department of Energy, Washington, D.C. Retrieved April 2017, from
https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=pet&s=mcrfpok1&f=a.
53
U.S. Federal Housing Finance Agency, 2017. All-transactions house price index for Oklahoma
City, OK (MSA). Retrieved May 2017, from
https://fred.stlouisfed.org/series/ATNHPIUS36420Q.
U.S. Geological Survey. 2017a. Did you feel it? U.S. Department of the Interior, Washington,
D.C. Retrieved February 2017, from https://earthquake.usgs.gov/data/dyfi/.
U.S. Geological Survey. 2017b. Magnitude/intensity comparison. U.S. Department of the
Interior, Washington, D.C. Retrieved March 2017, from
https://earthquake.usgs.gov/learn/topics/mag_vs_int.php.
U.S. Geological Survey. 2017c. DYFI Scientific Background. U.S. Department of the Interior,
Washington, D.C. Retrieved November 2017, from
https://earthquake.usgs.gov/data/dyfi/background.php. Throupe, R., R. Simons, and X. Mao. 2013. A review of hydro “fracking” and its potential
effects on real estate. Journal of Real Estate Literature, 21(2), 205-232.
Wald, D.J., V. Quitoriano, C.B. Worden, M. Hopper, and J.W. Dewey. 2011. USGS “Did You
Feel It?” internet-based macroseismic intensity maps. Annals of Geophysics, 54 (6), 688-
707.
Weber, J.G., J.W. Burnett, and I.M. Xiarchos. 2016. Broadening benefits from natural resource
extraction: Housing values and taxation of natural gas wells as property. Journal of Policy
Analysis and Management, 35(3), 587-614.
Wertz, J. January 19, 2012. Oil and gas and what a ‘boom’ means to big-energy Oklahoma. State
Impact, National Public Radio. Retrieved December 2017, from
https://stateimpact.npr.org/oklahoma/2012/01/19/oil-gas-and-what-a-boom-means-to-big-
energy-oklahoma/.
54
Wooldridge, J.M. 2012. Introductory econometrics: A modern approach. Boston: Cengage
Learning.
55
APPENDIX
A1 Auxiliary Regression of quantiles of housing prices against cumulative earthquake
activities
Table A1. Regression of quantiles of housing prices against cumulative earthquake activities Total
Sample First
Quartile (0 to 25%)
Second Quartile
(25 - 50%)
Third Quartile
(50 - 75%)
Fourth Quartile
(75 - 100%)
Total Sample
(interaction effects)
Cumulative Earthquake Activity
-0.0436*** (0.0009)
-0.0731*** (0.0026)
-0.0272*** (0.0009)
-0.0106*** (0.0006)
-0.0130*** (0.0011)
-0.0264*** (0.0008)
Cumulative Earthquake Activity ´ Price Group 1
-0.0689*** (0.0008)
Cumulative Earthquake Activity ´ Price Group 2
-0.0309*** (0.0005)
Cumulative Earthquake Activity ´ Price Group 3
-0.0180*** (0.0004)
Constant 10.30 10.36*** 10.83 11.13*** 10.33*** 10.34 (37.1100) (0.0456) (0.0166) (0.0294) Observations 152,522 37,291 38,655 38,039 38,537 152,522 R-squared 0.829 0.489 0.473 0.473 0.748 0.841 Adjusted R-squared 0.829 0.486 0.470 0.470 0.747 0.841 Notes: Robust standard errors in parentheses. The asterisk marks indicate the following: *** p<0.01, ** p<0.05, and * p<0.1.
56
A2 Oklahoma County Tax Assessor’s Office: Assessment Characteristics
Table A2: Tax Assessor variables used within this particular study. Variable Name Description Units Type
Acct_Num Account Number Unique Parcel Identifier String Sale_Price Nominal Sale Price Nominal Dollars Number
Sale_Month Sale Month Number Sale_Day Sale Day Number Sale_Year Sale Year Number
Sale_Date_2 Sale Date in MDY
Format Days since 1/1/1960 Number Prop_Type Property Type String Ngbh_Code Neighborhood Code String Square_Feet Square Feet Square Feet Number
Basement_Sq Basement Square Feet Square Feet Number Garage_Sq Garage Square Feet Square Feet Number Carport_Sq Carport Square Feet Square Feet Number
Rooms Total Number of Rooms Count Number
Bedrooms Total Number of
Bedrooms Count Number Baths Total Number of Baths Count Number
Year_Built Year Built Number Year_Remodeled Year Remodeled Number Street_Address Street Address Strings
City City Strings Zip_Code Zip Code Strings Lot_Size Lot Size in Square Feet Square Feet Number Vacant -1 Indicates Vacant Number
Tax_District Tax District String X_Coor X Coordinate in DD Number Y_Coor Y Coordinate in DD Number
Table A2 offers a list of the assessment characteristics, as provided by the Oklahoma County Tax
Assessor’s Office, used within the study.
