1# RECORD 98c105ab42ddbb39da40f571a9e40579ebb0b16f2# TITLE: The impact of the Coronavirus pandemic on New York City real estate: First evidence3# YEAR: 2021.0 VENUE: Journal of Regional Science DOI: 10.1111/jors.125914# AUTHORS: Jeffrey P. Cohen; Felix L. Friedt; Jackson P. Lautier5# PAGES: 38 CANDIDATE TABLE PAGES (1-based): [22, 25, 28, 31]6# ABSTRACT: Abstract We investigate whether pandemic‐induced contagion disamenities and income effects arising due to COVID‐related unemployment adversely affected real estate prices of one‐ or two‐family owner‐occupied properties across New York City (NYC). First, ordinary least squares hedonic results indicate that greater COVID case numbers are concentrated in neighborhoods with lower‐valued properties. Second, we use a repeat‐sales approach for the period 2003–2020, and we find that both the possibility of contagion and pandemic‐induced income effects adversely impacted home sale prices. Estimates suggest sale prices fell by roughly $60,000 or around 8% in response to both of the following: 1000 additional infections per 100,000 residents and a 10‐percentage point increase in unemployment in a given Modified Zip Code Tabulation Area (MODZCTA). These price effects were more pronounced during the second wave of infections. On the basis of cumulative MODZCTA infection rates through 2020, the estimated COVID‐19 price discount ranged from approximately 1% to 50% in the most affected neighborhoods, and averaged 14%. The contagion effect intensified in the more affluent, but less densely populated NYC neighborhoods, while the income effect was more pronounced in the most densely populated neighborhoods with more rental properties and greater population shares of foreign‐born residents. This disparity implies the pandemic may have been correlated with a wider gap in housing wealth in NYC between homeowners in lower‐priced and higher‐priced neighborhoods.789===== PAGE 1 (score 10, numeric-paren 0, table-words 0, var-words 10) =====10111 12The Impact of the Coronavirus Pandemic on New York City 13Real Estate: First Evidence 14 15 16Abstract: We investigate whether pandemic-induced contagion disamenities and income effects arising 17due to COVID-related unemployment adversely affected real estate prices of one- or two-family owner-18occupied properties across New York City (NYC). First, OLS hedonic results indicate that greater COVID 19case numbers are concentrated in neighborhoods with lower-valued properties. Second, we use a repeat-20sales approach for the period 2003 to 2020, and we find that both the possibility of contagion and 21pandemic-induced income effects adversely impacted home sale prices. Estimates suggest sale prices fell 22by roughly $60,000 or around 8% in response to both of the following: 1,000 additional infections per 23100,000 residents; and a 10-percentage point increase in unemployment in a given Modified Zip Code 24Tabulation Area (MODZCTA). These price effects were more pronounced during the second wave of 25infections. Based on cumulative MODZCTA infection rates through 2020, the estimated COVID-19 price 26discount ranged from approximately 1% to 50% in the most affected neighborhoods, and averaged 14%. 27The contagion effect intensified in the more affluent, but less densely populated NYC neighborhoods, 28while the income effect was more pronounced in the most densely populated neighborhoods with more 29rental properties and greater population shares of foreign-born residents. This disparity implies the 30pandemic may have been correlated with a wider gap in housing wealth in NYC between homeowners in 31lower-priced and higher-priced neighborhoods. 