# RECORD d3812508998e06d0728e9af8b9c2c72e4dad465a
# TITLE: Crime Measures and Housing Prices: an Analysis Using Quantile Regression and Spatial Autocorrelation
# YEAR: 2025.0  VENUE: Journal of real estate finance and economics  DOI: 10.1007/s11146-024-09997-w
# AUTHORS: J. Kallberg; Yoshikiyo Shimizu
# PAGES: 53  CANDIDATE TABLE PAGES (1-based): [18, 26, 27, 28, 29, 30, 34, 35, 36, 37, 38, 41, 42, 43, 44, 47, 48]
# ABSTRACT: Crime is a disamenity, so buyers should be willing to pay more for a house (all else equal) in a low crime area, suggesting that high crime rates depress housing prices. Conversely, it is plausible that criminals prefer wealthier areas because of the higher expected returns from their transgressions. This study examines the link between measures of crime and prices of residential housing. Our data begin in 2008 and end in 2020 for Seattle, Washington, including all reported felonies (756,304); 911 calls (1,528,303); all recorded residential real estate transactions (61,902 after filtering), as well as the corresponding property characteristics; demographic data and the associated changes. The inherent endogeneity between crime rates and housing prices forces us to find an instrument for crime rates. After rejecting several plausible choices, based on the Wu-Hausman test, the key variable in our empirical analysis is the number of 911 calls (contemporaneous and lagged) in a given beat. Our results present somewhat mixed evidence on the impact of crime on housing prices. Without adjustment for spatial autocorrelation, a 1 percentage point increase in crime rates (instrumented by the number of 911 calls) leads to approximately a 0.55% decrease in house prices. However adjusting for spatial autocorrelation changes this figure to a 0.80% increase. We further show that distance to a crime hotspot is significantly negatively related to housing prices, suggesting that criminals choose to operate in wealthier areas, which is consistent with our findings incorporating spatial autocorrelation.


===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) =====

Vol.:(0123456789)
The Journal of Real Estate Finance and Economics (2025) 72:477–529
https://doi.org/10.1007/s11146-024-09997-w
Crime Measures and Housing Prices: an Analysis Using 
Quantile Regression and Spatial Autocorrelation
Jarl G. Kallberg1 · Yoshiki Shimizu2 
Accepted: 18 August 2024 / Published online: 30 January 2025 
© The Author(s) 2025
Abstract
Crime is a disamenity, so buyers should be willing to pay more for a house (all else 
equal) in a low crime area, suggesting that high crime rates depress housing prices. 
Conversely, it is plausible that criminals prefer wealthier areas because of the higher 
expected returns from their transgressions. This study examines the link between 
measures of crime and prices of residential housing. Our data begin in 2008 and 
end in 2020 for Seattle, Washington, including all reported felonies (756,304); 911 
calls (1,528,303); all recorded residential real estate transactions (61,902 after filter-
ing), as well as the corresponding property characteristics; demographic data and 
the associated changes. The inherent endogeneity between crime rates and hous-
ing prices forces us to find an instrument for crime rates. After rejecting several 
plausible choices, based on the Wu-Hausman test, the key variable in our empiri-
cal analysis is the number of 911 calls (contemporaneous and lagged) in a given 
beat. Our results present somewhat mixed evidence on the impact of crime on hous-
ing prices. Without adjustment for spatial autocorrelation, a 1 percentage point 
increase in crime rates (instrumented by the number of 911 calls) leads to approxi-
mately a 0.55% decrease in house prices. However adjusting for spatial autocorre-
lation changes this figure to a 0.80% increase. We further show that distance to a 
crime hotspot is significantly negatively related to housing prices, suggesting that 
criminals choose to operate in wealthier areas, which is consistent with our findings 
incorporating spatial autocorrelation.
Keywords  Crime rates · Crime hotspots · Hedonic models · Quantile regression · 
Spatial autocorrelation · Wu-Hausman test
 *	 Yoshiki Shimizu 
	
yshimizu@hub.hit-u.ac.jp
1	
Carson College of Business, Washington State University, Pullman, WA 99164‑4746, USA
2	
Hitotsubashi University Business School (HUB), Hitotsubashi University, Tokyo 101‑8439, 
Japan


===== PAGE 18 (score 35, numeric-paren 0, table-words 3, var-words 14) =====

494
	
J. G. Kallberg, Y. Shimizu 
Table 2   Summary statistics
N
Mean
Median
Std Dev
Min
Max
Property-level characteristics
  Sale price
61,902
706,110
617,500
457,343
10,000
14,275,000
  WY
61,902
657,449
592,119
324,996
74,027
6,257,625
  Living area
61,902
1.891
1.720
0.792
0.350
4.590
  Lot area
61,902
4.660
4.650
2.636
0.375
14.466
  Bedrooms
61,902
3.192
3.000
0.921
0.000
11.000
  Bathrooms
61,902
1.485
1.000
0.689
0.000
7.000
  One story
61,902
0.398
0.000
0.490
0.000
1.000
  Age
61,902
55.571
65.000
38.399
0.000
114.000
  Renovated
61,902
0.048
0.000
0.214
0.000
1.000
  FirePlace
61,902
0.712
1.000
0.453
0.000
1.000
Environmental—natural views
  Lake/Mountain view
61,902
0.121
0.000
0.327
0.000
1.000
Distance to crime hotspots
  Straight distance (miles)
    Nearest
61,902
3.784
3.967
1.609
0.031
8.140
    2nd nearest
61,902
4.015
4.199
1.595
0.299
8.165
    3rd nearest
61,902
4.140
4.332
1.592
0.445
8.334
    4th nearest
61,902
4.294
4.478
1.609
0.592
8.645
    5th nearest
61,902
4.428
4.604
1.607
0.614
8.728
  Travel distance (miles)
    Nearest
61,902
4.795
5.088
1.962
0.046
10.902
    2nd nearest
61,902
5.195
5.506
1.979
0.476
11.156
    3rd nearest
61,902
5.428
5.745
2.021
0.638
11.348
    4th nearest
61,902
5.595
5.891
2.014
0.755
11.374
    5th nearest
61,902
5.802
6.114
2.011
0.825
11.396
  Travel time (minutes)
    Nearest
61,902
12.335
12.864
3.992
0.118
27.848
    2nd nearest
61,902
13.544
14.093
3.646
3.030
27.877
    3rd nearest
61,902
14.076
14.627
3.614
3.412
28.819
    4th nearest
61,902
14.835
15.399
3.546
4.221
29.700
    5th nearest
61,902
15.488
16.001
3.556
4.553
29.980
Sociodemographic characteristics (census tract-year observations)
N
Mean
Median
Std Dev
Min
Max
Population
1,235
5,138
4,900
1,683
1,915
11,930
% Male
1,235
0.492
0.491
0.018
0.439
0.567
% Female
1,235
0.508
0.509
0.018
0.433
0.561
% White
1,235
0.709
0.792
0.194
0.128
0.922
% Asian
1,235
0.139
0.095
0.113
0.032
0.639
% Black
1,235
0.083
0.033
0.097
0.005
0.426
% Old population
1,235
0.116
0.113
0.041
0.040
0.256
The sample consists of 61,902 single-family home sales in Seattle from 2008 to 2020. Definitions of the 
variables are provided in Appendix 1 Table 10


===== PAGE 19 (score 8, numeric-paren 0, table-words 2, var-words 4) =====

495
Crime Measures and Housing Prices: an Analysis Using Quantile…
The second choice seems the most frequent choice in the related literature and the 
one that we employ in our analysis.
To calculate Wy for home i’s sale (shown as a red dot • in Fig. 3), we first identify 
sales of its neighbor homes (j = 1,2,…) that occurred within 2,000 feet, within the 
past 24 months. The dark dots within the red circle are the sales of neighbor homes. 
For each neighbor home sale j, we measure the distance from home sale i and use 
the sale price yj for neighbor home j to calculate wijyj. Finally, we calculate the spa-
tial weight matrix for the sale of house i, [Wy]it, using the above equation.
Hotspot and Coldspot Characterization
We define a crime hotspot as a census tract with high cluster cores (densities) of 
crimes.39 As noted above, there is a substantial literature that documents the exist-
ence of crime hotspots, specifically in Seattle. For each year, we aggregate the 
number of reported crimes at the census tract level, and compute the Getis and Ord 
(1992) statistics, G∗
i for each tract i, defined as:
The numerator is the weighted average number of crimes reported in the neigh-
boring locations (census tracts j = 1,…,105). The denominator is the sum of all crime 
counts across all the census tracts. We categorize a census tract with a G∗
i larger than 
the mean as a “high-high cluster,” or “hotspot.” Similarly, a census tract with a G∗
i 
lower than the mean is categorized as a “low-low cluster,” or “coldspot.” For each 
year we select census tracts defined as hotspots based on G∗
i , and sort them from 
the highest to lowest within the High-High cluster. A census tract with the highest 
G∗
i statistics (with a p-value < 0.05 for significance) is defined as a hotspot, the tract 
with the highest crime density. We also define a census tract with the second, third, 
fourth and fifth highest and significant G∗
i statistics as the hotspot with the second, 
third, fourth and fifth highest crime density.
Figure 4 depicts census tracts with high crime density (red), those with low crime 
density (blue) or no significant density (gray) in 2010, 2015 and 2020. Those cat-
egorized as red and blue have G∗
i with p-value < 0.05.
Endogeneity
A major confounding issue in this analysis is the likely endogeneity of crime rates 
and housing prices, as noted earlier. This problem was initially handled in the con-
text of hedonic models by Burgess and Harmon (1991). To address this issue we 
use the Wu-Hausman test, which assumes the null hypothesis of no endogeneity.40 
It tests for the correlation between the regression residuals and a given explanatory 
G∗
i =
∑
j wijxj
∑
j xj
39  Ceccato and Wilhelmsson (2020) shows that housing prices are more sensitive to distance to crime 
hotspots than to overall crime rates.
40  Wu (1973) and Hausman (1978).


