1# RECORD d3812508998e06d0728e9af8b9c2c72e4dad465a2# TITLE: Crime Measures and Housing Prices: an Analysis Using Quantile Regression and Spatial Autocorrelation3# YEAR: 2025.0 VENUE: Journal of real estate finance and economics DOI: 10.1007/s11146-024-09997-w4# AUTHORS: J. Kallberg; Yoshikiyo Shimizu5# PAGES: 53 CANDIDATE TABLE PAGES (1-based): [18, 26, 27, 28, 29, 30, 34, 35, 36, 37, 38, 41, 42, 43, 44, 47, 48]6# 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.789===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) =====1011Vol.:(0123456789)12The Journal of Real Estate Finance and Economics (2025) 72:477–52913https://doi.org/10.1007/s11146-024-09997-w14Crime Measures and Housing Prices: an Analysis Using 15Quantile Regression and Spatial Autocorrelation16Jarl G. Kallberg1 · Yoshiki Shimizu2 17Accepted: 18 August 2024 / Published online: 30 January 2025 18© The Author(s) 202519Abstract20Crime is a disamenity, so buyers should be willing to pay more for a house (all else 21equal) in a low crime area, suggesting that high crime rates depress housing prices. 22Conversely, it is plausible that criminals prefer wealthier areas because of the higher 23expected returns from their transgressions. This study examines the link between 24measures of crime and prices of residential housing. Our data begin in 2008 and 25end in 2020 for Seattle, Washington, including all reported felonies (756,304); 911 26calls (1,528,303); all recorded residential real estate transactions (61,902 after filter-27ing), as well as the corresponding property characteristics; demographic data and 28the associated changes. The inherent endogeneity between crime rates and hous-29ing prices forces us to find an instrument for crime rates. After rejecting several 30plausible choices, based on the Wu-Hausman test, the key variable in our empiri-31cal analysis is the number of 911 calls (contemporaneous and lagged) in a given 32beat. Our results present somewhat mixed evidence on the impact of crime on hous-33ing prices. Without adjustment for spatial autocorrelation, a 1 percentage point 34increase in crime rates (instrumented by the number of 911 calls) leads to approxi-35mately a 0.55% decrease in house prices. However adjusting for spatial autocorre-36lation changes this figure to a 0.80% increase. We further show that distance to a 37crime hotspot is significantly negatively related to housing prices, suggesting that 38criminals choose to operate in wealthier areas, which is consistent with our findings 39incorporating spatial autocorrelation.40Keywords Crime rates · Crime hotspots · Hedonic models · Quantile regression · 41Spatial autocorrelation · Wu-Hausman test42 * Yoshiki Shimizu 43 44yshimizu@hub.hit-u.ac.jp451 46Carson College of Business, Washington State University, Pullman, WA 99164‑4746, USA472 48Hitotsubashi University Business School (HUB), Hitotsubashi University, Tokyo 101‑8439, 49Japan505152===== PAGE 18 (score 35, numeric-paren 0, table-words 3, var-words 14) =====535449455 56J. G. Kallberg, Y. Shimizu 57Table 2 Summary statistics58N59Mean60Median61Std Dev62Min63Max64Property-level characteristics65 Sale price6661,90267706,11068617,50069457,3437010,0007114,275,00072 WY7361,90274657,44975592,11976324,9967774,027786,257,62579 Living area8061,902811.891821.720830.792840.350854.59086 Lot area8761,902884.660894.650902.636910.3759214.46693 Bedrooms9461,902953.192963.000970.921980.0009911.000100 Bathrooms10161,9021021.4851031.0001040.6891050.0001067.000107 One story10861,9021090.3981100.0001110.4901120.0001131.000114 Age11561,90211655.57111765.00011838.3991190.000120114.000121 Renovated12261,9021230.0481240.0001250.2141260.0001271.000128 FirePlace12961,9021300.7121311.0001320.4531330.0001341.000135Environmental—natural views136 Lake/Mountain view13761,9021380.1211390.0001400.3271410.0001421.000143Distance to crime hotspots144 Straight distance (miles)145 Nearest14661,9021473.7841483.9671491.6091500.0311518.140152 2nd nearest15361,9021544.0151554.1991561.5951570.2991588.165159 3rd nearest16061,9021614.1401624.3321631.5921640.4451658.334166 4th nearest16761,9021684.2941694.4781701.6091710.5921728.645173 5th nearest17461,9021754.4281764.6041771.6071780.6141798.728180 Travel distance (miles)181 Nearest18261,9021834.7951845.0881851.9621860.04618710.902188 2nd nearest18961,9021905.1951915.5061921.9791930.47619411.156195 3rd nearest19661,9021975.4281985.7451992.0212000.63820111.348202 4th nearest20361,9022045.5952055.8912062.0142070.75520811.374209 5th nearest21061,9022115.8022126.1142132.0112140.82521511.396216 Travel time (minutes)217 Nearest21861,90221912.33522012.8642213.9922220.11822327.848224 2nd nearest22561,90222613.54422714.0932283.6462293.03023027.877231 3rd nearest23261,90223314.07623414.6272353.6142363.41223728.819238 4th nearest23961,90224014.83524115.3992423.5462434.22124429.700245 5th nearest24661,90224715.48824816.0012493.5562504.55325129.980252Sociodemographic characteristics (census tract-year observations)253N254Mean255Median256Std Dev257Min258Max259Population2601,2352615,1382624,9002631,6832641,91526511,930266% Male2671,2352680.4922690.4912700.0182710.4392720.567273% Female2741,2352750.5082760.5092770.0182780.4332790.561280% White2811,2352820.7092830.7922840.1942850.1282860.922287% Asian2881,2352890.1392900.0952910.1132920.0322930.639294% Black2951,2352960.0832970.0332980.0972990.0053000.426301% Old population3021,2353030.1163040.1133050.0413060.0403070.256308The sample consists of 61,902 single-family home sales in Seattle from 2008 to 2020. Definitions of the 309variables are provided in Appendix 1 Table 10310311312===== PAGE 19 (score 8, numeric-paren 0, table-words 2, var-words 4) =====313314495315Crime Measures and Housing Prices: an Analysis Using Quantile…316The second choice seems the most frequent choice in the related literature and the 317one that we employ in our analysis.318To calculate Wy for home i’s sale (shown as a red dot • in Fig. 3), we first identify 319sales of its neighbor homes (j = 1,2,…) that occurred within 2,000 feet, within the 320past 24 months. The dark dots within the red circle are the sales of neighbor homes. 321For each neighbor home sale j, we measure the distance from home sale i and use 322the sale price yj for neighbor home j to calculate wijyj. Finally, we calculate the spa-323tial weight matrix for the sale of house i, [Wy]it, using the above equation.324Hotspot and Coldspot Characterization325We define a crime hotspot as a census tract with high cluster cores (densities) of 326crimes.39 As noted above, there is a substantial literature that documents the exist-327ence of crime hotspots, specifically in Seattle. For each year, we aggregate the 328number of reported crimes at the census tract level, and compute the Getis and Ord 329(1992) statistics, G∗330i for each tract i, defined as:331The numerator is the weighted average number of crimes reported in the neigh-332boring locations (census tracts j = 1,…,105). The denominator is the sum of all crime 333counts across all the census tracts. We categorize a census tract with a G∗334i larger than 335the mean as a “high-high cluster,” or “hotspot.” Similarly, a census tract with a G∗336i 337lower than the mean is categorized as a “low-low cluster,” or “coldspot.” For each 338year we select census tracts defined as hotspots based on G∗339i , and sort them from 340the highest to lowest within the High-High cluster. A census tract with the highest 341G∗342i statistics (with a p-value < 0.05 for significance) is defined as a hotspot, the tract 343with the highest crime density. We also define a census tract with the second, third, 344fourth and fifth highest and significant G∗345i statistics as the hotspot with the second, 346third, fourth and fifth highest crime density.347Figure 4 depicts census tracts with high crime density (red), those with low crime 348density (blue) or no significant density (gray) in 2010, 2015 and 2020. Those cat-349egorized as red and blue have G∗350i with p-value < 0.05.351Endogeneity352A major confounding issue in this analysis is the likely endogeneity of crime rates 353and housing prices, as noted earlier. This problem was initially handled in the con-354text of hedonic models by Burgess and Harmon (1991). To address this issue we 355use the Wu-Hausman test, which assumes the null hypothesis of no endogeneity.40 356It tests for the correlation between the regression residuals and a given explanatory 357G∗358i =359∑360j wijxj361∑362j xj36339 Ceccato and Wilhelmsson (2020) shows that housing prices are more sensitive to distance to crime 364hotspots than to overall crime rates.36540 Wu (1973) and Hausman (1978).366367368===== PAGE 26 (score 11, numeric-paren 5, table-words 1, var-words 4) =====369370502371 372J. G. Kallberg, Y. Shimizu 373in years 2010, 2015 and 2020. Consistent with our earlier observation, we find a 374concentration of crime around the Westlake Center area consistently throughout 3752010, 2015 and 2020. The area with crime intensity further expands in 2020.376Wu‑Hausman Test and Correlations377We identify a set of 9 possible instruments. We then regress these potentially endog-378enous variables on 911 Calls as well as other exogenous variables on the right-hand 379side of Eq. 1 and obtain the residuals (v). We include v in Eq. 1 and test if its coef-380ficient is statistically different from zero.381In Table 5 Panel A, the computed Wu-Hausman statistic has p-values < 0.05, 382meaning that we reject the Ho (the variable is exogenous), for Number of 383Crimes – Total, Property and Social; for changes: ∆ Crimes – Total and Prop-384erty. With these factors as a right-hand side variable, an OLS model is biased 385and hence an instrumental variable approach is necessary. Four variables pass 386the Wu-Hausman test (Number of Crimes – Violent, ∆ Crimes—Violent, ∆ 387Crimes—Social and