32Keywords: COVID-19, Hedonic, Repeat Sales, Price Discount, Housing Wealth Inequality 33JEL Codes: R31 34 35∗ Jeffrey P. Cohen, Kinnard Scholar in Real Estate and Professor of Finance, Department of Finance and Center for Real Estate 36and Urban Economic Studies, School of Business, University of Connecticut, 2100 Hillside Rd, Storrs, CT 06269; and Research 37Fellow, Institute for Economic Equity, Federal Reserve Bank of St. Louis. jeffrey.cohen@business.uconn.edu 38† Felix L. Friedt, Assistant Professor of Economics, Department of Economics, Macalester College, 1600 Grand Avenue, Saint 39Paul, MN 55105; ffriedt@macalester.edu. 40‡ Jackson Lautier, Department of Statistics, University of Connecticut, Philip E. Austin Building, 215 Glenbrook Road, Storrs, 41CT 06269-4120, jackson.lautier@uconn.edu. 42 43Acknowledgements: Cohen’s work on this paper was completed in part while he has been a Research Fellow at the Institute for 44Economic Equity, Federal Reserve Bank of St. Louis. Lautier’s efforts on this paper were based upon work supported in part by the 45National Science Foundation Graduate Research Fellowship under Grant No. DHE 1747453. The authors would also like to thank 46Jason Barr for his helpful suggestions on the availability of related New York City datasets. Session participants at the 2022 47ASSA/AREUEA Annual Meetings (virtual) provided helpful comments. Any remaining errors are our own. The views expressed 48in this paper are those of the authors and do not necessarily represent the Federal Reserve Bank of St. Louis, the Federal Reserve 49Board, or the Federal Reserve System. Revised 1/27/2022. 50Jeffrey P. Cohen∗ 51Felix L. Friedt† 52Jackson P. Lautier‡ 53University of Connecticut 54Macalester College 55University of Connecticut 565758===== PAGE 22 (score 162, numeric-paren 141, table-words 3, var-words 15) =====596022 61 62Table 2: Changes in Pre- and Post-Pandemic NYC Real Estate Market Characteristics 63 64 (1) 65(2) 66(3) 67(4) 68(5) 69(6) 70(7) 71(8) 72(9) 73(10) 74 75Sales of once-sold Properties 76Sales of repeatedly-sold Properties 77Characteristic 78Pre- 79Pre- 80Post- 81Post- 82∆ in 83Pre- 84Pre- 85Post- 86Post- 87∆ in 88Mean 89S.D. 90Mean 91S.D. 92Means 93Mean 94S.D. 95Mean 96S.D. 97Means 98Panel A: First Wave of Infections (03/01/2019-07/31/2019 vs. 03/03/2020-07/31/2020) 99Sale Price ($ ‘000) 100784.74 101(600.4) 102793.64 103(579.2) 1048.90 105660.50 106(470.3) 107763.98 108(384.9) 109103.48*** 110# of Sales 11181.20 112(60.3) 11349.43 114(34.7) 115-31.78*** 11610.24 117(6.5) 1185.09 119(3.2) 120-5.15*** 121Unemployment Rate (%) 1224.04 123(2.8) 12414.07 125(8.1) 12610.03*** 1275.13 128(3.4) 12915.48 130(8.1) 13110.35*** 132Renovated Home (yes==1) 1330.04 134(0.2) 1350.05 136(0.2) 1370.01 1380.03 139(0.2) 1400.04 141(0.2) 1420.01 143Building Area (’00 sqft) 14424.34 145(164.6) 14628.95 147(222.0) 1484.61 14916.57 150(6.2) 15117.52 152(7.4) 1530.95 154Lot Area (’000 sqft) 1557.25 156(109.4) 15710.93 158(162.4) 1593.68 1602.88 161(1.5) 1622.90 163(1.9) 1640.02 165# of Floors 1662.03 167(0.5) 1682.03 169(0.5) 170-0.00 1712.04 172(0.5) 1732.09 174(0.5) 1750.05 176Basement (yes==1) 1771.00 178(0.0) 1791.00 180(0.0) 1810.00 1821.00 183(0.0) 1841.00 185(0.0) 1860.00 187Population (’000) 18853.94 189(25.2) 19054.89 191(25.3) 1920.95 19354.27 194(24.9) 19557.51 196(24.0) 1973.24 198Share of White Residents (%) 19939.46 200(17.1) 20138.05 202(16.9) 203-1.41*** 20438.56 205(18.3) 20638.96 207(18.0) 2080.40 209Share of Rental Units (%) 21045.95 211(19.9) 21246.45 213(20.0) 2140.50 21547.16 216(20.5) 21748.07 218(21.5) 2190.90 