===== PAGE 26 (score 11, numeric-paren 5, table-words 1, var-words 4) =====

502
	
J. G. Kallberg, Y. Shimizu 
in years 2010, 2015 and 2020. Consistent with our earlier observation, we find a 
concentration of crime around the Westlake Center area consistently throughout 
2010, 2015 and 2020. The area with crime intensity further expands in 2020.
Wu‑Hausman Test and Correlations
We identify a set of 9 possible instruments. We then regress these potentially endog-
enous variables on 911 Calls as well as other exogenous variables on the right-hand 
side of Eq. 1 and obtain the residuals (v). We include v in Eq. 1 and test if its coef-
ficient is statistically different from zero.
In Table 5 Panel A, the computed Wu-Hausman statistic has p-values < 0.05, 
meaning that we reject the Ho (the variable is exogenous), for Number of 
Crimes – Total, Property and Social; for changes: ∆ Crimes – Total and Prop-
erty. With these factors as a right-hand side variable, an OLS model is biased 
and hence an instrumental variable approach is necessary. Four variables pass 
the Wu-Hausman test (Number of Crimes – Violent, ∆ Crimes—Violent, ∆ 
Crimes—Social and Distance to Crime Hotspot) and can thus be used in an 
OLS regression.
The crime variables that did not pass the Wu-Hausman test are instrumented by 
the 911 Calls variables.46 To validate that 911 Calls is a viable instrument, we per-
form a Montiel—Pfluger test to identify a weak instrument: in both cases the vari-
ables pass the test.
Panel B reports correlations between these variables. For levels (changes), we 
compute the aggregate number of 911 calls and crimes per month for Number of 
Crimes – Total, Violent, Property and Social, and Distance to Crime Hotspot 
(Straight Miles, Travel Miles and Travel Time). The correlations between Number of 
Crimes – Total and subsets are positive, as expected. Interestingly, the correlation of 
911 Calls (both for levels and changes) is strongest for Violent (0.465), slightly less 
positive for Property (0.268) and Social (0.202).
Our 3 measures of Distance to Crime Hotspot variable are only significantly 
correlated with Violent (− 0.325). These 3 variables, however, are positively cor-
related with 911 Calls, suggesting that the farther away from a crime hotspot, 
the likelier it is that a 911 call will be made. This can be partially explained by 
the higher allocation of police resources to areas with the highest incidences of 
crime. The Center for Evidence-Based Crime Policy (CEBCP) reports that the 
effort to allocate more resources in crime hotspot areas resulted in reduced num-
bers of calls for service in the affected areas.47 Consistent with our intuition, our 
3 measures of Distance to Crime Coldspot are significantly correlated with Vio-
lent (0.276) and Social (0.519).
46  Number of Crimes and ∆ Crimes are instrumented by Number of 911 Calls and ∆ 911 Calls, respec-
tively.
47  https://​haven​gear.​com/​blog/​incre​asing-​hot-​spots-​polic​ing-​resul​ts-​in-​lower-​crime-​rates/.


===== PAGE 27 (score 25, numeric-paren 0, table-words 4, var-words 11) =====

503
Crime Measures and Housing Prices: an Analysis Using Quantile…
2SLS and Quantile Regressions
Results Without Spatial Autocorrelation Adjustments
Table 6 displays the results of running 2SLS and quantile regressions without incor-
porating adjustments for spatial autocorrelation or Distance to Crime Coldspot; 
when these influences are included we find important changes in our results. All 
estimates use month and year dummies.48 We have 61,902 observations using 911 
Calls as an instrument for Number of Crimes – Total. 
Our 2SLS estimates in model 1 are generally consistent with similar hedonic resi-
dential pricing models. We find a positive significant impact from: a larger living 
area or lot size; a younger home; more bathrooms; homes with a fireplace or view. 
48  In the existing hedonic literature, area-fixed effects are usually included to control for unobserved 
area-specific characteristics to address an omitted variable problem. Abbott and Klaiber (2011) docu-
ments that the omitted variable bias remains even with the use of area-fixed effects, both defined at 
coarse levels (cities or counties) or at much finer scales (zip areas or census tracts/blocks). Our baseline 
models (Eqs. 1 and 2) exclude area-fixed effects, and we address the omitted variable issue by including 
the spatial autocorrelation variable, Wy (Eqs. 3 and 4), to capture important neighborhood characteristics 
shared with nearby residents.
Table 4   Getis-Ord statistics and crime hot/coldspots
This table reports Getis-Ord statistics and the number of census tracts identified as crime hot/coldspots in 
each year during the 2008 – 2020 period. For each census tract i, we compute the Getis-Ord (Gi
*) statistic 
as the weighted-average number of crimes reported in its neighboring census tract j (j = 1, 2, …), divided 
by the sum of all crime counts across all the census tracts. A census tract j is defined as census tract i’s 
neighbor if it shares a border (including a vertex) with tract i based on the queen contiguity. A census 
tract with Gi
* greater (smaller) than the average Gi
* across all the census tracts in a given year is defined 
as a hot (cold) spot
Year
# of crimes per 
tract
Getis-ord statistics 
(G*)
Hotspot tracts
Coldspot tracts Non-hot/cold 
spot tracts
Mean
Median
Mean
Median
N
(%)
N
(%)
N
(%)
2008
460
337
0.775%
0.621%
13
9.63%
9
6.67%
113
83.70%
2009
468
350
0.777%
0.621%
13
9.63%
12
8.89%
110
81.48%
2010
452
339
0.776%
0.640%
14
10.37%
12
8.89%
109
80.74%
2011
422
316
0.776%
0.652%
11
8.15%
10
7.41%
114
84.44%
2012
421
318
0.778%
0.633%
12
8.89%
12
8.89%
111
82.22%
2013
456
332
0.774%
0.653%
13
9.63%
12
8.89%
110
81.48%
2014
500
365
0.778%
0.639%
12
8.89%
11
8.15%
112
82.96%
2015
487
349
0.777%
0.640%
13
9.63%
11
8.15%
111
82.22%
2016
505
379
0.777%
0.639%
11
8.15%
7
5.19%
117
86.67%
2017
522
363
0.778%
0.629%
12
8.89%
10
7.41%
113
83.70%
2018
537
352
0.782%
0.622%
14
10.37%
10
7.41%
111
82.22%
2019
500
356
0.783%
0.626%
14
10.37%
9
6.67%
112
82.96%
2020
530
432
0.771%
0.653%
15
11.11%
8
5.93%
112
82.96%


===== PAGE 28 (score 20, numeric-paren 11, table-words 2, var-words 3) =====

504
	
J. G. Kallberg, Y. Shimizu 
Table 5   Endogeneity, instrumental variables and correlations
Panel A: Tests for endogeneity and weak instrument
Wu-Hausman F-Stats (p-value)
(H0: Variable is exogenous)
Need an IV?
Number of 
crimes
16.496***
(< 0.001)
Yes
Number of 
crimes—
property
22.232***
(< 0.001)
Yes
Number of 
crimes—
violent
0.008
(0.928)
No
Number of 
crimes—
social
29.296***
(< 0.001)
Yes
Δ crimes
5.627**
(0.018)
Yes
Δ crimes—
property
4.320**
(0.038)
Yes
Δ crimes—
violent
0.358
(0.550)
No
Δ crimes—
social
0.889
(0.342)
No
Distance 
to crime 
hotspot
0.725
(0.346)
No
*, **, and *** indicate significance at the 10, 5 and 1 percent levels, respectively
Montiel—Pflu-
eger effective 
F-Stats
(H0: Variable 
is a weak 
instrument)
Critical 
value
(tau = 5%)
Weak IV?
Number of 911 calls
3,374.996
37.418
No
Δ 911 calls
374.025
37.418
No


===== PAGE 29 (score 76, numeric-paren 56, table-words 2, var-words 3) =====

505
Crime Measures and Housing Prices: an Analysis Using Quantile…
Table 5   (continued)
Panel B: Correlations
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(1) Number of 911 Calls
1.000
(2) Number of Crimes—
Total
0.344*
1.000
(3) Number of Crimes—
Violent
0.465*
0.497*
1.000
(4) Number of Crimes—
Property
0.268*
0.964*
0.301*
1.000
(5) Number of Crimes—
Social
0.202*
0.288*
0.478*
0.050
1.000
(6) Δ 911 Calls
0.245*
0.183*
0.158
0.159
0.088
1.000
(7) Δ Crimes—Total
0.141
0.595*
0.139
0.601*
0.153
0.481*
1.000
(8) Δ Crimes—Violent
0.154
0.218*
0.319*
0.166
0.128
0.542*
0.448* 1.000
(9) Δ Crimes—Property
0.130
0.606*
0.120
0.622*
0.120
0.441*
0.991* 0.368*
1.000
(10) Δ Crimes—Social
0.117
0.135
0.046
0.086
0.286*
0.362*
0.533* 0.292*
0.446* 1.000
(11) Distance to Crime 
Hotspot (Straight Miles)
0.410*
 − 0.067
 − 0.325*
0.012
 − 0.186*
0.027
0.054
0.007
0.049
0.074
1.000
(12) Distance to Crime 
Hotspot (Travel Miles)
0.423*
 − 0.010
 − 0.252*
0.072
 − 0.234*
0.013
0.061
0.004
0.057
0.083
0.966*
1.000
(13) Distance to Crime 
Hotspot (Travel Time)
0.428*
 − 0.024
 − 0.317*
0.092
 − 0.363*
0.033
0.054
0.003
0.053
0.054
0.902*
0.948*
1.000
(14) Distance to Crime 
Coldspot (Straight 
Miles)
0.054
0.005
0.276*  − 0.137
0.519*
0.021
 − 0.001
0.049
 − 0.009
0.029  − 0.108
 − 0.152
 − 0.255*
1.000
(15) Distance to Crime 
Coldspot (Travel Miles)
0.098
0.019
0.348*  − 0.141
0.564*
0.022
0.000
0.054
 − 0.008
0.023  − 0.144
 − 0.191*
 − 0.291*
0.957* 1.000
(16) Distance to Crime 
Coldspot (Travel Time)
0.231*
0.122
0.480*  − 0.051
0.580*
0.019
0.007
0.059
0.001
0.018  − 0.193*
 − 0.217*
 − 0.311*
0.900* 0.951* 1.000
Panel A reports the results of the Wu-Hausman test for endogeneity of the crime measure variables (as defined in Appendix 1 Table 10) and the result of the Montiel – 
Pflueger weak instrument test for Number of 911 Calls and Δ911 Calls. The Wu-Hausman test is performed based on Eq. 1. Panel B reports the correlations between the 
instrumental variables and the crime measure variables
*  shows significance at the 0.05 level