Distance to Crime Hotspot) and can thus be used in an 388OLS regression.389The crime variables that did not pass the Wu-Hausman test are instrumented by 390the 911 Calls variables.46 To validate that 911 Calls is a viable instrument, we per-391form a Montiel—Pfluger test to identify a weak instrument: in both cases the vari-392ables pass the test.393Panel B reports correlations between these variables. For levels (changes), we 394compute the aggregate number of 911 calls and crimes per month for Number of 395Crimes – Total, Violent, Property and Social, and Distance to Crime Hotspot 396(Straight Miles, Travel Miles and Travel Time). The correlations between Number of 397Crimes – Total and subsets are positive, as expected. Interestingly, the correlation of 398911 Calls (both for levels and changes) is strongest for Violent (0.465), slightly less 399positive for Property (0.268) and Social (0.202).400Our 3 measures of Distance to Crime Hotspot variable are only significantly 401correlated with Violent (− 0.325). These 3 variables, however, are positively cor-402related with 911 Calls, suggesting that the farther away from a crime hotspot, 403the likelier it is that a 911 call will be made. This can be partially explained by 404the higher allocation of police resources to areas with the highest incidences of 405crime. The Center for Evidence-Based Crime Policy (CEBCP) reports that the 406effort to allocate more resources in crime hotspot areas resulted in reduced num-407bers of calls for service in the affected areas.47 Consistent with our intuition, our 4083 measures of Distance to Crime Coldspot are significantly correlated with Vio-409lent (0.276) and Social (0.519).41046 Number of Crimes and ∆ Crimes are instrumented by Number of 911 Calls and ∆ 911 Calls, respec-411tively.41247 https://havengear.com/blog/increasing-hot-spots-policing-results-in-lower-crime-rates/.413414415===== PAGE 27 (score 25, numeric-paren 0, table-words 4, var-words 11) =====416417503418Crime Measures and Housing Prices: an Analysis Using Quantile…4192SLS and Quantile Regressions420Results Without Spatial Autocorrelation Adjustments421Table 6 displays the results of running 2SLS and quantile regressions without incor-422porating adjustments for spatial autocorrelation or Distance to Crime Coldspot; 423when these influences are included we find important changes in our results. All 424estimates use month and year dummies.48 We have 61,902 observations using 911 425Calls as an instrument for Number of Crimes – Total. 426Our 2SLS estimates in model 1 are generally consistent with similar hedonic resi-427dential pricing models. We find a positive significant impact from: a larger living 428area or lot size; a younger home; more bathrooms; homes with a fireplace or view. 42948 In the existing hedonic literature, area-fixed effects are usually included to control for unobserved 430area-specific characteristics to address an omitted variable problem. Abbott and Klaiber (2011) docu-431ments that the omitted variable bias remains even with the use of area-fixed effects, both defined at 432coarse levels (cities or counties) or at much finer scales (zip areas or census tracts/blocks). Our baseline 433models (Eqs. 1 and 2) exclude area-fixed effects, and we address the omitted variable issue by including 434the spatial autocorrelation variable, Wy (Eqs. 3 and 4), to capture important neighborhood characteristics 435shared with nearby residents.436Table 4 Getis-Ord statistics and crime hot/coldspots437This table reports Getis-Ord statistics and the number of census tracts identified as crime hot/coldspots in 438each year during the 2008 – 2020 period. For each census tract i, we compute the Getis-Ord (Gi439*) statistic 440as the weighted-average number of crimes reported in its neighboring census tract j (j = 1, 2, …), divided 441by the sum of all crime counts across all the census tracts. A census tract j is defined as census tract i’s 442neighbor if it shares a border (including a vertex) with tract i based on the queen contiguity. A census 443tract with Gi444* greater (smaller) than the average Gi445* across all the census tracts in a given year is defined 446as a hot (cold) spot447Year448# of crimes per 449tract450Getis-ord statistics 451(G*)452Hotspot tracts453Coldspot tracts Non-hot/cold 454spot tracts455Mean456Median457Mean458Median459N460(%)461N462(%)463N464(%)46520084664604673374680.775%4690.621%470134719.63%47294736.67%47411347583.70%47620094774684783504790.777%4800.621%481134829.63%483124848.89%48511048681.48%48720104884524893394900.776%4910.640%4921449310.37%494124958.89%49610949780.74%49820114994225003165010.776%5020.652%503115048.15%505105067.41%50711450884.44%50920125104215113185120.778%5130.633%514125158.89%516125178.89%51811151982.22%52020135214565223325230.774%5240.653%525135269.63%527125288.89%52911053081.48%53120145325005333655340.778%5350.639%536125378.89%538115398.15%54011254182.96%54220155434875443495450.777%5460.640%547135489.63%549115508.15%55111155282.22%55320165545055553795560.777%5570.639%558115598.15%56075615.19%56211756386.67%56420175655225663635670.778%5680.629%569125708.89%571105727.41%57311357483.70%57520185765375773525780.782%5790.622%5801458110.37%582105837.41%58411158582.22%58620195875005883565890.783%5900.626%5911459210.37%59395946.67%59511259682.96%59720205985305994326000.771%6010.653%6021560311.11%60486055.93%60611260782.96%608609610===== PAGE 28 (score 20, numeric-paren 11, table-words 2, var-words 3) =====611612504613 614J. G. Kallberg, Y. Shimizu 615Table 5 Endogeneity, instrumental variables and correlations616Panel A: Tests for endogeneity and weak instrument617Wu-Hausman F-Stats (p-value)618(H0: Variable is exogenous)619Need an IV?620Number of 621crimes62216.496***623(< 0.001)624Yes625Number of 626crimes—627property62822.232***629(< 0.001)630Yes631Number of 632crimes—633violent6340.008635(0.928)636No637Number of 638crimes—639social64029.296***641(< 0.001)642Yes643Δ crimes6445.627**645(0.018)646Yes647Δ crimes—648property6494.320**650(0.038)651Yes652Δ crimes—653violent6540.358655(0.550)656No657Δ crimes—658social6590.889660(0.342)661No662Distance 663to crime 664hotspot6650.725666(0.346)667No668*, **, and *** indicate significance at the 10, 5 and 1 percent levels, respectively669Montiel—Pflu-670eger effective 671F-Stats672(H0: Variable 673is a weak 674instrument)675Critical 676value677(tau = 5%)678Weak IV?679Number of 911 calls6803,374.99668137.418682No683Δ 911 calls684374.02568537.418686No687688689===== PAGE 29 (score 76, numeric-paren 56, table-words 2, var-words 3) =====690691505692Crime Measures and Housing Prices: an Analysis Using Quantile…693Table 5 (continued)694Panel B: Correlations695(1)696(2)697(3)698(4)699(5)700(6)701(7)702(8)703(9)704(10)705(11)706(12)707(13)708(14)709(15)710(16)711(1) Number of 911 Calls7121.000713(2) Number of Crimes—714Total7150.344*7161.000717(3) Number of Crimes—718Violent7190.465*7200.497*7211.000722(4) Number of Crimes—723Property7240.268*7250.964*7260.301*7271.000728(5) Number of Crimes—729Social7300.202*7310.288*7320.478*7330.0507341.000735(6) Δ 911 Calls7360.245*7370.183*7380.1587390.1597400.0887411.000742(7) Δ Crimes—Total7430.1417440.595*7450.1397460.601*7470.1537480.481*7491.000750(8) Δ Crimes—Violent7510.1547520.218*7530.319*7540.1667550.1287560.542*7570.448* 1.000758(9) Δ Crimes—Property7590.1307600.606*7610.1207620.622*7630.1207640.441*7650.991* 0.368*7661.000767(10) Δ Crimes—Social7680.1177690.1357700.0467710.0867720.286*7730.362*7740.533* 0.292*7750.446* 1.000776(11) Distance to Crime 777Hotspot (Straight Miles)7780.410*779 − 0.067780 − 0.325*7810.012782 − 0.186*7830.0277840.0547850.0077860.0497870.0747881.000789(12) Distance to Crime 790Hotspot (Travel Miles)7910.423*792 − 0.010793 − 0.252*7940.072795 − 0.234*7960.0137970.0617980.0047990.0578000.0838010.966*8021.000803(13) Distance to Crime 804Hotspot (Travel Time)8050.428*806 − 0.024807 − 0.317*8080.092809 − 0.363*8100.0338110.0548120.0038130.0538140.0548150.902*8160.948*8171.000818(14) Distance to Crime 819Coldspot (Straight 820Miles)8210.0548220.0058230.276* − 0.1378240.519*8250.021826 − 0.0018270.049828 − 0.0098290.029 − 0.108830 − 0.152831 − 0.255*8321.000833(15) Distance to Crime 834Coldspot (Travel Miles)8350.0988360.0198370.348* − 0.1418380.564*8390.0228400.0008410.054842 − 0.0088430.023 − 0.144844 − 0.191*845 − 0.291*8460.957* 1.000847(16) Distance to Crime 848Coldspot (Travel Time)8490.231*8500.1228510.480* − 0.0518520.580*8530.0198540.0078550.0598560.0018570.018 − 0.193*858 − 0.217*859 − 0.311*8600.900* 0.951* 1.000861Panel 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 – 862Pflueger 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 863instrumental variables and the crime measure variables864* shows significance at the 0.05 level865866867===== PAGE 30 (score 238, numeric-paren 195, table-words 4, var-words 15) =====868869506870 871J. G. Kallberg, Y. Shimizu 872Table 6 2SLS and quantile regressions – results without spatial autocorrelation adjustments873This table reports the results of 2SLS regression (model 1, as estimated in Eq. 1) and quantile regressions 874(models 2 – 6, as estimated in Eq. 2) for 5, 25, 50, 75 and 95 percent quantiles, respectively. Definitions of 875the variables are provided in Appendix 1 Table 10. Number of crimes is instrumented by Number of 911 876calls. Standard errors are reported in parentheses: for the 2SLS regression (model 1), standard errors are 877clustered by year and month. For quantile regressions (models 2 – 6), standard errors are obtained through 878bootstrap replications. *, ** and *** denote significance at the 10, 5 and 1 percent levels, respectively879Using Number of 911 Calls as an IV for Number of Crimes880Dependent variable:881(1)882(2)883(3)884(4)885(5)886(6)887Log (Sale Price)8882SLS889Q: 0.05890Q: 0.25891Q: 0.50892Q: 0.75893Q: 