220Density per Unit 2212.79 222(0.4) 2232.78 224(0.4) 225-0.02 2262.88 227(0.4) 2282.86 229(0.4) 230-0.03 231English Proficiency 23211.71 233(10.2) 23411.73 235(10.2) 2360.02 2379.73 238(8.9) 23910.90 240(9.9) 2411.17 242Median Income ($ ’000) 24373.67 244(18.6) 24572.98 246(18.2) 247-0.69 24870.26 249(19.0) 25069.64 251(20.5) 252-0.61 253Share College Educated (%) 25456.33 255(8.9) 25656.30 257(8.8) 258-0.03 25953.91 260(8.9) 26153.85 262(9.9) 263-0.06 264Observations 2655,670 2663,471 2679,141 268542 269277 270819 271Panel B: Second Wave of Infections (08/01/2019-12/31/2019 vs. 08/01/2020-12/31/2020) 272Sale Price ($ ’000) 273769.41 274(517.7) 275806.35 276(754.2) 27736.94** 278647.07 279(381.2) 280740.55 281(483.4) 28293.48** 283# of Sales 28489.37 285(63.4) 28690.64 287(74.8) 2881.27 2899.26 290(6.4) 2918.79 292(9.2) 293-0.47 294Unemployment Rate (%) 2953.59 296(2.5) 29712.09 298(4.0) 2998.50*** 3004.41 301(2.9) 30212.59 303(3.7) 3048.18*** 305Renovated Home (yes==1) 3060.04 307(0.2) 3080.04 309(0.2) 3100.00 3110.02 312(0.1) 3130.03 314(0.2) 3150.01 316Building Area (’00 sqft) 31724.42 318(170.0) 31928.29 320(211.5) 3213.87 32232.48 323(258.6) 32429.69 325(243.7) 326-2.80 327Lot Area (’000 sqft) 3287.76 329(120.8) 3309.87 331(143.7) 3322.11 33312.05 334(166.3) 33513.50 336(190.8) 3371.45 338# of Floors 3392.03 340(0.5) 3412.02 342(0.5) 343-0.01 3442.00 345(0.5) 3462.02 347(0.5) 3480.02 349Basement (yes==1) 3501.00 351(0.0) 3521.00 353(0.0) 3540.00 3551.00 356(0.0) 3571.00 358(0.0) 3590.00 360Population (’000) 36154.49 362(25.6) 36352.82 364(25.2) 365-1.66*** 36653.94 367(24.5) 36854.13 369(24.1) 3700.19 371Share of White Residents (%) 37239.38 373(17.0) 37440.96 375(17.8) 3761.58*** 37740.09 378(18.6) 37943.97 380(19.2) 3813.89** 382Share of Rental Units(%) 38346.06 384(19.8) 38544.28 386(20.1) 387-1.78*** 38846.22 389(20.1) 39042.67 391(20.8) 392-3.55* 393Density per Unit 3942.79 395(0.4) 3962.76 397(0.4) 398-0.03*** 3992.86 400(0.4) 4012.79 402(0.4) 403-0.06* 404English Proficiency 40511.88 406(10.4) 40711.64 408(10.4) 409-0.24 4109.88 411(9.2) 4129.51 413(8.4) 414-0.37 415Median Income ($ ’000) 41673.64 417(18.4) 41875.43 419(18.8) 4201.79*** 42171.62 422(19.4) 42375.82 424(19.2) 4254.20** 426Share College Educated (%) 42756.39 428(8.6) 42957.00 430(8.8) 4310.62*** 43254.67 433(8.8) 43456.56 435(8.7) 4361.89** 437Observations 4386,345 4395,845 44012,190 441495 442354 443849 444Notes: Data on the number of one- and two-family property transactions and sale prices are published by the 445NYC Department of Finance. Information on NYC MODZCTA unemployment rates are published by DEEP-446MAPS. Statistics on housing characteristics are publicly available through the PLUTOTM database maintained by 447the NYC Department of City Planning. Neighborhood characteristics represent the sample averages for the 448respective samples of observed sales and are based on 2019 data available through American Community Survey 449published by the U.S. Census. Based on our sample, we define distinct post outbreak periods including the 1) the 4501st wave of infections from March 1, 2020 through July 31, 2020; and 2) 2nd wave of infections from August 1, 4512020 through December 31, 2020. For the purposes of comparison, we calculate pre-outbreak statistics based on 452a limited sample restricted to 1) March 1 to July 31, 2019; and 2) August 1 to December 31, 2019. 