57
A3 Community Decimal Index Algorithm
Table A3. Did You Feel It Questionnaire.
The Community Decimal Intensity (or CDI) is an aggregate of the weighted sums of the various
indices of the DYFI questionnaires. According to the U.S. Geological Survey (2017c), there are
eight questions, which are listed in Table A3, used in the calculation of CDI. The far-left column
demonstrates the weights associated with each question and the corresponding response. The
second column shows the range of possible responses, based on a Likert scale, for each question
within the survey. Finally, U.S. Geological Survey (2017c) defines the algorithm as follows:
1. We turn each answer into a numeric value from 0 (not felt/no effect) to 1 or more. 2. We take the average of all answers to that question within that community. Note that
unanswered questions or answers of “not applicable” are NOT counted in this average.
3. We take the weighted sum of all the averages to form the community weighted sum (CWS).
4. We calculate the DYFI Intensity as follows: 5. CDI = 3.40 ln(CWS) - 4.38 6. CDI is rounded off to the first decimal place. We set a minimum CDI of 2 if the CWS
is nonzero (so the result is at least 2 “Felt”, or 1 “Not felt”), and cap the result at 9.0. A further description of the survey can be found at Wald et al. (2012).
Weight Range Question
5x 0-1 Did you feel it?*
1x 0-5 How would you describe the shaking?
1x 0-5 How did you react?
2x 0-1 Was it difficult to stand or walk?
5x 0-1 Did objects rattle, topple over, or fall off shelves?
2x 0-1 Did pictures move of get knocked askew?
3x 0-1 Did furniture slide, topple, or become displaced?
5x 0-3 Was there damage to the building?
58
A4 Sensitivity Analysis
Table A4. Regression analyses demonstrating alternative instrumental variables IV - Cumulative CDI IV - Cumulative CDI
Variables Personal Income New Home Count Earthquake Impact -0.0243*** -0.0332*** (0.00920) (0.00895) Personal Income -2.56e-07 (1.61e-05) New Home Count -0.000290*** (3.98e-05) Square Feet 0.346*** 0.345*** (0.00293) (0.00293) Basement square feet 0.245*** 0.244*** (0.0116) (0.0116) Garage square feet 0.146*** 0.146*** (0.00510) (0.00509) Carport square feet 0.0240 0.0242 (0.0210) (0.0210) Bedrooms 0.00209 0.00232 (0.00170) (0.00170) Baths 0.0719*** 0.0720*** (0.00243) (0.00243) Age -0.0100*** -0.0102*** (0.000153) (0.000151) Square of age 4.34e-05*** 4.65e-05*** (1.90e-06) (1.90e-06) Remodeled 0.0659*** 0.0663*** (0.00137) (0.00137) Log of lot size 0.0646*** 0.0647*** (0.00223) (0.00223) Wells (1.5 km) -0.00680*** -0.00691*** (0.000859) (0.000909) Wells (20 km) 0.000293*** 0.000240*** (4.50e-05) (5.63e-05) Ground water x wells (1.5 km) -0.000196 -0.000794 (0.00209) (0.00209) Ground water x wells (20 km) -2.89e-05** -3.28e-05** (1.33e-05) (1.35e-05) Distance to Oklahoma City -0.00274 -0.00315 (0.00201) (0.00202) Constant 9.826*** 9.849*** (0.0663) (0.0672)
Observations 152,522 152,522 R-squared 0.829 0.829
Adjusted R-squared 0.829 0.829 Notes: Each regression also includes a time trend, an indicator for years after 2008 and 238 census tract fixed effects estimates, which have been omitted for the sake of exposition. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