0.95894Living area8950.320***8960.286***8970.307***8980.321***8990.341***9000.348***901(0.003)902(0.006)903(0.003)904(0.003)905(0.004)906(0.005)907Lot area9080.021***9090.008***9100.013***9110.020***9120.026***9130.038***914(0.001)915(0.002)916(0.001)917(0.001)918(0.001)919(0.002)920Bedrooms921 − 0.019***922 − 0.009**923 − 0.012***924 − 0.012***925 − 0.014***926 − 0.018***927(0.002)928(0.003)929(0.002)930(0.001)931(0.001)932(0.003)933Bathrooms (Full)9340.003935 − 0.010**9360.0029370.005***9380.004**9390.010***940(0.002)941(0.004)942(0.003)943(0.002)944(0.002)945(0.003)946Age947 − 0.000***948 − 0.001***949 − 0.000***950 − 0.000***951 − 0.000***952 − 0.000953(0.000)954(0.000)955(0.000)956(0.000)957(0.000)958(0.000)959One story960 − 0.093***961 − 0.062***962 − 0.076***963 − 0.081***964 − 0.087***965 − 0.099***966(0.007)967(0.005)968(0.002)969(0.003)970(0.002)971(0.003)972Renovated973 − 0.046***974 − 0.275***975 − 0.107***976 − 0.0069770.025***9780.047***979(0.011)980(0.017)981(0.012)982(0.006)983(0.005)984(0.011)985Fireplace9860.040***9870.062***9880.041***9890.029***9900.020***9910.014***992(0.005)993(0.008)994(0.002)995(0.001)996(0.002)997(0.003)998Lake/Mountain view9990.105***10000.065***10010.069***10020.081***10030.104***10040.176***1005(0.006)1006(0.007)1007(0.004)1008(0.004)1009(0.006)1010(0.008)1011Population10120.012**1013 − 0.027***10140.00010150.014***10160.022***10170.027***1018(0.005)1019(0.008)1020(0.006)1021(0.004)1022(0.004)1023(0.007)1024% Male population10250.726***10261.706***10271.043***10280.762***10290.521***10301.083***1031(0.153)1032(0.206)1033(0.100)1034(0.093)1035(0.101)1036(0.237)1037% White population10380.793***10391.093***10400.792***10410.780***10420.773***10430.749***1044(0.034)1045(0.163)1046(0.022)1047(0.024)1048(0.019)1049(0.026)1050% Asian population10510.04310520.08010530.01110540.088**10550.160***10560.230***1057(0.032)1058(0.149)1059(0.023)1060(0.035)1061(0.022)1062(0.053)1063% Old population10640.283***10650.427***10660.124***10670.080***10680.078**10690.372***1070(0.042)1071(0.107)1072(0.040)1073(0.029)1074(0.031)1075(0.057)1076Number of crimes (IV)1077 − 0.828**1078 − 0.1171079 − 0.948***1080 − 0.798***1081 − 0.757***1082 − 1.104***1083(0.347)1084(0.790)1085(0.170)1086(0.176)1087(0.136)1088(0.183)1089Distance to crime hotspot1090 − 0.072***1091 − 0.072***1092 − 0.068***1093 − 0.070***1094 − 0.073***1095 − 0.084***1096(0.003)1097(0.002)1098(0.001)1099(0.001)1100(0.001)1101(0.001)1102Constant110312.295***110411.659***110512.199***110612.274***110712.385***110812.216***1109(0.078)1110(0.150)1111(0.055)1112(0.033)1113(0.047)1114(0.088)1115Year dummies1116Yes1117Yes1118Yes1119Yes1120Yes1121Yes1122Month dummies1123Yes1124Yes1125Yes1126Yes1127Yes1128Yes1129N113061,902113161,902113261,902113361,902113461,902113561,902113611371138===== PAGE 31 (score 25, numeric-paren 0, table-words 6, var-words 13) =====113911405071141Crime Measures and Housing Prices: an Analysis Using Quantile…1142Additional bedrooms and houses with a single floor exhibit a negative relation, since 1143living area is already factored into the model. In terms of demographics, the 2SLS 1144estimates are significantly positive for % White, % Male and % Old Population.1145The quantile regressions in models 2 to 6 allow us to examine how these 2SLS 1146results vary across the price quantiles: 0.05 (low end), 0.25, 0.50 (median), 0.75 1147and 0.95 (high end). The most evident trends are increasing significance (from low 1148to high prices) of: lot area and view and the decreasing significance of fireplace. 1149The sign on % Asian Population is positive in the higher quantiles.When we include 1150zip code fixed effects and re-run these regressions (results are reported in Appen-1151dix 2 Table 11), the sign on % Asian Population becomes significant and negative in 1152all quantiles.49 The coefficients on other demographic variables generally retain the 1153same sign with zip code fixed effects.1154Our focus however is on the impact of crime and we see a significant negative 1155relation between 911 Calls and housing prices in each of the quantile regressions, 1156except for the lowest priced houses, where there is no significant relation. This fact 1157is consistent with the notion that increasing crime rates (as instrumented by 911 1158Calls) depress housing prices. We find a significant negative relation between the 1159distance to a crime hotspot and housing prices. The latter, perhaps surprising result, 1160is discussed further below.1161We re-estimate this regression using different crime variables.50 When we substi-1162tute Number of Crimes—Violent for Number of 911 Calls, its coefficient is signifi-1163cantly negative across all the quantiles. The coefficients on the demographic vari-1164ables generally retain the same signs and approximately the same magnitudes as in 1165Table 6.1166We next use ∆ Violent Crime as a crime rate proxy and employ changes in demo-1167graphic variables.51 The coefficients on the demographic variable changes again 1168generally retain the same sign as in Table 6, except for the ∆ Old Population vari-1169able now has a significant negative sign, suggesting that an increasingly old popula-1170tion depresses prices. The coefficient on ∆ Violent Crime is significantly positive 1171in all but the median quantile. The negative estimates for coefficient on distance to 1172crime hotspot are virtually identical to those obtained previously.1173Next we employ ∆ Social Crime. The coefficients are positive and significant in 1174each quartile. In this and in the previous estimate, we find a strong negative impact 1175from increasing population. This effect is dampened when we refine our estimation 1176in the following tables. Furthermore, our discussions with Port of Seattle police have 117749 The inclusion of zip code fixed effects accounts for certain unobserved neighborhood characteris-1178tics, one of which could be gentrification. Hwang (2020) studies Seattle’s gentrification from the 1970’s 1179through 2010’s, and shows that neighborhoods with a higher Asian population experienced less gentri-1180fication, while those with a higher Black population were gentrified more frequently. She explains that 1181newly arriving immigrants in Seattle, who are predominantly Asian recently, tend to settle in areas with 1182an existing significant Asian population. This influx can deter gentrification in such areas, potentially 1183leading to lower housing prices (compared to areas which recently have been gentrified).118450 Here we only report the most relevant results for these alternative measures. The complete regressions 1185are available from the authors.118651 When we use changes in proxies for crime rates, we also use changes in demographic variables.118711881189===== PAGE 34 (score 261, numeric-paren 217, table-words 3, var-words 15) =====119011915101192 1193J. G. Kallberg, Y. Shimizu 1194Table 7 2SLS and instrumental quantile regressions – results with spatial autocorrelation adjustments1195Panel A: Using Number of 911 Calls as an IV for Number of Crimes1196Dependent variable:1197(1)1198(2)1199(3)1200(4)1201(5)1202(6)1203Log (Sale Price)12042SLS1205Q: 0.051206Q: 0.251207Q: 0.501208Q: 0.751209Q: 0.951210Log (WY)12110.394***12120.414***12130.368***12140.364***12150.364***12160.376***1217(0.013)1218(0.017)1219(0.006)1220(0.004)1221(0.008)1222(0.013)1223Living area12240.261***12250.241***12260.252***12270.267***12280.286***12290.299***1230(0.004)1231(0.006)1232(0.003)1233(0.003)1234(0.002)1235(0.005)1236Lot area12370.018***12380.008***12390.014***12400.018***12410.022***12420.034***1243(0.001)1244(0.002)1245(0.000)1246(0.001)1247(0.001)1248(0.001)1249Bedrooms1250 − 0.007***1251 − 0.0001252 − 0.002*1253 − 0.002*1254 − 0.002*1255 − 0.008**1256(0.001)1257(0.004)1258(0.001)1259(0.001)1260(0.001)1261(0.004)1262Bathrooms (Full)12630.008***1264 − 0.007***12650.006***12660.009***12670.009***12680.015***1269(0.001)1270(0.002)1271(0.002)1272(0.002)1273(0.002)1274(0.002)1275Age1276 − 0.001***1277 − 0.002***1278 − 0.001***1279 − 0.001***1280 − 0.000***1281 − 0.000***1282(0.000)1283(0.000)1284(0.000)1285(0.000)1286(0.000)1287(0.000)1288One story1289 − 0.076***1290 − 0.053***1291 − 0.064***1292 − 0.067***1293 − 0.071***1294 − 0.084***1295(0.005)1296(0.007)1297(0.002)1298(0.002)1299(0.002)1300(0.004)1301Renovated1302 − 0.028***1303 − 0.264***1304 − 0.102***13050.021***13060.048***13070.069***1308(0.011)1309(0.012)1310(0.012)1311(0.006)1312(0.002)1313(0.010)1314Fireplace13150.025***13160.044***13170.025***13180.013***13190.010***13200.007***1321(0.004)1322(0.007)1323(0.002)1324(0.002)1325(0.002)1326(0.002)1327Lake/Mountain view13280.063***13290.037***13300.036***13310.047***13320.059***13330.107***1334(0.005)1335(0.006)1336(0.003)1337(0.003)1338(0.003)1339(0.004)1340Population1341 − 0.0011342 − 0.0171343 − 0.00513440.00113450.00313460.013**1347(0.002)1348(0.013)1349(0.003)1350(0.004)1351(0.004)1352(0.005)1353% Male population1354 − 0.188*1355 − 0.24113560.07013570.05213580.00913590.377***1360(0.099)1361(0.234)1362(0.060)1363(0.079)1364(0.074)1365(0.142)1366% White population13670.605***13680.883***13690.626***13700.557***13710.520***13720.458***1373(0.029)1374(0.067)1375(0.018)1376(0.015)1377(0.018)1378(0.020)1379% Asian population13800.147***13810.234**13820.102***13830.129***13840.152***13850.151***1386(0.016)1387(0.104)1388(0.031)1389(0.027)1390(0.019)1391(0.032)1392% Old population1393 − 0.167***1394 − 0.246***1395 − 0.184***1396 − 0.206***1397 − 0.227***1398 − 0.222***1399(0.028)1400(0.053)1401(0.030)1402(0.025)1403(0.015)1404(0.032)1405Number of crimes (IV)14060.730***14071.294***14080.599***14090.373***14100.214*1411 − 0.1111412(0.148)1413(0.310)1414(0.094)1415(0.100)1416(0.118)1417(0.126)1418Distance to crime hotspot1419 − 0.038***1420 − 0.036***1421 − 0.038***1422 − 0.040***1423 − 0.042***1424 − 