453 454455456===== PAGE 23 (score 0, numeric-paren 0, table-words 0, var-words 0) =====45745823 459 460 461 462Figure 3: Cumulative Case Rates and MODZCTA Changes in the # of Property Sales and Avg. Sale 463Prices by Wave of Infection 464465466===== PAGE 25 (score 120, numeric-paren 101, table-words 5, var-words 9) =====46746825 469Table 3: COVID-19 Case Effect on NYC One & Two Family Home Prices 470 471 472(1) 473(2) 474(3) 475(4) 476(5) 477(6) 478(7) 479(8) 480 481First Wave (03/01-07/31, 2003-2020) 482Second Wave (08/01-12/31, 2003-2020) 483 484Cum. Cases 485Cum. Cases Cum. 486Cases 487New 488Cases 489Cum. 490Cases 491Cum. Cases Cum. Cases 492New Cases 493Cases per 100,000 (’000) 494-0.382** 495-0.398** 496-0.706* 49710.162 498-0.611** 499-0.618*** 500-0.633*** 501-30.759*** 502(0.172) 503(0.174) 504(0.358) 505(12.208) 506(0.236) 507(0.235) 508(0.241) 509(7.905) 510Unemployment Rate (%) 511-0.064*** 512-0.039 513-0.039 514-0.072*** 515-0.060** 516-0.063*** 517(0.012) 518(0.024) 519(0.024) 520(0.013) 521(0.024) 522(0.024) 523Investor Owned (%) 5240.004 5250.004 526-0.009 527-0.009 5280.002 5290.002 530-0.011 531-0.010 532(0.009) 533(0.009) 534(0.013) 535(0.013) 536(0.010) 537(0.010) 538(0.014) 539(0.014) 540Altered (yes==1) 5411.356*** 5421.351*** 5432.432*** 5442.432*** 5451.997*** 5461.989*** 5473.056*** 5483.058*** 549(0.228) 550(0.228) 551(0.339) 552(0.340) 553(0.315) 554(0.315) 555(0.591) 556(0.591) 557Year Built 5580.002 5590.002 5600.003 5610.003 562(0.002) 563(0.002) 564(0.002) 565(0.002) 566Home Sqft (’000) 5672.174*** 5682.163*** 5692.174*** 5702.159*** 571(0.280) 572(0.281) 573(0.341) 574(0.341) 575Lot Sqft (’00,000) 5760.032** 5770.032** 5780.027* 5790.026* 580(0.014) 581(0.013) 582(0.015) 583(0.015) 584# of Floors 5850.051 5860.060 5870.045 5880.067 589(0.069) 590(0.071) 591(0.159) 592(0.158) 593Basement (yes==1) 5940.519*** 5950.502*** 5960.530*** 5970.506*** 598(0.109) 599(0.112) 600(0.133) 601(0.132) 602Constant 603-1.475 604-1.372 6050.790*** 6060.765*** 607-3.643 608-3.317 6090.753*** 6100.722*** 611(3.674) 612(3.633) 613(0.085) 614(0.083) 615(4.459) 616(4.413) 617(0.072) 618(0.069) 619N 620122,630 621121,980 62215,540 62315,540 624133,072 625132,393 62615,733 62715,733 628adj. R2 6290.606 6300.608 6310.146 6320.146 6330.551 6340.552 6350.131 6360.131 637F 63855.2*** 63947.4*** 64014.2*** 64113.4*** 64235.1*** 64332.3*** 64410.4*** 64511.5*** 646Month-of-sale FE 647Y 648Y 649Y 650Y 651Y 652Y 653Y 654Y 655Month-of-previous-sale FE 656N 657N 658Y 659Y 660N 661N 662Y 663Y 664MODZCTA FE 665Y 666Y 667N 668N 669Y 670Y 671N 672N 673Notes: Standard errors are reported in parenthesis. Across all specifications, we cluster standard errors at the 674MODZCTA level. The estimation sample underlying the results presented in columns (1) through (4) focuses on 675the pandemic-induced effects during first wave of infections and includes transactions of one or two-family 676homes sold in NYC between March and July for each year from 2003 through 2020, whereas the estimation 677sample underlying the results shown in columns (5) and (8) focus on the second wave and includes transactions 678of one or two-family homes sold in NYC between August and December for each year from 2003 through 2020. 679Estimation samples underlying the results presented in columns (1) and (2) as well as (5) and (6) include all sales 680of one or two-family home in NYC and each variable is given in levels, i.e. the dependent variable is given by 681the recorded sale price in $100,000. In contrast, estimation samples underlying the results presented in columns 682(3) and (4) as well as (7) and (8) are restricted to repeatedly-sold one or two-family properties. Accordingly, all 683variables are given in changes between sales, i.e. the dependent variable is given by the change in sale prices 684between sales in $100,000. Depending on the sample restriction and selection of control variables, these 685estimation samples include 149 to 160 MODZCTAs. The cumulative and new case rates are defined per 100,000 686MODZCTA residents. Across all estimations investor-owned properties are excluded. The variable “Investor 687Owned” controls for the share of investor-owned properties within a given home’s MODZCTA. Statistical 688significance at the conventional 10%, 5%, and 1% thresholds is given by *, **, ***, respectively. 