===== PAGE 30 (score 238, numeric-paren 195, table-words 4, var-words 15) =====

506
	
J. G. Kallberg, Y. Shimizu 
Table 6   2SLS and quantile regressions – results without spatial autocorrelation adjustments
This table reports the results of 2SLS regression (model 1, as estimated in Eq. 1) and quantile regressions 
(models 2 – 6, as estimated in Eq. 2) for 5, 25, 50, 75 and 95 percent quantiles, respectively. Definitions of 
the variables are provided in Appendix 1 Table 10. Number of crimes is instrumented by Number of 911 
calls. Standard errors are reported in parentheses: for the 2SLS regression (model 1), standard errors are 
clustered by year and month. For quantile regressions (models 2 – 6), standard errors are obtained through 
bootstrap replications. *, ** and *** denote significance at the 10, 5 and 1 percent levels, respectively
Using Number of 911 Calls as an IV for Number of Crimes
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Living area
0.320***
0.286***
0.307***
0.321***
0.341***
0.348***
(0.003)
(0.006)
(0.003)
(0.003)
(0.004)
(0.005)
Lot area
0.021***
0.008***
0.013***
0.020***
0.026***
0.038***
(0.001)
(0.002)
(0.001)
(0.001)
(0.001)
(0.002)
Bedrooms
 − 0.019***
 − 0.009**
 − 0.012***
 − 0.012***
 − 0.014***
 − 0.018***
(0.002)
(0.003)
(0.002)
(0.001)
(0.001)
(0.003)
Bathrooms (Full)
0.003
 − 0.010**
0.002
0.005***
0.004**
0.010***
(0.002)
(0.004)
(0.003)
(0.002)
(0.002)
(0.003)
Age
 − 0.000***
 − 0.001***
 − 0.000***
 − 0.000***
 − 0.000***
 − 0.000
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.093***
 − 0.062***
 − 0.076***
 − 0.081***
 − 0.087***
 − 0.099***
(0.007)
(0.005)
(0.002)
(0.003)
(0.002)
(0.003)
Renovated
 − 0.046***
 − 0.275***
 − 0.107***
 − 0.006
0.025***
0.047***
(0.011)
(0.017)
(0.012)
(0.006)
(0.005)
(0.011)
Fireplace
0.040***
0.062***
0.041***
0.029***
0.020***
0.014***
(0.005)
(0.008)
(0.002)
(0.001)
(0.002)
(0.003)
Lake/Mountain view
0.105***
0.065***
0.069***
0.081***
0.104***
0.176***
(0.006)
(0.007)
(0.004)
(0.004)
(0.006)
(0.008)
Population
0.012**
 − 0.027***
0.000
0.014***
0.022***
0.027***
(0.005)
(0.008)
(0.006)
(0.004)
(0.004)
(0.007)
% Male population
0.726***
1.706***
1.043***
0.762***
0.521***
1.083***
(0.153)
(0.206)
(0.100)
(0.093)
(0.101)
(0.237)
% White population
0.793***
1.093***
0.792***
0.780***
0.773***
0.749***
(0.034)
(0.163)
(0.022)
(0.024)
(0.019)
(0.026)
% Asian population
0.043
0.080
0.011
0.088**
0.160***
0.230***
(0.032)
(0.149)
(0.023)
(0.035)
(0.022)
(0.053)
% Old population
0.283***
0.427***
0.124***
0.080***
0.078**
0.372***
(0.042)
(0.107)
(0.040)
(0.029)
(0.031)
(0.057)
Number of crimes (IV)
 − 0.828**
 − 0.117
 − 0.948***
 − 0.798***
 − 0.757***
 − 1.104***
(0.347)
(0.790)
(0.170)
(0.176)
(0.136)
(0.183)
Distance to crime hotspot
 − 0.072***
 − 0.072***
 − 0.068***
 − 0.070***
 − 0.073***
 − 0.084***
(0.003)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Constant
12.295***
11.659***
12.199***
12.274***
12.385***
12.216***
(0.078)
(0.150)
(0.055)
(0.033)
(0.047)
(0.088)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902


===== PAGE 31 (score 25, numeric-paren 0, table-words 6, var-words 13) =====

507
Crime Measures and Housing Prices: an Analysis Using Quantile…
Additional bedrooms and houses with a single floor exhibit a negative relation, since 
living area is already factored into the model. In terms of demographics, the 2SLS 
estimates are significantly positive for % White, % Male and % Old Population.
The quantile regressions in models 2 to 6 allow us to examine how these 2SLS 
results vary across the price quantiles: 0.05 (low end), 0.25, 0.50 (median), 0.75 
and 0.95 (high end). The most evident trends are increasing significance (from low 
to high prices) of: lot area and view and the decreasing significance of fireplace. 
The sign on % Asian Population is positive in the higher quantiles.When we include 
zip code fixed effects and re-run these regressions (results are reported in Appen-
dix 2 Table 11), the sign on % Asian Population becomes significant and negative in 
all quantiles.49 The coefficients on other demographic variables generally retain the 
same sign with zip code fixed effects.
Our focus however is on the impact of crime and we see a significant negative 
relation between 911 Calls and housing prices in each of the quantile regressions, 
except for the lowest priced houses, where there is no significant relation. This fact 
is consistent with the notion that increasing crime rates (as instrumented by 911 
Calls) depress housing prices. We find a significant negative relation between the 
distance to a crime hotspot and housing prices. The latter, perhaps surprising result, 
is discussed further below.
We re-estimate this regression using different crime variables.50 When we substi-
tute Number of Crimes—Violent for Number of 911 Calls, its coefficient is signifi-
cantly negative across all the quantiles. The coefficients on the demographic vari-
ables generally retain the same signs and approximately the same magnitudes as in 
Table 6.
We next use ∆ Violent Crime as a crime rate proxy and employ changes in demo-
graphic variables.51 The coefficients on the demographic variable changes again 
generally retain the same sign as in Table 6, except for the ∆ Old Population vari-
able now has a significant negative sign, suggesting that an increasingly old popula-
tion depresses prices. The coefficient on ∆ Violent Crime is significantly positive 
in all but the median quantile. The negative estimates for coefficient on distance to 
crime hotspot are virtually identical to those obtained previously.
Next we employ ∆ Social Crime. The coefficients are positive and significant in 
each quartile. In this and in the previous estimate, we find a strong negative impact 
from increasing population. This effect is dampened when we refine our estimation 
in the following tables. Furthermore, our discussions with Port of Seattle police have 
49  The inclusion of zip code fixed effects accounts for certain unobserved neighborhood characteris-
tics, one of which could be gentrification. Hwang (2020) studies Seattle’s gentrification from the 1970’s 
through 2010’s, and shows that neighborhoods with a higher Asian population experienced less gentri-
fication, while those with a higher Black population were gentrified more frequently. She explains that 
newly arriving immigrants in Seattle, who are predominantly Asian recently, tend to settle in areas with 
an existing significant Asian population. This influx can deter gentrification in such areas, potentially 
leading to lower housing prices (compared to areas which recently have been gentrified).
50  Here we only report the most relevant results for these alternative measures. The complete regressions 
are available from the authors.
51  When we use changes in proxies for crime rates, we also use changes in demographic variables.


===== PAGE 34 (score 261, numeric-paren 217, table-words 3, var-words 15) =====

510
	
J. G. Kallberg, Y. Shimizu 
Table 7   2SLS and instrumental quantile regressions – results with spatial autocorrelation adjustments
Panel A: Using Number of 911 Calls as an IV for Number of Crimes
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.394***
0.414***
0.368***
0.364***
0.364***
0.376***
(0.013)
(0.017)
(0.006)
(0.004)
(0.008)
(0.013)
Living area
0.261***
0.241***
0.252***
0.267***
0.286***
0.299***
(0.004)
(0.006)
(0.003)
(0.003)
(0.002)
(0.005)
Lot area
0.018***
0.008***
0.014***
0.018***
0.022***
0.034***
(0.001)
(0.002)
(0.000)
(0.001)
(0.001)
(0.001)
Bedrooms
 − 0.007***
 − 0.000
 − 0.002*
 − 0.002*
 − 0.002*
 − 0.008**
(0.001)
(0.004)
(0.001)
(0.001)
(0.001)
(0.004)
Bathrooms (Full)
0.008***
 − 0.007***
0.006***
0.009***
0.009***
0.015***
(0.001)
(0.002)
(0.002)
(0.002)
(0.002)
(0.002)
Age
 − 0.001***
 − 0.002***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.076***
 − 0.053***
 − 0.064***
 − 0.067***
 − 0.071***
 − 0.084***
(0.005)
(0.007)
(0.002)
(0.002)
(0.002)
(0.004)
Renovated
 − 0.028***
 − 0.264***
 − 0.102***
0.021***
0.048***
0.069***
(0.011)
(0.012)
(0.012)
(0.006)
(0.002)
(0.010)
Fireplace
0.025***
0.044***
0.025***
0.013***
0.010***
0.007***
(0.004)
(0.007)
(0.002)
(0.002)
(0.002)
(0.002)
Lake/Mountain view
0.063***
0.037***
0.036***
0.047***
0.059***
0.107***
(0.005)
(0.006)
(0.003)
(0.003)
(0.003)
(0.004)
Population
 − 0.001
 − 0.017
 − 0.005
0.001
0.003
0.013**
(0.002)
(0.013)
(0.003)
(0.004)
(0.004)
(0.005)
% Male population
 − 0.188*
 − 0.241
0.070
0.052
0.009
0.377***
(0.099)
(0.234)
(0.060)
(0.079)
(0.074)
(0.142)
% White population
0.605***
0.883***
0.626***
0.557***
0.520***
0.458***
(0.029)
(0.067)
(0.018)
(0.015)
(0.018)
(0.020)
% Asian population
0.147***
0.234**
0.102***
0.129***
0.152***
0.151***
(0.016)
(0.104)
(0.031)
(0.027)
(0.019)
(0.032)
% Old population
 − 0.167***
 − 0.246***
 − 0.184***
 − 0.206***
 − 0.227***
 − 0.222***
(0.028)
(0.053)
(0.030)
(0.025)
(0.015)
(0.032)
Number of crimes (IV)
0.730***
1.294***
0.599***
0.373***
0.214*
 − 0.111
(0.148)
(0.310)
(0.094)
(0.100)
(0.118)
(0.126)
Distance to crime hotspot
 − 0.038***
 − 0.036***
 − 0.038***
 − 0.040***
 − 0.042***
 − 0.046***
(0.002)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Constant
7.511***
6.972***
7.715***
7.855***
7.932***
7.670***
(0.180)
(0.221)
(0.076)
(0.078)
(0.076)
(0.181)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
Panel B: Using Number of Crimes—Violent Only
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.376***
0.371***
0.348***
0.350***
0.354***
0.372***
(0.012)
(0.016)
(0.006)
(0.003)
(0.007)
(0.012)