0.046***1425(0.002)1426(0.002)1427(0.001)1428(0.001)1429(0.001)1430(0.001)1431Constant14327.511***14336.972***14347.715***14357.855***14367.932***14377.670***1438(0.180)1439(0.221)1440(0.076)1441(0.078)1442(0.076)1443(0.181)1444Year dummies1445Yes1446Yes1447Yes1448Yes1449Yes1450Yes1451Month dummies1452Yes1453Yes1454Yes1455Yes1456Yes1457Yes1458N145961,902146061,902146161,902146261,902146361,902146461,9021465Panel B: Using Number of Crimes—Violent Only1466Dependent variable:1467(1)1468(2)1469(3)1470(4)1471(5)1472(6)1473Log (Sale Price)14742SLS1475Q: 0.051476Q: 0.251477Q: 0.501478Q: 0.751479Q: 0.951480Log (WY)14810.376***14820.371***14830.348***14840.350***14850.354***14860.372***1487(0.012)1488(0.016)1489(0.006)1490(0.003)1491(0.007)1492(0.012)149314941495===== PAGE 35 (score 290, numeric-paren 246, table-words 2, var-words 15) =====149614975111498Crime Measures and Housing Prices: an Analysis Using Quantile…1499Table 7 (continued)1500Living area15010.260***15020.237***15030.252***15040.266***15050.285***15060.299***1507(0.004)1508(0.005)1509(0.003)1510(0.002)1511(0.002)1512(0.005)1513Lot area15140.018***15150.005***15160.013***15170.017***15180.022***15190.034***1520(0.001)1521(0.002)1522(0.000)1523(0.001)1524(0.001)1525(0.001)1526Bedrooms1527 − 0.008***1528 − 0.0031529 − 0.003*1530 − 0.003**1531 − 0.003**1532 − 0.008**1533(0.001)1534(0.004)1535(0.001)1536(0.001)1537(0.001)1538(0.004)1539Bathrooms (Full)15400.008***1541 − 0.006***15420.005*15430.009***15440.009***15450.014***1546(0.001)1547(0.002)1548(0.003)1549(0.001)1550(0.002)1551(0.002)1552Age1553 − 0.001***1554 − 0.002***1555 − 0.001***1556 − 0.001***1557 − 0.000***1558 − 0.000***1559(0.000)1560(0.000)1561(0.000)1562(0.000)1563(0.000)1564(0.000)1565One story1566 − 0.080***1567 − 0.061***1568 − 0.068***1569 − 0.070***1570 − 0.072***1571 − 0.085***1572(0.005)1573(0.005)1574(0.002)1575(0.003)1576(0.003)1577(0.003)1578Renovated1579 − 0.028***1580 − 0.262***1581 − 0.104***15820.021***15830.048***15840.070***1585(0.011)1586(0.012)1587(0.017)1588(0.006)1589(0.003)1590(0.009)1591Fireplace15920.025***15930.047***15940.026***15950.014***15960.010***15970.006**1598(0.004)1599(0.005)1600(0.002)1601(0.002)1602(0.002)1603(0.003)1604Lake/Mountain view16050.063***16060.036***16070.037***16080.046***16090.060***16100.107***1611(0.005)1612(0.005)1613(0.002)1614(0.004)1615(0.003)1616(0.004)1617Population16180.012***16190.01116200.012***16210.015***16220.009***16230.015***1624(0.002)1625(0.009)1626(0.002)1627(0.003)1628(0.003)1629(0.005)1630% Male population16310.307***16320.787***16330.525***16340.440***16350.274***16360.465***1637(0.070)1638(0.128)1639(0.041)1640(0.047)1641(0.062)1642(0.106)1643% White population16440.455***16450.602***16460.476***16470.454***16480.449***16490.410***1650(0.023)1651(0.037)1652(0.017)1653(0.012)1654(0.014)1655(0.021)1656% Asian population16570.00616580.0211659 − 0.02616600.041*16610.078***16620.097**1663(0.028)1664(0.053)1665(0.029)1666(0.022)1667(0.018)1668(0.039)1669% Old population1670 − 0.183***1671 − 0.172***1672 − 0.212***1673 − 0.223***1674 − 0.222***1675 − 0.214***1676(0.022)1677(0.044)1678(0.027)1679(0.022)1680(0.015)1681(0.026)1682Number of crimes—violent1683 − 2.139***1684 − 3.152***1685 − 2.447***1686 − 2.130***1687 − 1.472***1688 − 1.577***1689(0.455)1690(0.578)1691(0.322)1692(0.211)1693(0.151)1694(0.331)1695Distance to crime hotspot1696 − 0.040***1697 − 0.040***1698 − 0.039***1699 − 0.041***1700 − 0.043***1701 − 0.046***1702(0.002)1703(0.002)1704(0.001)1705(0.000)1706(0.001)1707(0.001)1708Constant17097.622***17107.208***17117.827***17127.884***17137.981***17147.709***1715(0.177)1716(0.196)1717(0.076)1718(0.069)1719(0.065)1720(0.168)1721Year dummies1722Yes1723Yes1724Yes1725Yes1726Yes1727Yes1728Month dummies1729Yes1730Yes1731Yes1732Yes1733Yes1734Yes1735N173661,902173761,902173861,902173961,902174061,902174161,9021742Panel C: Using Δ Crimes—Violent1743Dependent variable1744(1)1745(2)1746(3)1747(4)1748(5)1749(6)1750Log (Sale Price)17512SLS1752Q: 0.051753Q: 0.251754Q: 0.501755Q: 0.751756Q: 0.951757Log (WY)17580.425***17590.431***17600.395***17610.388***17620.398***17630.413***1764(0.018)1765(0.017)1766(0.005)1767(0.006)1768(0.011)1769(0.014)1770Living area17710.268***17720.252***17730.259***17740.276***17750.288***17760.302***1777(0.003)1778(0.006)1779(0.002)1780(0.003)1781(0.002)1782(0.005)1783Lot area17840.013***1785 − 0.00017860.009***17870.012***17880.018***17890.030***1790(0.001)1791(0.001)1792(0.000)1793(0.001)1794(0.001)1795(0.001)1796Bedrooms1797 − 0.011***1798 − 0.009**1799 − 0.007***1800 − 0.006***1801 − 0.004***1802 − 0.008**1803(0.002)1804(0.004)1805(0.001)1806(0.001)1807(0.001)1808(0.004)180918101811===== PAGE 36 (score 281, numeric-paren 238, table-words 2, var-words 14) =====181218135121814 1815J. G. Kallberg, Y. Shimizu 1816Table 7 (continued)1817Bathrooms (Full)18180.005***1819 − 0.012**18200.00018210.005**18220.008***18230.016***1824(0.001)1825(0.005)1826(0.001)1827(0.002)1828(0.001)1829(0.003)1830Age1831 − 0.001***1832 − 0.001***1833 − 0.001***1834 − 0.000***1835 − 0.000***1836 − 0.0001837(0.000)1838(0.000)1839(0.000)1840(0.000)1841(0.000)1842(0.000)1843One story1844 − 0.080***1845 − 0.066***1846 − 0.068***1847 − 0.067***1848 − 0.071***1849 − 0.085***1850(0.005)1851(0.005)1852(0.002)1853(0.002)1854(0.003)1855(0.004)1856Renovated1857 − 0.031***1858 − 0.259***1859 − 0.103***18600.015*18610.049***18620.065***1863(0.011)1864(0.010)1865(0.012)1866(0.008)1867(0.003)1868(0.011)1869Fireplace18700.031***18710.061***18720.031***18730.016***18740.014***18750.010***1876(0.005)1877(0.006)1878(0.003)1879(0.002)1880(0.002)1881(0.003)1882Lake/Mountain view18830.060***18840.039***18850.040***18860.045***18870.053***18880.093***1889(0.005)1890(0.004)1891(0.003)1892(0.003)1893(0.003)1894(0.005)1895Δ Population1896 − 6.758***1897 − 7.272***1898 − 8.306***1899 − 6.953***1900 − 6.174***1901 − 7.300***1902(1.261)1903(1.100)1904(0.488)1905(0.637)1906(0.737)1907(0.742)1908Δ Male population19090.5411910 − 0.05719111.574**19120.60719130.40319142.223***1915(1.205)1916(1.394)1917(0.632)1918(0.503)1919(0.497)1920(0.691)1921Δ White population19223.432***19233.445***19243.560***19253.744***19263.715***19273.541***1928(0.952)1929(0.238)1930(0.218)1931(0.307)1932(0.263)1933(0.282)1934Δ Asian population19353.347***19364.176***19373.646***19383.242***19392.895***19402.525***1941(0.259)1942(0.234)1943(0.085)1944(0.062)1945(0.076)1946(0.141)1947Δ Old population1948 − 0.652**1949 − 0.968***1950 − 0.867***1951 − 0.741***1952 − 0.581***1953 − 0.404***1954(0.301)1955(0.251)1956(0.095)1957(0.091)1958(0.058)1959(0.098)1960Δ Crimes—violent1961 − 0.00219620.00219630.00219640.0001965 − 0.00019660.005*1967(0.009)1968(0.011)1969(0.002)1970(0.004)1971(0.002)1972(0.003)1973Distance to crime hotspot1974 − 0.031***1975 − 0.028***1976 − 0.032***1977 − 0.032***1978 − 0.034***1979 − 0.037***1980(0.004)1981(0.002)1982(0.001)1983(0.001)1984(0.001)1985(0.001)1986Constant19877.352***19887.077***19897.708***19907.857***19917.783***19927.663***1993(0.246)1994(0.204)1995(0.079)1996(0.082)1997(0.146)1998(0.180)1999Year dummies2000Yes2001Yes2002Yes2003Yes2004Yes2005Yes2006Month dummies2007Yes2008Yes2009Yes2010Yes2011Yes2012Yes2013N201461,902201561,902201661,902201761,902201861,902201961,9022020Panel D: Using Δ Crimes—Social2021Dependent variable:2022(1)2023(2)2024(3)2025(4)2026(5)2027(6)2028Log (Sale Price)20292SLS2030Q: 0.052031Q: 0.252032Q: 0.502033Q: 0.752034Q: 0.952035Log (WY)20360.424***20370.429***20380.395***20390.388***20400.397***20410.413***2042(0.017)2043(0.016)2044(0.005)2045(0.006)2046(0.011)2047(0.014)2048Living area20490.268***20500.252***20510.259***20520.276***20530.288***20540.302***2055(0.003)2056(0.005)2057(0.002)2058(0.002)2059(0.002)2060(0.005)2061Lot area20620.013***2063 − 0.00120640.009***20650.012***20660.018***20670.030***2068(0.001)2069(0.001)2070(0.000)2071(0.001)2072(0.001)2073(0.001)2074Bedrooms2075 − 0.011***2076 − 0.009**2077 − 0.007***2078 − 0.006***2079 − 0.004***2080 − 0.008**2081(0.002)2082(0.004)2083(0.001)2084(0.001)2085(0.001)2086(0.004)2087Bathrooms (Full)20880.005***2089 − 0.012**20900.00020910.005**20920.008***20930.016***2094(0.001)2095(0.005)2096(0.001)2097(0.002)2098(0.001)2099(0.003)2100Age2101 − 0.001***2102 − 0.001***2103 − 0.001***2104 − 0.000***2105 − 0.000***2106 − 0.0002107(0.000)2108(0.000)2109(0.000)2110(0.000)2111(0.000)2112(0.000)2113One story2114 − 0.080***2115 − 0.066***2116 − 0.068***2117 − 0.067***2118 − 0.071***2119 − 0.085***2120(0.005)2121(0.005)2122(0.002)2123(0.003)2124(0.003)2125(0.004)212621272128===== PAGE 37 (score 283, numeric-paren 240, table-words 2, var-words 14) =====212921305132131Crime Measures and Housing Prices: an Analysis Using Quantile…2132Table 7 (continued)2133Renovated2134 − 0.031***2135 − 0.260***2136 − 0.103***21370.014*21380.049***21390.066***2140(0.011)2141(0.011)2142(0.012)2143(0.007)2144(0.003)2145(0.012)2146Fireplace21470.031***21480.061***21490.031***21500.016***21510.014***21520.010***2153(0.005)2154(0.006)2155(0.003)2156(0.002)2157(0.002)2158(0.003)2159Lake/Mountain view21600.060***21610.040***21620.040***21630.045***21640.054***21650.093***2166(0.005)2167(0.005)2168(0.003)2169(0.003)2170(0.003)2171(0.005)2172Δ Population2173 − 6.743***2174 − 7.292***2175 − 8.304***2176 − 6.972***2177 − 6.166***2178 − 7.275***2179(1.257)2180(1.092)2181(0.527)2182(0.616)2183(0.733)2184(0.749)2185Δ Male population21860.5132187 − 0.07921881.570**21890.61821900.37021912.208***2192(1.207)2193(1.382)2194(0.626)2195(0.532)2196(0.539)2197(0.674)2198Δ White population21993.440***22003.465***22013.560***22023.745***22033.724***22043.545***2205(0.950)2206(0.215)2207(0.215)2208(0.310)2209(0.259)2210(0.290)2211Δ Asian population22123.347***22134.180***22143.645***22153.239***22162.901***22172.519***2218(0.260)2219(0.219)2220(0.093)2221(0.057)2222(0.075)2223(0.139)2224Δ Old population2225 − 0.652**2226 − 0.982***2227 − 0.866***2228 − 0.732***2229 − 0.574***2230 − 0.406***2231(0.299)2232(0.243)2233(0.088)2234(0.092)2235(0.062)2236(0.096)2237Δ Crimes—social22380.00522390.007**22400.00222410.006***22420.006***22430.0042244(0.006)2245(0.003)2246(0.002)2247(0.002)2248(0.002)2249(0.004)2250Distance to crime hotspot2251 − 0.031***2252 − 0.029***2253 − 0.032***2254 − 0.032***2255 − 0.034***2256 − 0.037***2257(0.004)2258(0.002)2259(0.001)2260(0.001)2261(0.001)2262(0.001)2263Constant22647.361***22657.107***22667.710***22677.863***22687.788***22697.665***2270(0.243)2271(0.201)2272(0.079)2273(0.085)2274(0.145)2275(0.175)2276Year dummies2277Yes2278Yes2279Yes2280Yes2281Yes2282Yes2283Month dummies2284Yes2285Yes2286Yes2287Yes2288Yes2289Yes2290N229161,902229261,902229361,902229461,902229561,902229661,9022297Panel E: Using 911 Calls (Lag 1) as an IV for Number of Crimes (Lag 1)2298Dependent variable:2299(1)2300(2)2301(3)2302(4)2303(5)2304(6)2305Log (Sale Price)23062SLS2307Q: 0.052308Q: 0.252309Q: 0.502310Q: 0.752311Q: 0.952312Log (WY)23130.398***23140.428***23150.370***23160.365***23170.363***23180.377***2319(0.015)2320(0.010)2321(0.008)2322(0.005)2323(0.004)2324(0.007)2325Living area23260.262***23270.242***23280.253***23290.267***23300.287***23310.300***2332(0.004)2333(0.006)2334(0.002)2335(0.003)2336(0.004)2337(0.003)2338Lot area23390.018***23400.009***23410.014***23420.018***23430.022***23440.034***2345(0.001)2346(0.001)2347(0.000)2348(0.001)2349(0.001)2350(0.001)2351Bedrooms2352 − 0.007***23530.0012354 − 0.0012355 − 0.0012356 − 0.0022357 − 0.008**2358(0.001)2359(0.003)2360(0.002)2361(0.002)2362(0.003)2363(0.004)2364Bathrooms (Full)23650.009***2366 − 0.00523670.007***23680.009***23690.009***23700.015***2371(0.001)2372(0.006)2373(0.001)2374(0.002)2375(0.001)2376(0.001)2377Age2378 − 0.001***2379 − 0.002***2380 − 0.001***2381 − 0.001***2382 − 0.000***2383 − 0.000***2384(0.000)2385(0.000)2386(0.000)2387(0.000)2388(0.000)2389(0.000)2390One story2391 − 0.075***2392 − 0.050***2393 − 0.064***2394 − 0.066***2395 − 0.071***2396 − 0.083***2397(0.005)2398(0.005)2399(0.004)2400(0.003)2401(0.003)2402(0.004)2403Renovated2404 − 0.027**2405 − 0.265***2406 − 0.102***24070.022***24080.049***24090.072***2410(0.011)2411(0.015)2412(0.015)2413(0.006)2414(0.004)2415(0.014)2416Fireplace24170.025***24180.043***24190.024***24200.014***24210.011***24220.006*2423(0.004)2424(0.006)2425(0.002)2426(0.003)2427(0.002)2428(0.003)2429Lake/Mountain view24300.062***24310.038***24320.035***24330.047***24340.059***24350.103***2436(0.005)2437(0.008)2438(0.005)2439(0.004)2440(0.005)2441(0.012)244224432444===== PAGE 38 (score 112, numeric-paren 90, table-words 4, var-words 5) =====244524465142447 2448J. G. Kallberg, Y. Shimizu 2449When we focus on 911 calls the situation is less uniform. In Table 6 the coef-2450ficient is negative and significant, suggesting that a higher frequency of 911 calls 2451depresses prices. However when we incorporate spatial correlation the sign becomes 2452positive and significant. Our estimates are similar when we use lagged 911 calls by 2453one year. These results suggest that more 911 calls are made in an area with higher 2454house prices, once we incorporate spatial autocorrelation. One potential explana-2455tion is that Wy captures characteristics specific to certain neighborhoods that are 2456not observable to econometricians, such as “collective efficacy.58” Once such char-2457acteristics are controlled for, an increase in the number of 911 calls can indicate 2458the residents’ greater efforts to monitor the safety of their residence, to report any 2459suspicious persons/activities in the area, etc. A study by Vera Institute of Justice59 246058 Collective efficacy is defined in Sampson et al. (1997) as (from the abstract): social cohesion among 2461neighbors combined with their willingness to intervene on behalf of the common good. They show that 2462neighborhoods with high collective efficacy tend to experience lower rates of violent crime.246359 https://www.vera.org/publications/making-911-accessible2464Table 7 (continued)2465Population2466 − 0.006**2467 − 0.026**2468 − 0.009**2469 − 0.00224700.00024710.012***2472(0.003)2473(0.011)2474(0.004)2475(0.003)2476(0.003)2477(0.004)2478% Male population2479 − 0.339**2480 − 0.674**2481 − 0.0182482 − 0.0572483 − 0.07024840.316**2485(0.142)2486(0.316)2487(0.131)2488(0.114)2489(0.126)2490(0.148)2491% White population24920.636***24930.922***24940.639***24950.578***24960.540***24970.467***2498(0.038)2499(0.063)2500(0.016)2501(0.016)2502(0.009)2503(0.016)2504% Asian population25050.174***25060.220**25070.105***25080.147***25090.169***25100.167***2511(0.023)2512(0.104)2513(0.026)2514(0.026)2515(0.018)2516(0.028)2517% Old population2518 − 0.155***2519 − 0.338***2520 − 0.176***2521 − 0.208***2522 − 0.223***2523 − 0.236***2524(0.031)2525(0.091)2526(0.032)2527(0.030)2528(0.031)2529(0.051)2530Number of crimes (IV—2531Lag1)25321.106***25331.773***25340.784***25350.575***25360.429***25370.0192538(0.223)2539(0.403)2540(0.130)2541(0.095)2542(0.139)2543(0.197)2544Distance to crime hotspot2545 − 0.038***2546 − 0.035***2547 − 0.038***2548 − 0.040***2549 − 0.042***2550 − 0.046***2551(0.002)2552(0.002)2553(0.000)2554(0.001)2555(0.001)2556(0.002)2557Constant25587.522***25597.001***25607.739***25617.880***25627.973***25637.670***2564(0.189)2565(0.162)2566(0.051)2567(0.050)2568(0.062)2569(0.105)2570Year dummies2571Yes2572Yes2573Yes2574Yes2575Yes2576Yes2577Month dummies2578Yes2579Yes2580Yes2581Yes2582Yes2583Yes2584N258558,301258658,301258758,301258858,301258958,301259058,3012591This table reports the results of 2SLS regression (model 1, as estimated in Eq. 3) and instrumental quantile 2592regressions (models 2 – 6, as estimated in Eq. 4) for 5, 25, 50, 75 and 95 percent quantiles, respectively. Defini-2593tions of the variables are provided in Appendix 1 Table 10. Panels A and B use Number of crimes (instrumented 2594by Number of 911 calls) and Number of crimes-violent), respectively, as well as levels of the demographic vari-2595ables (Population, % Male, % White, % Asian and % Old population) as control variables. Panels C and D 2596use annual changes for Number of crimes- violent and Number of crimes-social, respectively, as well as for the 2597demographic variables. Panel E uses lagged values of Number of crimes (instrumented by lagged Number of 2598911 calls). Standard errors are reported in parentheses: for 2SLS regressions (model 1), standard errors are clus-2599tered by year and month. For quantile regressions (models 2 – 6), standard errors are obtained through bootstrap 2600replications. *, ** and *** denote significance at the 10, 5 and 1 percent levels, respectively260126022603===== PAGE 39 (score 25, numeric-paren 0, table-words 8, var-words 9) =====260426055152606Crime Measures and Housing Prices: an Analysis Using Quantile…2607corroborates this perspective: among different types of 911 calls made in New 2608Orleans, they report that 911 calls made for property crimes and those for “suspi-2609cious behaviors” are the two predominant call types, followed by domestic violence, 2610violent crime, drugs and sex offenses. This trend is observed in both census tracts 2611with the lowest and highest 10 percent median income. Thus, the positive relation 2612between the frequency of 911 calls and housing prices (once we incorporate spa-2613tial autocorrelation) can be indicative of residents’ collective efficacy or criminals’ 2614attempts to target higher priced properties.2615Robustness2616Our first robustness check restricts our 911 Calls to only those with high priority. 2617Each 911 call is assigned a priority level, ranging from 1 (highest) to 9 (lowest), 2618based on the severity of an event. 911 Calls (High Priority) are defined as those with 2619the two highest priority levels. In our data, this subsample represents 68.1% of the 26201,528,303 calls in our total sample. By excluding calls of less severity, this measure 2621of crime levels can hopefully become more accurate.2622Table 8 Panel A presents our basic regression without adjustment for spatial auto-2623correlation. Here the coefficient on 911 Calls (High Priority) is negative and signifi-2624cant in each quantile (except for the lowest quantile), with higher values than those 2625reported in the previous tables, suggesting that this measure is a more efficient proxy 2626for crime rates than the total of 911 Calls. Panel B now include the Wy variable. As in 2627Tables 6 and 7, we now observe positive and significant estimates for 911 Calls (High 2628Priority) in the lowest 3 price quantiles. This flip from negative to positive signs on 2629the crime proxies when we include spatial autocorrelation parallels our earlier results. 