===== PAGE 35 (score 290, numeric-paren 246, table-words 2, var-words 15) =====

511
Crime Measures and Housing Prices: an Analysis Using Quantile…
Table 7   (continued)
Living area
0.260***
0.237***
0.252***
0.266***
0.285***
0.299***
(0.004)
(0.005)
(0.003)
(0.002)
(0.002)
(0.005)
Lot area
0.018***
0.005***
0.013***
0.017***
0.022***
0.034***
(0.001)
(0.002)
(0.000)
(0.001)
(0.001)
(0.001)
Bedrooms
 − 0.008***
 − 0.003
 − 0.003*
 − 0.003**
 − 0.003**
 − 0.008**
(0.001)
(0.004)
(0.001)
(0.001)
(0.001)
(0.004)
Bathrooms (Full)
0.008***
 − 0.006***
0.005*
0.009***
0.009***
0.014***
(0.001)
(0.002)
(0.003)
(0.001)
(0.002)
(0.002)
Age
 − 0.001***
 − 0.002***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.080***
 − 0.061***
 − 0.068***
 − 0.070***
 − 0.072***
 − 0.085***
(0.005)
(0.005)
(0.002)
(0.003)
(0.003)
(0.003)
Renovated
 − 0.028***
 − 0.262***
 − 0.104***
0.021***
0.048***
0.070***
(0.011)
(0.012)
(0.017)
(0.006)
(0.003)
(0.009)
Fireplace
0.025***
0.047***
0.026***
0.014***
0.010***
0.006**
(0.004)
(0.005)
(0.002)
(0.002)
(0.002)
(0.003)
Lake/Mountain view
0.063***
0.036***
0.037***
0.046***
0.060***
0.107***
(0.005)
(0.005)
(0.002)
(0.004)
(0.003)
(0.004)
Population
0.012***
0.011
0.012***
0.015***
0.009***
0.015***
(0.002)
(0.009)
(0.002)
(0.003)
(0.003)
(0.005)
% Male population
0.307***
0.787***
0.525***
0.440***
0.274***
0.465***
(0.070)
(0.128)
(0.041)
(0.047)
(0.062)
(0.106)
% White population
0.455***
0.602***
0.476***
0.454***
0.449***
0.410***
(0.023)
(0.037)
(0.017)
(0.012)
(0.014)
(0.021)
% Asian population
0.006
0.021
 − 0.026
0.041*
0.078***
0.097**
(0.028)
(0.053)
(0.029)
(0.022)
(0.018)
(0.039)
% Old population
 − 0.183***
 − 0.172***
 − 0.212***
 − 0.223***
 − 0.222***
 − 0.214***
(0.022)
(0.044)
(0.027)
(0.022)
(0.015)
(0.026)
Number of crimes—violent
 − 2.139***
 − 3.152***
 − 2.447***
 − 2.130***
 − 1.472***
 − 1.577***
(0.455)
(0.578)
(0.322)
(0.211)
(0.151)
(0.331)
Distance to crime hotspot
 − 0.040***
 − 0.040***
 − 0.039***
 − 0.041***
 − 0.043***
 − 0.046***
(0.002)
(0.002)
(0.001)
(0.000)
(0.001)
(0.001)
Constant
7.622***
7.208***
7.827***
7.884***
7.981***
7.709***
(0.177)
(0.196)
(0.076)
(0.069)
(0.065)
(0.168)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
Panel C: Using Δ Crimes—Violent
Dependent variable
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.425***
0.431***
0.395***
0.388***
0.398***
0.413***
(0.018)
(0.017)
(0.005)
(0.006)
(0.011)
(0.014)
Living area
0.268***
0.252***
0.259***
0.276***
0.288***
0.302***
(0.003)
(0.006)
(0.002)
(0.003)
(0.002)
(0.005)
Lot area
0.013***
 − 0.000
0.009***
0.012***
0.018***
0.030***
(0.001)
(0.001)
(0.000)
(0.001)
(0.001)
(0.001)
Bedrooms
 − 0.011***
 − 0.009**
 − 0.007***
 − 0.006***
 − 0.004***
 − 0.008**
(0.002)
(0.004)
(0.001)
(0.001)
(0.001)
(0.004)


===== PAGE 36 (score 281, numeric-paren 238, table-words 2, var-words 14) =====

512
	
J. G. Kallberg, Y. Shimizu 
Table 7   (continued)
Bathrooms (Full)
0.005***
 − 0.012**
0.000
0.005**
0.008***
0.016***
(0.001)
(0.005)
(0.001)
(0.002)
(0.001)
(0.003)
Age
 − 0.001***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
 − 0.000
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.080***
 − 0.066***
 − 0.068***
 − 0.067***
 − 0.071***
 − 0.085***
(0.005)
(0.005)
(0.002)
(0.002)
(0.003)
(0.004)
Renovated
 − 0.031***
 − 0.259***
 − 0.103***
0.015*
0.049***
0.065***
(0.011)
(0.010)
(0.012)
(0.008)
(0.003)
(0.011)
Fireplace
0.031***
0.061***
0.031***
0.016***
0.014***
0.010***
(0.005)
(0.006)
(0.003)
(0.002)
(0.002)
(0.003)
Lake/Mountain view
0.060***
0.039***
0.040***
0.045***
0.053***
0.093***
(0.005)
(0.004)
(0.003)
(0.003)
(0.003)
(0.005)
Δ Population
 − 6.758***
 − 7.272***
 − 8.306***
 − 6.953***
 − 6.174***
 − 7.300***
(1.261)
(1.100)
(0.488)
(0.637)
(0.737)
(0.742)
Δ Male population
0.541
 − 0.057
1.574**
0.607
0.403
2.223***
(1.205)
(1.394)
(0.632)
(0.503)
(0.497)
(0.691)
Δ White population
3.432***
3.445***
3.560***
3.744***
3.715***
3.541***
(0.952)
(0.238)
(0.218)
(0.307)
(0.263)
(0.282)
Δ Asian population
3.347***
4.176***
3.646***
3.242***
2.895***
2.525***
(0.259)
(0.234)
(0.085)
(0.062)
(0.076)
(0.141)
Δ Old population
 − 0.652**
 − 0.968***
 − 0.867***
 − 0.741***
 − 0.581***
 − 0.404***
(0.301)
(0.251)
(0.095)
(0.091)
(0.058)
(0.098)
Δ Crimes—violent
 − 0.002
0.002
0.002
0.000
 − 0.000
0.005*
(0.009)
(0.011)
(0.002)
(0.004)
(0.002)
(0.003)
Distance to crime hotspot
 − 0.031***
 − 0.028***
 − 0.032***
 − 0.032***
 − 0.034***
 − 0.037***
(0.004)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Constant
7.352***
7.077***
7.708***
7.857***
7.783***
7.663***
(0.246)
(0.204)
(0.079)
(0.082)
(0.146)
(0.180)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
Panel D: Using Δ Crimes—Social
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.424***
0.429***
0.395***
0.388***
0.397***
0.413***
(0.017)
(0.016)
(0.005)
(0.006)
(0.011)
(0.014)
Living area
0.268***
0.252***
0.259***
0.276***
0.288***
0.302***
(0.003)
(0.005)
(0.002)
(0.002)
(0.002)
(0.005)
Lot area
0.013***
 − 0.001
0.009***
0.012***
0.018***
0.030***
(0.001)
(0.001)
(0.000)
(0.001)
(0.001)
(0.001)
Bedrooms
 − 0.011***
 − 0.009**
 − 0.007***
 − 0.006***
 − 0.004***
 − 0.008**
(0.002)
(0.004)
(0.001)
(0.001)
(0.001)
(0.004)
Bathrooms (Full)
0.005***
 − 0.012**
0.000
0.005**
0.008***
0.016***
(0.001)
(0.005)
(0.001)
(0.002)
(0.001)
(0.003)
Age
 − 0.001***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
 − 0.000
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.080***
 − 0.066***
 − 0.068***
 − 0.067***
 − 0.071***
 − 0.085***
(0.005)
(0.005)
(0.002)
(0.003)
(0.003)
(0.004)


===== PAGE 37 (score 283, numeric-paren 240, table-words 2, var-words 14) =====

513
Crime Measures and Housing Prices: an Analysis Using Quantile…
Table 7   (continued)
Renovated
 − 0.031***
 − 0.260***
 − 0.103***
0.014*
0.049***
0.066***
(0.011)
(0.011)
(0.012)
(0.007)
(0.003)
(0.012)
Fireplace
0.031***
0.061***
0.031***
0.016***
0.014***
0.010***
(0.005)
(0.006)
(0.003)
(0.002)
(0.002)
(0.003)
Lake/Mountain view
0.060***
0.040***
0.040***
0.045***
0.054***
0.093***
(0.005)
(0.005)
(0.003)
(0.003)
(0.003)
(0.005)
Δ Population
 − 6.743***
 − 7.292***
 − 8.304***
 − 6.972***
 − 6.166***
 − 7.275***
(1.257)
(1.092)
(0.527)
(0.616)
(0.733)
(0.749)
Δ Male population
0.513
 − 0.079
1.570**
0.618
0.370
2.208***
(1.207)
(1.382)
(0.626)
(0.532)
(0.539)
(0.674)
Δ White population
3.440***
3.465***
3.560***
3.745***
3.724***
3.545***
(0.950)
(0.215)
(0.215)
(0.310)
(0.259)
(0.290)
Δ Asian population
3.347***
4.180***
3.645***
3.239***
2.901***
2.519***
(0.260)
(0.219)
(0.093)
(0.057)
(0.075)
(0.139)
Δ Old population
 − 0.652**
 − 0.982***
 − 0.866***
 − 0.732***
 − 0.574***
 − 0.406***
(0.299)
(0.243)
(0.088)
(0.092)
(0.062)
(0.096)
Δ Crimes—social
0.005
0.007**
0.002
0.006***
0.006***
0.004
(0.006)
(0.003)
(0.002)
(0.002)
(0.002)
(0.004)
Distance to crime hotspot
 − 0.031***
 − 0.029***
 − 0.032***
 − 0.032***
 − 0.034***
 − 0.037***
(0.004)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Constant
7.361***
7.107***
7.710***
7.863***
7.788***
7.665***
(0.243)
(0.201)
(0.079)
(0.085)
(0.145)
(0.175)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
Panel E: Using 911 Calls (Lag 1) as an IV for Number of Crimes (Lag 1)
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.398***
0.428***
0.370***
0.365***
0.363***
0.377***
(0.015)
(0.010)
(0.008)
(0.005)
(0.004)
(0.007)
Living area
0.262***
0.242***
0.253***
0.267***
0.287***
0.300***
(0.004)
(0.006)
(0.002)
(0.003)
(0.004)
(0.003)
Lot area
0.018***
0.009***
0.014***
0.018***
0.022***
0.034***
(0.001)
(0.001)
(0.000)
(0.001)
(0.001)
(0.001)
Bedrooms
 − 0.007***
0.001
 − 0.001
 − 0.001
 − 0.002
 − 0.008**
(0.001)
(0.003)
(0.002)
(0.002)
(0.003)
(0.004)
Bathrooms (Full)
0.009***
 − 0.005
0.007***
0.009***
0.009***
0.015***
(0.001)
(0.006)
(0.001)
(0.002)
(0.001)
(0.001)
Age
 − 0.001***
 − 0.002***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.075***
 − 0.050***
 − 0.064***
 − 0.066***
 − 0.071***
 − 0.083***
(0.005)
(0.005)
(0.004)
(0.003)
(0.003)
(0.004)
Renovated
 − 0.027**
 − 0.265***
 − 0.102***
0.022***
0.049***
0.072***
(0.011)
(0.015)
(0.015)
(0.006)
(0.004)
(0.014)
Fireplace
0.025***
0.043***
0.024***
0.014***
0.011***
0.006*
(0.004)
(0.006)
(0.002)
(0.003)
(0.002)
(0.003)
Lake/Mountain view
0.062***
0.038***
0.035***
0.047***
0.059***
0.103***
(0.005)
(0.008)
(0.005)
(0.004)
(0.005)
(0.012)


===== PAGE 38 (score 112, numeric-paren 90, table-words 4, var-words 5) =====

514
	
J. G. Kallberg, Y. Shimizu 
When we focus on 911 calls the situation is less uniform. In Table 6 the coef-
ficient is negative and significant, suggesting that a higher frequency of 911 calls 
depresses prices. However when we incorporate spatial correlation the sign becomes 
positive and significant. Our estimates are similar when we use lagged 911 calls by 
one year. These results suggest that more 911 calls are made in an area with higher 
house prices, once we incorporate spatial autocorrelation. One potential explana-
tion is that Wy captures characteristics specific to certain neighborhoods that are 
not observable to econometricians, such as “collective efficacy.58” Once such char-
acteristics are controlled for, an increase in the number of 911 calls can indicate 
the residents’ greater efforts to monitor the safety of their residence, to report any 
suspicious persons/activities in the area, etc. A study by Vera Institute of Justice59 
58  Collective efficacy is defined in Sampson et al. (1997) as (from the abstract): social cohesion among 
neighbors combined with their willingness to intervene on behalf of the common good. They show that 
neighborhoods with high collective efficacy tend to experience lower rates of violent crime.
59  https://​www.​vera.​org/​publi​catio​ns/​making-​911-​acces​sible
Table 7   (continued)
Population
 − 0.006**
 − 0.026**
 − 0.009**
 − 0.002
0.000
0.012***
(0.003)
(0.011)
(0.004)
(0.003)
(0.003)
(0.004)
% Male population
 − 0.339**
 − 0.674**
 − 0.018
 − 0.057
 − 0.070
0.316**
(0.142)
(0.316)
(0.131)
(0.114)
(0.126)
(0.148)
% White population
0.636***
0.922***
0.639***
0.578***
0.540***
0.467***
(0.038)
(0.063)
(0.016)
(0.016)
(0.009)
(0.016)
% Asian population
0.174***
0.220**
0.105***
0.147***
0.169***
0.167***
(0.023)
(0.104)
(0.026)
(0.026)
(0.018)
(0.028)
% Old population
 − 0.155***
 − 0.338***
 − 0.176***
 − 0.208***
 − 0.223***
 − 0.236***
(0.031)
(0.091)
(0.032)
(0.030)
(0.031)
(0.051)
Number of crimes (IV—
Lag1)
1.106***
1.773***
0.784***
0.575***
0.429***
0.019
(0.223)
(0.403)
(0.130)
(0.095)
(0.139)
(0.197)
Distance to crime hotspot
 − 0.038***
 − 0.035***
 − 0.038***
 − 0.040***
 − 0.042***
 − 0.046***
(0.002)
(0.002)
(0.000)
(0.001)
(0.001)
(0.002)
Constant
7.522***
7.001***
7.739***
7.880***
7.973***
7.670***
(0.189)
(0.162)
(0.051)
(0.050)
(0.062)
(0.105)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
58,301
58,301
58,301
58,301
58,301
58,301
This table reports the results of 2SLS regression (model 1, as estimated in Eq. 3) and instrumental quantile 
regressions (models 2 – 6, as estimated in Eq. 4) for 5, 25, 50, 75 and 95 percent quantiles, respectively. Defini-
tions of the variables are provided in Appendix 1 Table 10. Panels A and B use Number of crimes (instrumented 
by Number of 911 calls) and Number of crimes-violent), respectively, as well as levels of the demographic vari-
ables (Population, % Male, % White, % Asian and % Old population) as control variables. Panels C and D 
use annual changes for Number of crimes- violent and Number of crimes-social, respectively, as well as for the 
demographic variables. Panel E uses lagged values of Number of crimes (instrumented by lagged Number of 
911 calls). Standard errors are reported in parentheses: for 2SLS regressions (model 1), standard errors are clus-
tered by year and month. For quantile regressions (models 2 – 6), standard errors are obtained through bootstrap 
replications. *, ** and *** denote significance at the 10, 5 and 1 percent levels, respectively


===== PAGE 39 (score 25, numeric-paren 0, table-words 8, var-words 9) =====

515
Crime Measures and Housing Prices: an Analysis Using Quantile…
corroborates this perspective: among different types of 911 calls made in New 
Orleans, they report that 911 calls made for property crimes and those for “suspi-
cious behaviors” are the two predominant call types, followed by domestic violence, 
violent crime, drugs and sex offenses. This trend is observed in both census tracts 
with the lowest and highest 10 percent median income. Thus, the positive relation 
between the frequency of 911 calls and housing prices (once we incorporate spa-
tial autocorrelation) can be indicative of residents’ collective efficacy or criminals’ 
attempts to target higher priced properties.
Robustness
Our first robustness check restricts our 911 Calls to only those with high priority. 
Each 911 call is assigned a priority level, ranging from 1 (highest) to 9 (lowest), 
based on the severity of an event. 911 Calls (High Priority) are defined as those with 
the two highest priority levels. In our data, this subsample represents 68.1% of the 
1,528,303 calls in our total sample. By excluding calls of less severity, this measure 
of crime levels can hopefully become more accurate.
Table 8 Panel A presents our basic regression without adjustment for spatial auto-
correlation. Here the coefficient on 911 Calls (High Priority) is negative and signifi-
cant in each quantile (except for the lowest quantile), with higher values than those 
reported in the previous tables, suggesting that this measure is a more efficient proxy 
for crime rates than the total of 911 Calls. Panel B now include the Wy variable. As in 
Tables 6 and 7, we now observe positive and significant estimates for 911 Calls (High 
Priority) in the lowest 3 price quantiles. This flip from negative to positive signs on 
the crime proxies when we include spatial autocorrelation parallels our earlier results. 
The positive sign is more consistent with our estimates involving distance to crime 
hotspots, which could suggest that ignoring spatial autocorrelation can lead to spuri-
ous estimates. Very similar results obtain when we substitute lagged values of 911 
Calls (High Priority) for results with and without spatial autocorrelation adjustments.
Our next robustness check is presented in Table 9, which adds distance to crime 
coldspot (Distance to Crime Coldspot) to the estimation. The rationale is that con-
sumers should be willing to pay more for residences near low crime areas. Table 9 
Panel A presents the results using 911 calls as an instrument without correcting for 
spatial correlation. The estimates are very similar to those reported in Table 6 with 
the additional observation that Distance to Crime Coldspot has a weakly negative 
impact; its coefficient is small but significant. When we repeat this analysis using 
Number of Crimes – Violent, ∆ Violent Crime, ∆ Social Crime and 911 Calls (Lag 
1), we generate very similar estimates for the Distance to Crime Coldspot variable.
Table  9 Panel B now incorporates the Wy variable and uses 911 Calls as the 
proxy for crime rates. We observe the same pattern as in Panel A: the estimates on 
Distance to Crime Coldspot are negative and significant in each quantile, however 
the average coefficient is only 45% of the value obtained in Panel A. Overall these 
estimates provide consistent evidence of a small positive pricing impact of being 
closer to a crime coldspot. When we re-estimate this regression using the other 4 
crime rate proxies the results are qualitatively similar.