2630The positive sign is more consistent with our estimates involving distance to crime 2631hotspots, which could suggest that ignoring spatial autocorrelation can lead to spuri-2632ous estimates. Very similar results obtain when we substitute lagged values of 911 2633Calls (High Priority) for results with and without spatial autocorrelation adjustments.2634Our next robustness check is presented in Table 9, which adds distance to crime 2635coldspot (Distance to Crime Coldspot) to the estimation. The rationale is that con-2636sumers should be willing to pay more for residences near low crime areas. Table 9 2637Panel A presents the results using 911 calls as an instrument without correcting for 2638spatial correlation. The estimates are very similar to those reported in Table 6 with 2639the additional observation that Distance to Crime Coldspot has a weakly negative 2640impact; its coefficient is small but significant. When we repeat this analysis using 2641Number of Crimes – Violent, ∆ Violent Crime, ∆ Social Crime and 911 Calls (Lag 26421), we generate very similar estimates for the Distance to Crime Coldspot variable.2643Table 9 Panel B now incorporates the Wy variable and uses 911 Calls as the 2644proxy for crime rates. We observe the same pattern as in Panel A: the estimates on 2645Distance to Crime Coldspot are negative and significant in each quantile, however 2646the average coefficient is only 45% of the value obtained in Panel A. Overall these 2647estimates provide consistent evidence of a small positive pricing impact of being 2648closer to a crime coldspot. When we re-estimate this regression using the other 4 2649crime rate proxies the results are qualitatively similar.265026512652===== PAGE 41 (score 262, numeric-paren 219, table-words 3, var-words 14) =====265326545172655Crime Measures and Housing Prices: an Analysis Using Quantile…2656Table 8 Robustness check – quantile regression results using 911 calls with high priority2657Panel A: Using 911 Calls (High Priority), without spatial autocorrelation adjustments2658Dependent variable:2659(1)2660(2)2661(3)2662(4)2663(5)2664(6)2665Log (Sale Price)26662SLS2667Q: 0.052668Q: 0.252669Q: 0.502670Q: 0.752671Q: 0.952672Living area26730.318***26740.284***26750.306***26760.320***26770.340***26780.347***2679(0.004)2680(0.006)2681(0.003)2682(0.003)2683(0.004)2684(0.006)2685Lot area26860.021***26870.007***26880.013***26890.019***26900.026***26910.038***2692(0.001)2693(0.002)2694(0.001)2695(0.001)2696(0.001)2697(0.002)2698Bedrooms2699 − 0.019***2700 − 0.008**2701 − 0.012***2702 − 0.012***2703 − 0.014***2704 − 0.019***2705(0.002)2706(0.004)2707(0.002)2708(0.001)2709(0.001)2710(0.003)2711Bathrooms (Full)27120.0032713 − 0.010***27140.00227150.006***27160.004*27170.009***2718(0.002)2719(0.004)2720(0.003)2721(0.002)2722(0.002)2723(0.003)2724Age2725 − 0.000***2726 − 0.001***2727 − 0.000***2728 − 0.000***2729 − 0.000**2730 − 0.0002731(0.000)2732(0.000)2733(0.000)2734(0.000)2735(0.000)2736(0.000)2737One story2738 − 0.095***2739 − 0.064***2740 − 0.077***2741 − 0.082***2742 − 0.088***2743 − 0.100***2744(0.008)2745(0.005)2746(0.002)2747(0.003)2748(0.003)2749(0.003)2750Renovated2751 − 0.046***2752 − 0.269***2753 − 0.107***2754 − 0.00727550.026***27560.049***2757(0.011)2758(0.017)2759(0.012)2760(0.006)2761(0.004)2762(0.010)2763Fireplace27640.040***27650.059***27660.042***27670.029***27680.021***27690.013***2770(0.005)2771(0.008)2772(0.002)2773(0.001)2774(0.001)2775(0.003)2776Lake/Mountain view27770.104***27780.066***27790.069***27800.080***27810.103***27820.173***2783(0.005)2784(0.007)2785(0.004)2786(0.003)2787(0.006)2788(0.007)2789Population27900.017**2791 − 0.019**27920.00427930.017***27940.025***27950.030***2796(0.007)2797(0.009)2798(0.005)2799(0.004)2800(0.004)2801(0.007)2802% Male population28030.907***28041.745***28051.120***28060.870***28070.647***28081.337***2809(0.229)2810(0.183)2811(0.102)2812(0.088)2813(0.091)2814(0.209)2815% White population28160.747***28170.978***28180.762***28190.756***28200.751***28210.728***2822(0.042)2823(0.162)2824(0.021)2825(0.024)2826(0.018)2827(0.020)2828% Asian population28290.00828300.0072831 − 0.00228320.077**28330.146***28340.232***2835(0.053)2836(0.130)2837(0.022)2838(0.032)2839(0.023)2840(0.040)2841% Old population28420.268***28430.381***28440.091***28450.059**28460.065*28470.416***2848(0.038)2849(0.083)2850(0.032)2851(0.027)2852(0.035)2853(0.054)2854Number of crimes (IV)2855 − 1.255**2856 − 0.9002857 − 1.298***2858 − 1.062***2859 − 0.994***2860 − 1.371***2861(0.534)2862(0.776)2863(0.185)2864(0.163)2865(0.109)2866(0.163)2867Distance to crime hotspot2868 − 0.073***2869 − 0.073***2870 − 0.068***2871 − 0.070***2872 − 0.074***2873 − 0.085***2874(0.003)2875(0.002)2876(0.001)2877(0.001)2878(0.001)2879(0.001)2880Constant288112.241***288211.742***288312.183***288412.235***288512.341***288612.109***2887(0.102)2888(0.136)2889(0.053)2890(0.035)2891(0.050)2892(0.071)2893Year dummies2894Yes2895Yes2896Yes2897Yes2898Yes2899Yes2900Month dummies2901Yes2902Yes2903Yes2904Yes2905Yes2906Yes2907N290861,902290961,902291061,902291161,902291261,902291361,9022914Panel B: Using 911 Calls (High Priority), with spatial autocorrelation adjustments2915Dependent variable:2916(1)2917(2)2918(3)2919(4)2920(5)2921(6)2922Log (Sale Price)29232SLS2924Q: 0.052925Q: 0.252926Q: 0.502927Q: 0.752928Q: 0.952929Log (WY)29300.389***29310.402***29320.363***29330.360***29340.361***29350.373***2936(0.012)2937(0.017)2938(0.007)2939(0.003)2940(0.007)2941(0.013)2942Living area29430.261***29440.240***29450.252***29460.266***29470.285***29480.300***2949(0.004)2950(0.006)2951(0.003)2952(0.003)2953(0.002)2954(0.005)295529562957===== PAGE 42 (score 222, numeric-paren 183, table-words 3, var-words 14) =====295829595182960 2961J. G. Kallberg, Y. Shimizu 2962Table 8 (continued)2963Lot area29640.018***29650.007***29660.013***29670.018***29680.022***29690.033***2970(0.001)2971(0.002)2972(0.000)2973(0.001)2974(0.001)2975(0.001)2976Bedrooms2977 − 0.007***2978 − 0.0012979 − 0.002*2980 − 0.002**2981 − 0.002*2982 − 0.009**2983(0.001)2984(0.004)2985(0.001)2986(0.001)2987(0.001)2988(0.004)2989Bathrooms (Full)29900.008***2991 − 0.007***29920.006***29930.009***29940.009***29950.014***2996(0.001)2997(0.002)2998(0.002)2999(0.002)3000(0.002)3001(0.002)3002Age3003 − 0.001***3004 − 0.002***3005 − 0.001***3006 − 0.001***3007 − 0.000***3008 − 0.000***3009(0.000)3010(0.000)3011(0.000)3012(0.000)3013(0.000)3014(0.000)3015One story3016 − 0.077***3017 − 0.055***3018 − 0.065***3019 − 0.067***3020 − 0.072***3021 − 0.085***3022(0.006)3023(0.007)3024(0.002)3025(0.002)3026(0.002)3027(0.003)3028Renovated3029 − 0.028***3030 − 0.264***3031 − 0.102***30320.021***30330.048***30340.069***3035(0.011)3036(0.013)3037(0.013)3038(0.007)3039(0.002)3040(0.011)3041Fireplace30420.025***30430.045***30440.025***30450.013***30460.010***30470.006**3048(0.004)3049(0.007)3050(0.002)3051(0.002)3052(0.002)3053(0.002)3054Lake/Mountain view30550.062***30560.035***30570.037***30580.046***30590.059***30600.106***3061(0.005)3062(0.006)3063(0.003)3064(0.003)3065(0.003)3066(0.004)3067Population30680.0033069 − 0.0113070 − 0.00130710.00530720.006*30730.014***3074(0.003)3075(0.014)3076(0.003)3077(0.004)3078(0.003)3079(0.005)3080% Male population3081 − 0.04830820.09730830.171***30840.144**30850.11530860.496***3087(0.098)3088(0.226)3089(0.065)3090(0.068)3091(0.088)3092(0.164)3093% White population30940.575***30950.844***30960.606***30970.543***30980.502***30990.443***3100(0.020)3101(0.072)3102(0.018)3103(0.014)3104(0.018)3105(0.018)3106% Asian population31070.122***31080.217**31090.088***31100.120***31110.135***31120.138***3113(0.018)3114(0.092)3115(0.030)3116(0.025)3117(0.019)3118(0.029)3119% Old population3120 − 0.173***3121 − 0.187***3122 − 0.188***3123 − 0.213***3124 − 0.228***3125 − 0.203***3126(0.025)3127(0.068)3128(0.025)3129(0.026)3130(0.017)3131(0.034)3132Number of crimes (IV)31330.409**31340.943**31350.386***31360.188**31370.0163138 − 0.285**3139(0.191)3140(0.371)3141(0.109)3142(0.088)3143(0.131)3144(0.115)3145Distance to crime hotspot3146 − 0.039***3147 − 0.037***3148 − 0.038***3149 − 0.040***3150 − 0.042***3151 − 0.047***3152(0.002)3153(0.002)3154(0.001)3155(0.001)3156(0.001)3157(0.001)3158Constant31597.530***31606.979***31617.728***31627.854***31637.934***31647.668***3165(0.176)3166(0.202)3167(0.085)3168(0.078)3169(0.073)3170(0.175)3171Year dummies3172Yes3173Yes3174Yes3175Yes3176Yes3177Yes3178Month dummies3179Yes3180Yes3181Yes3182Yes3183Yes3184Yes3185N318661,902318761,902318861,902318961,902319061,902319161,9023192This table reports the results of 2SLS regression (model 1, as estimated in Eq. 1 and Eq. 3) and instru-3193mental quantile regressions (models 2 – 6, as estimated in Eqs. 2 and 4) for 5, 25, 50, 75 and 95 per-3194cent quantiles, respectively. Number of crimes is instrumented by Number of 911 calls (High Priority), 3195defined as the number of 911 calls with priority levels of 1 (highest) or 2 (second highest), made in a 3196given year per police beat. Panel A (B) performs 2SLS and quantile regressions without (with) spatial 3197autocorrelation adjustments. Definitions of other variables are provided in Appendix 1 Table 10. Stand-3198ard errors are reported in parentheses: for 2SLS regressions (model 1), standard errors are clustered by 3199year and month. For quantile regressions (models 2 – 6), standard errors are obtained through bootstrap 3200replications. *, ** and *** denote significance at the 10, 5 and 1 percent levels, respectively320132023203===== PAGE 43 (score 262, numeric-paren 219, table-words 3, var-words 14) =====320432055193206Crime Measures and Housing Prices: an Analysis Using Quantile…3207Table 9 Robustness check – quantile regression results including distance to crime coldspot3208Panel A: Using Number of 911 Calls as an IV, without spatial autocorrelation adjustments3209Dependent variable:3210(1)3211(2)3212(3)3213(4)3214(5)3215(6)3216Log (Sale Price)32172SLS3218Q: 0.053219Q: 0.253220Q: 0.503221Q: 0.753222Q: 0.953223Living area32240.318***32250.285***32260.306***32270.321***32280.339***32290.348***3230(0.003)3231(0.005)3232(0.002)3233(0.003)3234(0.003)3235(0.005)3236Lot area32370.022***32380.009***32390.014***32400.020***32410.026***32420.038***3243(0.001)3244(0.002)3245(0.001)3246(0.001)3247(0.001)3248(0.002)3249Bedrooms3250 − 0.019***3251 − 0.008*3252 − 0.011***3253 − 0.013***3254 − 0.014***3255 − 0.018***3256(0.002)3257(0.005)3258(0.002)3259(0.002)3260(0.002)3261(0.003)3262Bathrooms (Full)32630.004*3264 − 0.009**32650.00332660.006***32670.005**32680.011***3269(0.002)3270(0.004)3271(0.002)3272(0.002)3273(0.002)3274(0.004)3275Age3276 − 0.000***3277 − 0.001***3278 − 0.000***3279 − 0.000***3280 − 0.000**3281 − 0.0003282(0.000)3283(0.000)3284(0.000)3285(0.000)3286(0.000)3287(0.000)3288One story3289 − 0.092***3290 − 0.063***3291 − 0.073***3292 − 0.079***3293 − 0.087***3294 − 0.099***3295(0.007)3296(0.005)3297(0.003)3298(0.003)3299(0.003)3300(0.005)3301Renovated3302 − 0.046***3303 − 0.278***3304 − 0.110***3305 − 