===== PAGE 41 (score 262, numeric-paren 219, table-words 3, var-words 14) =====

517
Crime Measures and Housing Prices: an Analysis Using Quantile…
Table 8   Robustness check – quantile regression results using 911 calls with high priority
Panel A: Using 911 Calls (High Priority), without spatial autocorrelation adjustments
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Living area
0.318***
0.284***
0.306***
0.320***
0.340***
0.347***
(0.004)
(0.006)
(0.003)
(0.003)
(0.004)
(0.006)
Lot area
0.021***
0.007***
0.013***
0.019***
0.026***
0.038***
(0.001)
(0.002)
(0.001)
(0.001)
(0.001)
(0.002)
Bedrooms
 − 0.019***
 − 0.008**
 − 0.012***
 − 0.012***
 − 0.014***
 − 0.019***
(0.002)
(0.004)
(0.002)
(0.001)
(0.001)
(0.003)
Bathrooms (Full)
0.003
 − 0.010***
0.002
0.006***
0.004*
0.009***
(0.002)
(0.004)
(0.003)
(0.002)
(0.002)
(0.003)
Age
 − 0.000***
 − 0.001***
 − 0.000***
 − 0.000***
 − 0.000**
 − 0.000
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.095***
 − 0.064***
 − 0.077***
 − 0.082***
 − 0.088***
 − 0.100***
(0.008)
(0.005)
(0.002)
(0.003)
(0.003)
(0.003)
Renovated
 − 0.046***
 − 0.269***
 − 0.107***
 − 0.007
0.026***
0.049***
(0.011)
(0.017)
(0.012)
(0.006)
(0.004)
(0.010)
Fireplace
0.040***
0.059***
0.042***
0.029***
0.021***
0.013***
(0.005)
(0.008)
(0.002)
(0.001)
(0.001)
(0.003)
Lake/Mountain view
0.104***
0.066***
0.069***
0.080***
0.103***
0.173***
(0.005)
(0.007)
(0.004)
(0.003)
(0.006)
(0.007)
Population
0.017**
 − 0.019**
0.004
0.017***
0.025***
0.030***
(0.007)
(0.009)
(0.005)
(0.004)
(0.004)
(0.007)
% Male population
0.907***
1.745***
1.120***
0.870***
0.647***
1.337***
(0.229)
(0.183)
(0.102)
(0.088)
(0.091)
(0.209)
% White population
0.747***
0.978***
0.762***
0.756***
0.751***
0.728***
(0.042)
(0.162)
(0.021)
(0.024)
(0.018)
(0.020)
% Asian population
0.008
0.007
 − 0.002
0.077**
0.146***
0.232***
(0.053)
(0.130)
(0.022)
(0.032)
(0.023)
(0.040)
% Old population
0.268***
0.381***
0.091***
0.059**
0.065*
0.416***
(0.038)
(0.083)
(0.032)
(0.027)
(0.035)
(0.054)
Number of crimes (IV)
 − 1.255**
 − 0.900
 − 1.298***
 − 1.062***
 − 0.994***
 − 1.371***
(0.534)
(0.776)
(0.185)
(0.163)
(0.109)
(0.163)
Distance to crime hotspot
 − 0.073***
 − 0.073***
 − 0.068***
 − 0.070***
 − 0.074***
 − 0.085***
(0.003)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Constant
12.241***
11.742***
12.183***
12.235***
12.341***
12.109***
(0.102)
(0.136)
(0.053)
(0.035)
(0.050)
(0.071)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
Panel B: Using 911 Calls (High Priority), with spatial autocorrelation adjustments
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.389***
0.402***
0.363***
0.360***
0.361***
0.373***
(0.012)
(0.017)
(0.007)
(0.003)
(0.007)
(0.013)
Living area
0.261***
0.240***
0.252***
0.266***
0.285***
0.300***
(0.004)
(0.006)
(0.003)
(0.003)
(0.002)
(0.005)


===== PAGE 42 (score 222, numeric-paren 183, table-words 3, var-words 14) =====

518
	
J. G. Kallberg, Y. Shimizu 
Table 8   (continued)
Lot area
0.018***
0.007***
0.013***
0.018***
0.022***
0.033***
(0.001)
(0.002)
(0.000)
(0.001)
(0.001)
(0.001)
Bedrooms
 − 0.007***
 − 0.001
 − 0.002*
 − 0.002**
 − 0.002*
 − 0.009**
(0.001)
(0.004)
(0.001)
(0.001)
(0.001)
(0.004)
Bathrooms (Full)
0.008***
 − 0.007***
0.006***
0.009***
0.009***
0.014***
(0.001)
(0.002)
(0.002)
(0.002)
(0.002)
(0.002)
Age
 − 0.001***
 − 0.002***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.077***
 − 0.055***
 − 0.065***
 − 0.067***
 − 0.072***
 − 0.085***
(0.006)
(0.007)
(0.002)
(0.002)
(0.002)
(0.003)
Renovated
 − 0.028***
 − 0.264***
 − 0.102***
0.021***
0.048***
0.069***
(0.011)
(0.013)
(0.013)
(0.007)
(0.002)
(0.011)
Fireplace
0.025***
0.045***
0.025***
0.013***
0.010***
0.006**
(0.004)
(0.007)
(0.002)
(0.002)
(0.002)
(0.002)
Lake/Mountain view
0.062***
0.035***
0.037***
0.046***
0.059***
0.106***
(0.005)
(0.006)
(0.003)
(0.003)
(0.003)
(0.004)
Population
0.003
 − 0.011
 − 0.001
0.005
0.006*
0.014***
(0.003)
(0.014)
(0.003)
(0.004)
(0.003)
(0.005)
% Male population
 − 0.048
0.097
0.171***
0.144**
0.115
0.496***
(0.098)
(0.226)
(0.065)
(0.068)
(0.088)
(0.164)
% White population
0.575***
0.844***
0.606***
0.543***
0.502***
0.443***
(0.020)
(0.072)
(0.018)
(0.014)
(0.018)
(0.018)
% Asian population
0.122***
0.217**
0.088***
0.120***
0.135***
0.138***
(0.018)
(0.092)
(0.030)
(0.025)
(0.019)
(0.029)
% Old population
 − 0.173***
 − 0.187***
 − 0.188***
 − 0.213***
 − 0.228***
 − 0.203***
(0.025)
(0.068)
(0.025)
(0.026)
(0.017)
(0.034)
Number of crimes (IV)
0.409**
0.943**
0.386***
0.188**
0.016
 − 0.285**
(0.191)
(0.371)
(0.109)
(0.088)
(0.131)
(0.115)
Distance to crime hotspot
 − 0.039***
 − 0.037***
 − 0.038***
 − 0.040***
 − 0.042***
 − 0.047***
(0.002)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Constant
7.530***
6.979***
7.728***
7.854***
7.934***
7.668***
(0.176)
(0.202)
(0.085)
(0.078)
(0.073)
(0.175)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
This table reports the results of 2SLS regression (model 1, as estimated in Eq. 1 and Eq. 3) and instru-
mental quantile regressions (models 2 – 6, as estimated in Eqs. 2 and 4) for 5, 25, 50, 75 and 95 per-
cent quantiles, respectively. Number of crimes is instrumented by Number of 911 calls (High Priority), 
defined as the number of 911 calls with priority levels of 1 (highest) or 2 (second highest), made in a 
given year per police beat. Panel A (B) performs 2SLS and quantile regressions without (with) spatial 
autocorrelation adjustments. Definitions of other variables are provided in Appendix 1 Table 10. Stand-
ard errors are reported in parentheses: for 2SLS regressions (model 1), standard errors are clustered by 
year and month. For quantile regressions (models 2 – 6), standard errors are obtained through bootstrap 
replications. *, ** and *** denote significance at the 10, 5 and 1 percent levels, respectively


===== PAGE 43 (score 262, numeric-paren 219, table-words 3, var-words 14) =====

519
Crime Measures and Housing Prices: an Analysis Using Quantile…
Table 9   Robustness check – quantile regression results including distance to crime coldspot
Panel A: Using Number of 911 Calls as an IV, without spatial autocorrelation adjustments
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Living area
0.318***
0.285***
0.306***
0.321***
0.339***
0.348***
(0.003)
(0.005)
(0.002)
(0.003)
(0.003)
(0.005)
Lot area
0.022***
0.009***
0.014***
0.020***
0.026***
0.038***
(0.001)
(0.002)
(0.001)
(0.001)
(0.001)
(0.002)
Bedrooms
 − 0.019***
 − 0.008*
 − 0.011***
 − 0.013***
 − 0.014***
 − 0.018***
(0.002)
(0.005)
(0.002)
(0.002)
(0.002)
(0.003)
Bathrooms (Full)
0.004*
 − 0.009**
0.003
0.006***
0.005**
0.011***
(0.002)
(0.004)
(0.002)
(0.002)
(0.002)
(0.004)
Age
 − 0.000***
 − 0.001***
 − 0.000***
 − 0.000***
 − 0.000**
 − 0.000
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.092***
 − 0.063***
 − 0.073***
 − 0.079***
 − 0.087***
 − 0.099***
(0.007)
(0.005)
(0.003)
(0.003)
(0.003)
(0.005)
Renovated
 − 0.046***
 − 0.278***
 − 0.110***
 − 0.003
0.023***
0.046***
(0.011)
(0.023)
(0.014)
(0.008)
(0.007)
(0.013)
Fireplace
0.039***
0.062***
0.038***
0.027***
0.020***
0.013***
(0.005)
(0.007)
(0.002)
(0.002)
(0.001)
(0.002)
Lake/Mountain view
0.109***
0.066***
0.074***
0.085***
0.106***
0.180***
(0.006)
(0.007)
(0.004)
(0.004)
(0.006)
(0.009)
Population
0.014***
 − 0.019
0.001
0.016***
0.022***
0.026***
(0.005)
(0.013)
(0.006)
(0.002)
(0.003)
(0.005)
% Male population
0.614***
1.484***
0.858***
0.629***
0.452***
1.008***
(0.133)
(0.272)
(0.115)
(0.117)
(0.079)
(0.241)
% White population
0.776***
1.106***
0.801***
0.754***
0.758***
0.738***
(0.037)
(0.154)
(0.023)
(0.017)
(0.021)
(0.023)
% Asian population
0.020
0.111
 − 0.000
0.067***
0.136***
0.221***
(0.032)
(0.107)
(0.024)
(0.016)
(0.018)
(0.046)
% Old population
0.329***
0.427***
0.196***
0.119***
0.127***
0.376***
(0.041)
(0.101)
(0.037)
(0.021)
(0.027)
(0.045)
Number of crimes (IV)
 − 0.694**
0.093
 − 0.588***
 − 0.713***
 − 0.659***
 − 1.007***
(0.305)
(0.744)
(0.164)
(0.097)
(0.120)
(0.202)
Distance to crime hotspot
 − 0.072***
 − 0.071***
 − 0.067***
 − 0.069***
 − 0.073***
 − 0.084***
(0.003)
(0.002)
(0.001)
(0.001)
(0.001)
(0.002)
Distance to crime coldspot
 − 0.010***
 − 0.011***
 − 0.011***
 − 0.009***
 − 0.008***
 − 0.006***
(0.001)
(0.001)
(0.000)
(0.001)
(0.001)
(0.001)
Constant
12.359***
11.687***
12.260***
12.353***
12.443***
12.282***
(0.067)
(0.196)
(0.090)
(0.057)
(0.046)
(0.078)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,898
61,898
61,898
61,898
61,898
61,898
Panel B: Using Number of 911 Calls as an IV, with spatial autocorrelation adjustments
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.389***
0.409***
0.364***
0.359***
0.360***
0.371***
(0.013)
(0.014)
(0.009)
(0.008)
(0.007)
(0.011)


===== PAGE 44 (score 244, numeric-paren 202, table-words 3, var-words 15) =====

520
	
J. G. Kallberg, Y. Shimizu 
Table 9   (continued)
Living area
0.260***
0.239***
0.250***
0.267***
0.285***
0.299***
(0.004)
(0.005)
(0.002)
(0.002)
(0.003)
(0.005)
Lot area
0.019***
0.008***
0.014***
0.018***
0.022***
0.034***
(0.001)
(0.001)
(0.001)
(0.000)
(0.001)
(0.002)
Bedrooms
 − 0.007***
0.000
 − 0.001
 − 0.002
 − 0.003*
 − 0.008***
(0.001)
(0.005)
(0.002)
(0.001)
(0.001)
(0.003)
Bathrooms (Full)
0.009***
 − 0.005
0.007***
0.010***
0.009***
0.016***
(0.001)
(0.004)
(0.002)
(0.001)
(0.002)
(0.003)
Age
 − 0.001***
 − 0.002***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.076***
 − 0.054***
 − 0.063***
 − 0.066***
 − 0.072***
 − 0.084***
(0.005)
(0.004)
(0.002)
(0.003)
(0.003)
(0.004)
Renovated
 − 0.029***
 − 0.265***
 − 0.104***
0.020***
0.048***
0.070***
(0.011)
(0.016)
(0.015)
(0.007)
(0.007)
(0.011)
Fireplace
0.025***
0.043***
0.024***
0.012***
0.010***
0.007***
(0.004)
(0.006)
(0.003)
(0.002)
(0.002)
(0.002)
Lake/Mountain view
0.065***
0.040***
0.039***
0.048***
0.063***
0.110***
(0.005)
(0.005)
(0.003)
(0.004)
(0.004)
(0.008)
Population
0.001
 − 0.015
 − 0.003
0.004
0.004
0.014***
(0.002)
(0.014)
(0.005)
(0.003)
(0.004)
(0.005)
% Male population
 − 0.217**
 − 0.322
0.010
0.036
 − 0.003
0.332***
(0.094)
(0.249)
(0.126)
(0.077)
(0.100)
(0.095)
% White population
0.592***
0.873***
0.618***
0.541***
0.507***
0.453***
(0.030)
(0.067)
(0.020)
(0.012)
(0.017)
(0.019)
% Asian population
0.129***
0.210**
0.090***
0.107***
0.133***
0.140***
(0.018)
(0.092)
(0.027)
(0.017)
(0.021)
(0.022)
% Old population
 − 0.138***
 − 0.224***
 − 0.157***
 − 0.175***
 − 0.205***
 − 0.211***
(0.032)
(0.076)
(0.041)
(0.019)
(0.020)
(0.041)
Number of crimes (IV)
0.734***
1.338***
0.612***
0.354***
0.195**
 − 0.072
(0.148)
(0.234)
(0.167)
(0.111)
(0.096)
(0.126)
Distance to crime hotspot
 − 0.039***
 − 0.036***
 − 0.038***
 − 0.040***
 − 0.042***
 − 0.046***
(0.002)
(0.002)
(0.001)
(0.001)
(0.001)
(0.001)
Distance to crime coldspot
 − 0.006***
 − 0.007***
 − 0.006***
 − 0.005***
 − 0.005***
 − 0.004***
(0.001)
(0.001)
(0.000)
(0.000)
(0.000)
(0.001)
Constant
7.602***
7.085***
7.801***
7.929***
8.008***
7.761***
(0.174)
(0.251)
(0.115)
(0.102)
(0.129)
(0.176)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,898
61,898
61,898
61,898
61,898
61,898
This table reports the results of 2SLS regressions, as well as those of quantile regressions by adding the 
Distance to crime coldspot variable to the estimations. Panel A (B) performs 2SLS and quantile regres-
sions without (with) spatial autocorrelation adjustments. Distance to crime coldspot is defined as the 
straight mile distance to the nearest crime coldspot, as identified by the Getis-Ord statistics. Definitions 
of other variables are provided in Appendix 1 Table 10. Standard errors are reported in parentheses: for 
2SLS regressions (model 1), standard errors are clustered by year and month. For quantile regressions 
(models 2 – 6), standard errors are obtained through bootstrap replications. *, ** and *** denote signifi-
cance at the 10, 5 and 1 percent levels, respectively