0.00333060.023***33070.046***3308(0.011)3309(0.023)3310(0.014)3311(0.008)3312(0.007)3313(0.013)3314Fireplace33150.039***33160.062***33170.038***33180.027***33190.020***33200.013***3321(0.005)3322(0.007)3323(0.002)3324(0.002)3325(0.001)3326(0.002)3327Lake/Mountain view33280.109***33290.066***33300.074***33310.085***33320.106***33330.180***3334(0.006)3335(0.007)3336(0.004)3337(0.004)3338(0.006)3339(0.009)3340Population33410.014***3342 − 0.01933430.00133440.016***33450.022***33460.026***3347(0.005)3348(0.013)3349(0.006)3350(0.002)3351(0.003)3352(0.005)3353% Male population33540.614***33551.484***33560.858***33570.629***33580.452***33591.008***3360(0.133)3361(0.272)3362(0.115)3363(0.117)3364(0.079)3365(0.241)3366% White population33670.776***33681.106***33690.801***33700.754***33710.758***33720.738***3373(0.037)3374(0.154)3375(0.023)3376(0.017)3377(0.021)3378(0.023)3379% Asian population33800.02033810.1113382 − 0.00033830.067***33840.136***33850.221***3386(0.032)3387(0.107)3388(0.024)3389(0.016)3390(0.018)3391(0.046)3392% Old population33930.329***33940.427***33950.196***33960.119***33970.127***33980.376***3399(0.041)3400(0.101)3401(0.037)3402(0.021)3403(0.027)3404(0.045)3405Number of crimes (IV)3406 − 0.694**34070.0933408 − 0.588***3409 − 0.713***3410 − 0.659***3411 − 1.007***3412(0.305)3413(0.744)3414(0.164)3415(0.097)3416(0.120)3417(0.202)3418Distance to crime hotspot3419 − 0.072***3420 − 0.071***3421 − 0.067***3422 − 0.069***3423 − 0.073***3424 − 0.084***3425(0.003)3426(0.002)3427(0.001)3428(0.001)3429(0.001)3430(0.002)3431Distance to crime coldspot3432 − 0.010***3433 − 0.011***3434 − 0.011***3435 − 0.009***3436 − 0.008***3437 − 0.006***3438(0.001)3439(0.001)3440(0.000)3441(0.001)3442(0.001)3443(0.001)3444Constant344512.359***344611.687***344712.260***344812.353***344912.443***345012.282***3451(0.067)3452(0.196)3453(0.090)3454(0.057)3455(0.046)3456(0.078)3457Year dummies3458Yes3459Yes3460Yes3461Yes3462Yes3463Yes3464Month dummies3465Yes3466Yes3467Yes3468Yes3469Yes3470Yes3471N347261,898347361,898347461,898347561,898347661,898347761,8983478Panel B: Using Number of 911 Calls as an IV, with spatial autocorrelation adjustments3479Dependent variable:3480(1)3481(2)3482(3)3483(4)3484(5)3485(6)3486Log (Sale Price)34872SLS3488Q: 0.053489Q: 0.253490Q: 0.503491Q: 0.753492Q: 0.953493Log (WY)34940.389***34950.409***34960.364***34970.359***34980.360***34990.371***3500(0.013)3501(0.014)3502(0.009)3503(0.008)3504(0.007)3505(0.011)350635073508===== PAGE 44 (score 244, numeric-paren 202, table-words 3, var-words 15) =====350935105203511 3512J. G. Kallberg, Y. Shimizu 3513Table 9 (continued)3514Living area35150.260***35160.239***35170.250***35180.267***35190.285***35200.299***3521(0.004)3522(0.005)3523(0.002)3524(0.002)3525(0.003)3526(0.005)3527Lot area35280.019***35290.008***35300.014***35310.018***35320.022***35330.034***3534(0.001)3535(0.001)3536(0.001)3537(0.000)3538(0.001)3539(0.002)3540Bedrooms3541 − 0.007***35420.0003543 − 0.0013544 − 0.0023545 − 0.003*3546 − 0.008***3547(0.001)3548(0.005)3549(0.002)3550(0.001)3551(0.001)3552(0.003)3553Bathrooms (Full)35540.009***3555 − 0.00535560.007***35570.010***35580.009***35590.016***3560(0.001)3561(0.004)3562(0.002)3563(0.001)3564(0.002)3565(0.003)3566Age3567 − 0.001***3568 − 0.002***3569 − 0.001***3570 − 0.001***3571 − 0.000***3572 − 0.000***3573(0.000)3574(0.000)3575(0.000)3576(0.000)3577(0.000)3578(0.000)3579One story3580 − 0.076***3581 − 0.054***3582 − 0.063***3583 − 0.066***3584 − 0.072***3585 − 0.084***3586(0.005)3587(0.004)3588(0.002)3589(0.003)3590(0.003)3591(0.004)3592Renovated3593 − 0.029***3594 − 0.265***3595 − 0.104***35960.020***35970.048***35980.070***3599(0.011)3600(0.016)3601(0.015)3602(0.007)3603(0.007)3604(0.011)3605Fireplace36060.025***36070.043***36080.024***36090.012***36100.010***36110.007***3612(0.004)3613(0.006)3614(0.003)3615(0.002)3616(0.002)3617(0.002)3618Lake/Mountain view36190.065***36200.040***36210.039***36220.048***36230.063***36240.110***3625(0.005)3626(0.005)3627(0.003)3628(0.004)3629(0.004)3630(0.008)3631Population36320.0013633 − 0.0153634 − 0.00336350.00436360.00436370.014***3638(0.002)3639(0.014)3640(0.005)3641(0.003)3642(0.004)3643(0.005)3644% Male population3645 − 0.217**3646 − 0.32236470.01036480.0363649 − 0.00336500.332***3651(0.094)3652(0.249)3653(0.126)3654(0.077)3655(0.100)3656(0.095)3657% White population36580.592***36590.873***36600.618***36610.541***36620.507***36630.453***3664(0.030)3665(0.067)3666(0.020)3667(0.012)3668(0.017)3669(0.019)3670% Asian population36710.129***36720.210**36730.090***36740.107***36750.133***36760.140***3677(0.018)3678(0.092)3679(0.027)3680(0.017)3681(0.021)3682(0.022)3683% Old population3684 − 0.138***3685 − 0.224***3686 − 0.157***3687 − 0.175***3688 − 0.205***3689 − 0.211***3690(0.032)3691(0.076)3692(0.041)3693(0.019)3694(0.020)3695(0.041)3696Number of crimes (IV)36970.734***36981.338***36990.612***37000.354***37010.195**3702 − 0.0723703(0.148)3704(0.234)3705(0.167)3706(0.111)3707(0.096)3708(0.126)3709Distance to crime hotspot3710 − 0.039***3711 − 0.036***3712 − 0.038***3713 − 0.040***3714 − 0.042***3715 − 0.046***3716(0.002)3717(0.002)3718(0.001)3719(0.001)3720(0.001)3721(0.001)3722Distance to crime coldspot3723 − 0.006***3724 − 0.007***3725 − 0.006***3726 − 0.005***3727 − 0.005***3728 − 0.004***3729(0.001)3730(0.001)3731(0.000)3732(0.000)3733(0.000)3734(0.001)3735Constant37367.602***37377.085***37387.801***37397.929***37408.008***37417.761***3742(0.174)3743(0.251)3744(0.115)3745(0.102)3746(0.129)3747(0.176)3748Year dummies3749Yes3750Yes3751Yes3752Yes3753Yes3754Yes3755Month dummies3756Yes3757Yes3758Yes3759Yes3760Yes3761Yes3762N376361,898376461,898376561,898376661,898376761,898376861,8983769This table reports the results of 2SLS regressions, as well as those of quantile regressions by adding the 3770Distance to crime coldspot variable to the estimations. Panel A (B) performs 2SLS and quantile regres-3771sions without (with) spatial autocorrelation adjustments. Distance to crime coldspot is defined as the 3772straight mile distance to the nearest crime coldspot, as identified by the Getis-Ord statistics. Definitions 3773of other variables are provided in Appendix 1 Table 10. Standard errors are reported in parentheses: for 37742SLS regressions (model 1), standard errors are clustered by year and month. For quantile regressions 3775(models 2 – 6), standard errors are obtained through bootstrap replications. *, ** and *** denote signifi-3776cance at the 10, 5 and 1 percent levels, respectively377737783779===== PAGE 45 (score 21, numeric-paren 0, table-words 2, var-words 17) =====378037815213782Crime Measures and Housing Prices: an Analysis Using Quantile…3783These limitations suggest several topics for further research, in particular 3784the impact of policing efforts. As noted, enforcement levels are related to crime 3785rates and to the frequency of 911 calls, complicating the already intricate rela-3786tions between crime and the pricing of real estate. In addition, the link between 3787actual crimes and 911calls requires additional analysis.3788Appendix 13789````3790Table 10 Variable definitions3791Variable3792Definition3793Transaction prices3794 Sale price3795Sale price of property ($)3796 Log (Sale Price)3797Log of sale price3798 WY3799The weighted average sale price ($) of neighbor properties 3800within 2000 feet3801 Log (WY)3802Log of WY3803Property-level characteristics3804 Living area3805Total living area in ft23806 Lot area3807Total lot area in ft23808 Bedrooms3809# of bedrooms3810 Bathrooms3811# of full-size bathrooms3812 One story3813A dummy variable equal to 1 for one-story property; zero 3814otherwise3815 Year built3816Year when property was built3817 Renovated3818A dummy variable equal to 1 if previously renovated; zero 3819otherwise3820 Age3821Year of sale—Year built3822 Fireplace3823A dummy variable equal to 1 for property with fireplace; zero 3824otherwise3825Environmental—natual views3826 Lake/Mountain view3827A dummy variable equal to 1 for property with natural view (Mt 3828Rainier, Cascades, Puget Sound, Lake Washington, Olympics, 3829Others); zero otherwise3830Crime variables3831 Number of crimes3832Number of reported crimes per census tract, divided by popula-3833tion per census tract3834 Change in crime #3835% change in number of reported crime from 3836previous year (t-1) per census tract, or 3837log3838(38391 + number of crimest3840) −log(1 + number of crimest−13841)384238433844===== PAGE 47 (score 254, numeric-paren 212, table-words 3, var-words 14) =====384538465233847Crime Measures and Housing Prices: an Analysis Using Quantile…3848Appendix 2 3849Table 11 Baseline results with zip code fixed effects3850Panel A: Without spatial autocorrelation adjustments3851Dependent variable:3852(1)3853(2)3854(3)3855(4)3856(5)3857(6)3858Log (Sale Price)38592SLS3860Q: 0.053861Q: 0.253862Q: 0.503863Q: 0.753864Q: 0.953865Living area38660.288***38670.255***38680.280***38690.294***38700.306***38710.312***3872(0.004)3873(0.006)3874(0.003)3875(0.002)3876(0.003)3877(0.004)3878Lot area38790.026***38800.016***38810.021***38820.024***38830.030***38840.041***3885(0.001)3886(0.001)3887(0.001)3888(0.001)3889(0.001)3890(0.002)3891Bedrooms3892 − 0.016***3893 − 0.0033894 − 0.009***3895 − 0.010***3896 − 0.011***3897 − 0.017***3898(0.002)3899(0.003)3900(0.002)3901(0.002)3902(0.001)3903(0.003)3904Bathrooms (Full)39050.007***39060.00139070.006***39080.008***39090.008***39100.013***3911(0.002)3912(0.004)3913(0.002)3914(0.002)3915(0.003)3916(0.004)3917Age3918 − 0.001***3919 − 0.001***3920 − 0.001***3921 − 0.001***3922 − 0.000***3923 − 0.000***3924(0.000)3925(0.000)3926(0.000)3927(0.000)3928(0.000)3929(0.000)3930One story3931 − 0.076***3932 − 0.053***3933 − 0.058***3934 − 0.065***3935 − 0.075***3936 − 0.086***3937(0.006)3938(0.006)3939(0.003)3940(0.003)3941(0.004)3942(0.004)3943Renovated3944 − 0.033***3945 − 0.265***3946 − 0.106***39470.018**39480.041***39490.052***3950(0.011)3951(0.015)3952(0.010)3953(0.009)3954(0.004)3955(0.009)3956Fireplace39570.020***39580.036***39590.021***39600.014***39610.006***3962 − 0.006**3963(0.005)3964(0.005)3965(0.002)3966(0.002)3967(0.001)3968(0.003)3969Lake/Mountain view39700.140***39710.094***39720.107***39730.114***39740.138***39750.211***3976(0.006)3977(0.004)3978(0.003)3979(0.003)3980(0.006)3981(0.010)3982Population39830.073***39840.080***39850.075***39860.071***39870.073***39880.087***3989(0.008)3990(0.007)3991(0.005)3992(0.003)3993(0.006)3994(0.014)3995% Male population39960.355*39970.304*39980.23339990.296**40000.627***40011.935***4002(0.188)4003(0.162)4004(0.145)4005(0.127)4006(0.114)4007(0.283)4008% White population40090.261***40100.262***40110.285***40120.250***40130.242***40140.1044015(0.043)4016(0.083)4017(0.032)4018(0.032)4019(0.036)4020(0.065)4021% Asian population4022 − 0.493***4023 − 0.437***4024 − 0.399***4025 − 0.418***4026 − 0.405***4027 − 0.576***4028(0.060)4029(0.078)4030(0.052)4031(0.033)4032(0.028)4033(0.061)4034% Old population40350.664***40360.246***40370.283***40380.493***40390.678***40401.148***4041(0.060)4042(0.054)4043(0.029)4044(0.048)4045(0.031)4046(0.183)4047Number of crimes (IV)4048 − 1.849***4049 − 3.070***4050 − 2.646***4051 − 2.263***4052 − 1.898***4053 − 1.870***4054(0.376)4055(0.515)4056(0.185)4057(0.180)4058(0.209)4059(0.336)4060Distance to crime hotspot4061 − 0.071***4062 − 0.082***4063 − 0.074***4064 − 0.071***4065 − 0.068***4066 − 0.065***4067(0.004)4068(0.005)4069(0.001)4070(0.001)4071(0.002)4072(0.002)4073Constant407412.432***407512.398***407612.487***407712.507***407812.330***407911.711***4080(0.113)4081(0.097)4082(0.074)4083(0.058)4084(0.059)4085(0.212)4086Year dummies4087Yes4088Yes4089Yes4090Yes4091Yes4092Yes4093Month dummies4094Yes4095Yes4096Yes4097Yes4098Yes4099Yes4100Zip area dummies4101Yes4102Yes4103Yes4104Yes4105Yes4106Yes4107N410861,902410961,902411061,902411161,902411261,902411361,9024114Panel B: With spatial autocorrelation adjustments4115Dependent variable:4116(1)4117(2)4118(3)4119(4)4120(5)4121(6)4122Log (Sale Price)41232SLS4124Q: 0.054125Q: 0.254126Q: 0.504127Q: 0.754128Q: 0.954129Log (WY)41300.292***41310.282***41320.271***41330.265***41340.262***41350.266***4136(0.013)4137(0.018)4138(0.009)4139(0.006)4140(0.010)4141(0.017)414241434144===== PAGE 48 (score 230, numeric-paren 191, table-words 2, var-words 15) =====414541465244147 4148J. G. Kallberg, Y. Shimizu 4149Table 11 (continued)4150Living area41510.256***41520.231***41530.249***41540.265***41550.281***41560.290***4157(0.004)4158(0.005)4159(0.002)4160(0.002)4161(0.002)4162(0.004)4163Lot area41640.023***41650.012***41660.018***41670.022***41680.027***41690.038***4170(0.001)4171(0.001)4172(0.001)4173(0.001)4174(0.000)4175(0.001)4176Bedrooms4177 − 0.008***4178 − 0.0014179 − 0.002**4180 − 0.003*4181 − 0.004***4182 − 0.008***4183(0.001)4184(0.004)4185(0.001)4186(0.002)4187(0.001)4188(0.003)4189Bathrooms (Full)41900.009***41910.00041920.008***41930.010***41940.009***41950.015***4196(0.001)4197(0.004)4198(0.002)4199(0.002)4200(0.002)4201(0.002)4202Age4203 − 0.001***4204 − 0.002***4205 − 0.001***4206 − 0.001***4207 − 0.001***4208 − 0.000***4209(0.000)4210(0.000)4211(0.000)4212(0.000)4213(0.000)4214(0.000)4215One story4216 − 0.070***4217 − 0.054***4218 − 0.058***4219 − 0.060***4220 − 0.067***4221 − 0.080***4222(0.005)4223(0.005)4224(0.003)4225(0.003)4226(0.002)4227(0.004)4228Renovated4229 − 0.025**4230 − 0.258***4231 − 0.102***42320.028***42330.053***42340.067***4235(0.011)4236(0.013)4237(0.014)4238(0.008)4239(0.003)4240(0.007)4241Fireplace42420.016***42430.032***42440.015***42450.008***42460.0024247 − 0.0014248(0.004)4249(0.005)4250(0.002)4251(0.002)4252(0.002)4253(0.004)4254Lake/Mountain view42550.095***42560.059***42570.066***42580.076***42590.094***42600.158***4261(0.005)4262(0.005)4263(0.003)4264(0.002)4265(0.004)4266(0.006)4267Population42680.031***42690.056***42700.045***42710.036***42720.028***42730.029***4274(0.004)4275(0.017)4276(0.003)4277(0.006)4278(0.007)4279(0.010)4280% Male population4281 − 0.315***42820.0894283 − 0.0904284 − 0.1374285 − 0.06142860.2104287(0.104)4288(0.100)4289(0.111)4290(0.089)4291(0.117)4292(0.273)4293% White population42940.359***42950.270***42960.309***42970.309***42980.334***42990.299***4300(0.023)4301(0.076)4302(0.026)4303(0.036)4304(0.021)4305(0.045)4306% Asian population4307 − 0.149***4308 − 0.200***4309 − 0.159***4310 − 0.151***4311 − 0.122***4312 − 0.166***4313(0.034)4314(0.060)4315(0.033)4316(0.027)4317(0.032)4318(0.046)4319% Old population43200.098**43210.00943220.02343230.088***43240.143***43250.198**4326(0.039)4327(0.054)4328(0.031)4329(0.026)4330(0.034)4331(0.094)4332Number of crimes (IV)4333 − 0.0554334 − 1.430***4335 − 0.956***4336 − 0.632***4337 − 0.249**4338 − 0.1244339(0.192)4340(0.521)4341(0.196)4342(0.195)4343(0.099)4344(0.274)4345Distance to crime hotspot4346 − 0.045***4347 − 0.056***4348 − 0.051***4349 − 0.048***4350 − 0.046***4351 − 0.044***4352(0.004)4353(0.004)4354(0.002)4355(0.001)4356(0.001)4357(0.002)4358Constant43598.929***43608.748***43619.057***43629.237***43639.280***43649.186***4365(0.197)4366(0.267)4367(0.147)4368(0.083)4369(0.107)4370(0.120)4371Year dummies4372Yes4373Yes4374Yes4375Yes4376Yes4377Yes4378Month dummies4379Yes4380Yes4381Yes4382Yes4383Yes4384Yes4385Zip area dummies4386Yes4387Yes4388Yes4389Yes4390Yes4391Yes4392N439361,902439461,902439561,902439661,902439761,902439861,9024399This table reports the results of baseline regressions (as estimated in Eq. 1 – 4) with zip code fixed 4400effects. Panel A reports the results from Eq. 1 (2SLS) and 2 (quantile regressions) without spatial auto-4401correlation adjustments. Panel B reports the results from Eq. 3 (2SLS) and 4 (instrumental variable quan-4402tile regressions) with spatial autocorrelation adjustments. Standard errors are reported in parentheses: for 44032SLS regressions (model 1), standard errors are clustered by year and month. For quantile regressions 4404(models 2 – 6), standard errors are obtained through bootstrap replications. *, ** and *** denote signifi-4405cance at the 10, 5 and 1 percent levels, respectively440644074408===== PAGE 49 (score 7, numeric-paren 0, table-words 0, var-words 7) =====440944105254411Crime Measures and Housing Prices: an Analysis Using Quantile…4412Acknowledgements We are very grateful to the editor, R. Kelley Pace, and to an anonymous referee 4413for suggestions that have significantly improved this paper. We would like to thank Geoffrey Bell, Chris 4414Cunningham, Eric DeVries (Port of Seattle Police Department), D. J. Fairhurst, Andrew Fluharty, Joseph 4415Grant, Chinmoy Ghosh, Shinichi Hirota, Toshiki Honda, Peggy Kallberg, Jin Man Lee, Dahui Li, Jiawen 4416Liu, Daisuke Miyakawa, Hidetoshi Nakagawa, Nobuhiro Nakamura, Mikiharu Noma, Kazuhiko Ohashi, 4417Yusuke Osaki, Liang Peng, Katsushi Suzuki, Daisuke Yokouchi, Wenqing Zhang and seminar and con-4418ference participants at the 2022 AsRes-AREUEA Conference in Tokyo, the 2022 FMA Conference in 4419Atlanta, Hitotsubashi University, University of Massachusetts, Dartmouth, University of Minnesota, 4420Duluth, Waseda University, and Washington State University for helpful comments.4421Funding Open access funding provided by Hitotsubashi University. The authors did not receive support 4422from any organization for the submitted work.4423Data Availability The data analyzed in this study come from the King County Assessor’s website (https://4424kingcounty.gov/depts/assessor.aspx), the Washington State Office of Financial Management website (https://4425ofm.wa.gov/), the Seattle Police Department website (https://www.seattle.gov/police), and the Seattle Open 4426Data Portal (https://data.seattle.gov/). The geographic data were analyzed using Maptitude Mapping Soft-4427ware (Caliper Corporation), GeoDa (https://geodacenter.github.io/), and Texas A&M University Geoser-4428vices (https://geoservices.tamu.edu/). The data from Caliper Corporation were used under license.4429Code Availability Not applicable.4430Declarations 4431Conflicts of Interests/Competing Interests The authors have no relevant financial or non-financial interests 4432to disclose.4433Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, 4434which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long 4435as you give appropriate credit to the original author(s) and the source, provide a link to the Creative 4436Commons licence, and indicate if changes were made. The images or other third party material in this 4437article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line 4438to the material. If material is not included in the article’s Creative Commons licence and your intended 4439use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permis-4440sion directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/4441licenses/by/4.0/.4442References4443Abbott, J. K., & Klaiber, H. A. (2011). An embarrassment of riches: Confronting omitted variable 4444bias and multi-scale capitalization in hedonic price models. Review of Economics and Statistics, 444593, 1331–1342. https://doi.org/10.1162/REST_a_001344446Andresen, M. A., & Malleson, N. (2013). Crime seasonality and its variations across space. Applied 4447Geography, 43, 25–35. https://doi.org/10.1016/j.apgeog.2013.06.0074448Anselin, L., Cohen, J., Cook, D., Gorr, W., & Tita, G. (2000). Spatial analyses of crime. Criminal 4449Justice, 4, 213–262.4450Becker, G. S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76, 4451169–217.4452Boggess, L. N., Greenbaum, R. T., & Tita, G. E. (2013). Does crime drive housing sales? Evidence 4453from Los Angeles. Journal of Crime and Justice, 36, 299–318. https://doi.org/10.1080/07356445448X.2013.8129764455Braakmann, N. (2012). The link between non-property crime and house prices: Evidence from UK 4456street-level data. MPRA Paper No. 44884.4457