===== PAGE 45 (score 21, numeric-paren 0, table-words 2, var-words 17) =====

521
Crime Measures and Housing Prices: an Analysis Using Quantile…
These limitations suggest several topics for further research, in particular 
the impact of policing efforts. As noted, enforcement levels are related to crime 
rates and to the frequency of 911 calls, complicating the already intricate rela-
tions between crime and the pricing of real estate. In addition, the link between 
actual crimes and 911calls requires additional analysis.
Appendix 1
````
Table 10   Variable definitions
Variable
Definition
Transaction prices
  Sale price
Sale price of property ($)
  Log (Sale Price)
Log of sale price
  WY
The weighted average sale price ($) of neighbor properties 
within 2000 feet
  Log (WY)
Log of WY
Property-level characteristics
  Living area
Total living area in ­ft2
  Lot area
Total lot area in ­ft2
  Bedrooms
# of bedrooms
  Bathrooms
# of full-size bathrooms
  One story
A dummy variable equal to 1 for one-story property; zero 
otherwise
  Year built
Year when property was built
  Renovated
A dummy variable equal to 1 if previously renovated; zero 
otherwise
  Age
Year of sale—Year built
  Fireplace
A dummy variable equal to 1 for property with fireplace; zero 
otherwise
Environmental—natual views
  Lake/Mountain view
A dummy variable equal to 1 for property with natural view (Mt 
Rainier, Cascades, Puget Sound, Lake Washington, Olympics, 
Others); zero otherwise
Crime variables
  Number of crimes
Number of reported crimes per census tract, divided by popula-
tion per census tract
  Change in crime #
% change in number of reported crime from 
previous year (t-1) per census tract, or 
log
(
1 + number of crimest
) −log(1 + number of crimest−1
)


===== PAGE 47 (score 254, numeric-paren 212, table-words 3, var-words 14) =====

523
Crime Measures and Housing Prices: an Analysis Using Quantile…
Appendix 2 
Table 11   Baseline results with zip code fixed effects
Panel A: Without spatial autocorrelation adjustments
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Living area
0.288***
0.255***
0.280***
0.294***
0.306***
0.312***
(0.004)
(0.006)
(0.003)
(0.002)
(0.003)
(0.004)
Lot area
0.026***
0.016***
0.021***
0.024***
0.030***
0.041***
(0.001)
(0.001)
(0.001)
(0.001)
(0.001)
(0.002)
Bedrooms
 − 0.016***
 − 0.003
 − 0.009***
 − 0.010***
 − 0.011***
 − 0.017***
(0.002)
(0.003)
(0.002)
(0.002)
(0.001)
(0.003)
Bathrooms (Full)
0.007***
0.001
0.006***
0.008***
0.008***
0.013***
(0.002)
(0.004)
(0.002)
(0.002)
(0.003)
(0.004)
Age
 − 0.001***
 − 0.001***
 − 0.001***
 − 0.001***
 − 0.000***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.076***
 − 0.053***
 − 0.058***
 − 0.065***
 − 0.075***
 − 0.086***
(0.006)
(0.006)
(0.003)
(0.003)
(0.004)
(0.004)
Renovated
 − 0.033***
 − 0.265***
 − 0.106***
0.018**
0.041***
0.052***
(0.011)
(0.015)
(0.010)
(0.009)
(0.004)
(0.009)
Fireplace
0.020***
0.036***
0.021***
0.014***
0.006***
 − 0.006**
(0.005)
(0.005)
(0.002)
(0.002)
(0.001)
(0.003)
Lake/Mountain view
0.140***
0.094***
0.107***
0.114***
0.138***
0.211***
(0.006)
(0.004)
(0.003)
(0.003)
(0.006)
(0.010)
Population
0.073***
0.080***
0.075***
0.071***
0.073***
0.087***
(0.008)
(0.007)
(0.005)
(0.003)
(0.006)
(0.014)
% Male population
0.355*
0.304*
0.233
0.296**
0.627***
1.935***
(0.188)
(0.162)
(0.145)
(0.127)
(0.114)
(0.283)
% White population
0.261***
0.262***
0.285***
0.250***
0.242***
0.104
(0.043)
(0.083)
(0.032)
(0.032)
(0.036)
(0.065)
% Asian population
 − 0.493***
 − 0.437***
 − 0.399***
 − 0.418***
 − 0.405***
 − 0.576***
(0.060)
(0.078)
(0.052)
(0.033)
(0.028)
(0.061)
% Old population
0.664***
0.246***
0.283***
0.493***
0.678***
1.148***
(0.060)
(0.054)
(0.029)
(0.048)
(0.031)
(0.183)
Number of crimes (IV)
 − 1.849***
 − 3.070***
 − 2.646***
 − 2.263***
 − 1.898***
 − 1.870***
(0.376)
(0.515)
(0.185)
(0.180)
(0.209)
(0.336)
Distance to crime hotspot
 − 0.071***
 − 0.082***
 − 0.074***
 − 0.071***
 − 0.068***
 − 0.065***
(0.004)
(0.005)
(0.001)
(0.001)
(0.002)
(0.002)
Constant
12.432***
12.398***
12.487***
12.507***
12.330***
11.711***
(0.113)
(0.097)
(0.074)
(0.058)
(0.059)
(0.212)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
Zip area dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
Panel B: With spatial autocorrelation adjustments
Dependent variable:
(1)
(2)
(3)
(4)
(5)
(6)
Log (Sale Price)
2SLS
Q: 0.05
Q: 0.25
Q: 0.50
Q: 0.75
Q: 0.95
Log (WY)
0.292***
0.282***
0.271***
0.265***
0.262***
0.266***
(0.013)
(0.018)
(0.009)
(0.006)
(0.010)
(0.017)


===== PAGE 48 (score 230, numeric-paren 191, table-words 2, var-words 15) =====

524
	
J. G. Kallberg, Y. Shimizu 
Table 11   (continued)
Living area
0.256***
0.231***
0.249***
0.265***
0.281***
0.290***
(0.004)
(0.005)
(0.002)
(0.002)
(0.002)
(0.004)
Lot area
0.023***
0.012***
0.018***
0.022***
0.027***
0.038***
(0.001)
(0.001)
(0.001)
(0.001)
(0.000)
(0.001)
Bedrooms
 − 0.008***
 − 0.001
 − 0.002**
 − 0.003*
 − 0.004***
 − 0.008***
(0.001)
(0.004)
(0.001)
(0.002)
(0.001)
(0.003)
Bathrooms (Full)
0.009***
0.000
0.008***
0.010***
0.009***
0.015***
(0.001)
(0.004)
(0.002)
(0.002)
(0.002)
(0.002)
Age
 − 0.001***
 − 0.002***
 − 0.001***
 − 0.001***
 − 0.001***
 − 0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
One story
 − 0.070***
 − 0.054***
 − 0.058***
 − 0.060***
 − 0.067***
 − 0.080***
(0.005)
(0.005)
(0.003)
(0.003)
(0.002)
(0.004)
Renovated
 − 0.025**
 − 0.258***
 − 0.102***
0.028***
0.053***
0.067***
(0.011)
(0.013)
(0.014)
(0.008)
(0.003)
(0.007)
Fireplace
0.016***
0.032***
0.015***
0.008***
0.002
 − 0.001
(0.004)
(0.005)
(0.002)
(0.002)
(0.002)
(0.004)
Lake/Mountain view
0.095***
0.059***
0.066***
0.076***
0.094***
0.158***
(0.005)
(0.005)
(0.003)
(0.002)
(0.004)
(0.006)
Population
0.031***
0.056***
0.045***
0.036***
0.028***
0.029***
(0.004)
(0.017)
(0.003)
(0.006)
(0.007)
(0.010)
% Male population
 − 0.315***
0.089
 − 0.090
 − 0.137
 − 0.061
0.210
(0.104)
(0.100)
(0.111)
(0.089)
(0.117)
(0.273)
% White population
0.359***
0.270***
0.309***
0.309***
0.334***
0.299***
(0.023)
(0.076)
(0.026)
(0.036)
(0.021)
(0.045)
% Asian population
 − 0.149***
 − 0.200***
 − 0.159***
 − 0.151***
 − 0.122***
 − 0.166***
(0.034)
(0.060)
(0.033)
(0.027)
(0.032)
(0.046)
% Old population
0.098**
0.009
0.023
0.088***
0.143***
0.198**
(0.039)
(0.054)
(0.031)
(0.026)
(0.034)
(0.094)
Number of crimes (IV)
 − 0.055
 − 1.430***
 − 0.956***
 − 0.632***
 − 0.249**
 − 0.124
(0.192)
(0.521)
(0.196)
(0.195)
(0.099)
(0.274)
Distance to crime hotspot
 − 0.045***
 − 0.056***
 − 0.051***
 − 0.048***
 − 0.046***
 − 0.044***
(0.004)
(0.004)
(0.002)
(0.001)
(0.001)
(0.002)
Constant
8.929***
8.748***
9.057***
9.237***
9.280***
9.186***
(0.197)
(0.267)
(0.147)
(0.083)
(0.107)
(0.120)
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Month dummies
Yes
Yes
Yes
Yes
Yes
Yes
Zip area dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
61,902
61,902
61,902
61,902
61,902
61,902
This table reports the results of baseline regressions (as estimated in Eq.  1 – 4) with zip code fixed 
effects. Panel A reports the results from Eq. 1 (2SLS) and 2 (quantile regressions) without spatial auto-
correlation adjustments. Panel B reports the results from Eq. 3 (2SLS) and 4 (instrumental variable quan-
tile regressions) with spatial autocorrelation adjustments. Standard errors are reported in parentheses: for 
2SLS regressions (model 1), standard errors are clustered by year and month. For quantile regressions 
(models 2 – 6), standard errors are obtained through bootstrap replications. *, ** and *** denote signifi-
cance at the 10, 5 and 1 percent levels, respectively


===== PAGE 49 (score 7, numeric-paren 0, table-words 0, var-words 7) =====

525
Crime Measures and Housing Prices: an Analysis Using Quantile…
Acknowledgements  We are very grateful to the editor, R. Kelley Pace, and to an anonymous referee 
for suggestions that have significantly improved this paper. We would like to thank Geoffrey Bell, Chris 
Cunningham, Eric DeVries (Port of Seattle Police Department), D. J. Fairhurst, Andrew Fluharty, Joseph 
Grant, Chinmoy Ghosh, Shinichi Hirota, Toshiki Honda, Peggy Kallberg, Jin Man Lee, Dahui Li, Jiawen 
Liu, Daisuke Miyakawa, Hidetoshi Nakagawa, Nobuhiro Nakamura, Mikiharu Noma, Kazuhiko Ohashi, 
Yusuke Osaki, Liang Peng, Katsushi Suzuki, Daisuke Yokouchi, Wenqing Zhang and seminar and con-
ference participants at the 2022 AsRes-AREUEA Conference in Tokyo, the 2022 FMA Conference in 
Atlanta, Hitotsubashi University, University of Massachusetts, Dartmouth, University of Minnesota, 
Duluth, Waseda University, and Washington State University for helpful comments.
Funding  Open access funding provided by Hitotsubashi University. The authors did not receive support 
from any organization for the submitted work.
Data Availability  The data analyzed in this study come from the King County Assessor’s website (https://​
kingc​ounty.​gov/​depts/​asses​sor.​aspx), the Washington State Office of Financial Management website (https://​
ofm.​wa.​gov/), the Seattle Police Department website (https://​www.​seatt​le.​gov/​police), and the Seattle Open 
Data Portal (https://​data.​seatt​le.​gov/). The geographic data were analyzed using Maptitude Mapping Soft-
ware (Caliper Corporation), GeoDa (https://​geoda​center.​github.​io/), and Texas A&M University Geoser-
vices (https://​geose​rvices.​tamu.​edu/). The data from Caliper Corporation were used under license.
Code Availability  Not applicable.
Declarations 
Conflicts of Interests/Competing Interests  The authors have no relevant financial or non-financial interests 
to disclose.
Open Access  This article is licensed under a Creative Commons Attribution 4.0 International License, 
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long 
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative 
Commons licence, and indicate if changes were made. The images or other third party material in this 
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line 
to the material. If material is not included in the article’s Creative Commons licence and your intended 
use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permis-
sion directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.
References
Abbott, J. K., & Klaiber, H. A. (2011). An embarrassment of riches: Confronting omitted variable 
bias and multi-scale capitalization in hedonic price models. Review of Economics and Statistics, 
93, 1331–1342. https://​doi.​org/​10.​1162/​REST_a_​00134
Andresen, M. A., & Malleson, N. (2013). Crime seasonality and its variations across space. Applied 
Geography, 43, 25–35. https://​doi.​org/​10.​1016/j.​apgeog.​2013.​06.​007
Anselin, L., Cohen, J., Cook, D., Gorr, W., & Tita, G. (2000). Spatial analyses of crime. Criminal 
Justice, 4, 213–262.
Becker, G. S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76, 
169–217.
Boggess, L. N., Greenbaum, R. T., & Tita, G. E. (2013). Does crime drive housing sales? Evidence 
from Los Angeles. Journal of Crime and Justice, 36, 299–318. https://​doi.​org/​10.​1080/​07356​
48X.​2013.​812976
Braakmann, N. (2012). The link between non-property crime and house prices: Evidence from UK 
street-level data. MPRA Paper No. 44884.