689 690 691 692 693 694695696===== PAGE 26 (score 16, numeric-paren 0, table-words 3, var-words 10) =====69769826 699positive effect on resale values. Other time-invariant property characteristics variables drop out when 700taking first differences for the repeat-sales estimations. 701Combined, the estimates reported in columns (1) through (8) suggest that larger COVID-19 case 702numbers occur in neighborhoods of lower-valued properties (see columns (1) and (2) as well as columns 703(5) and (6)) and that a worsening of the COVID-19 outbreak – especially during the second wave of 704infections – within any given neighborhood, when controlling for unemployment changes, is correlated 705with a reduction in the average value of repeat-sales properties within that MODZCTA (columns (3), (7) 706and (8)). These findings echo some of the recent reports on COVID-19, which have shown that the 707outbreak is more concentrated among poorer (i.e., unemployed) individuals (Goldstein, 2020), who are 708more likely to own or occupy lower-valued homes. 7095.2 Price Heterogeneity Analysis 710A priori, it is unclear whether these pandemic price effects are similar across price levels or how they may 711interact with buyer attributes and preferences. More affluent home buyers, for example, may be more 712sensitive to the possibility of contagion, while wealthier neighborhoods may be less likely to be impacted 713by contagion during the pandemic (i.e., less densely populated and more opportunity to isolate and work 714remotely). In the absence of observing buyer characteristics, we rely on pre-pandemic neighborhood 715attributes to test whether the contagion and income effects on home values vary with the characteristics 716of the residents living near the location of the sale. The pre-pandemic attributes are based on the 2019 717ACS and include population density per BBL, median neighborhood income, the share of rentals among 718all residential units, and the local population shares of residents using public transit, with limited English 719proficiency, having some college education, or who are foreign born. 720Table 4 provides the coefficient estimates for these interaction terms during the first wave of infections 721(Panel A) and the second wave of the pandemic (Panel B). Column (1) of Table 4 provides the baseline 722average effect estimates. In columns (2) and (3), we present price effect estimates for homes valued in the 723bottom and top 25th percentile of the observed local pre-pandemic price distribution, respectively. The 724results reveal an interesting pattern. While the contagion price discount is orders of magnitude larger for 725726727===== PAGE 28 (score 156, numeric-paren 132, table-words 7, var-words 10) =====72872928 730Table 4: Heterogeneity of COVID-19 Case and Income Effects on Sales Prices of Repeatedly-sold Homes 731 732 733 734(1) 735(2) 736(3) 737(4) 738(5) 739(6) 740(7) 741(8) 742(9) 743(10) 744 745 746 747Base- 748Price 749Price 750Interaction w. Neighborhood Attributes 751 752 753 754line 755< 25th 756> 75th 757BBL 758Public 759Median 760College 761Rental 762Limited 763Foreign 764 765 766 767Results 768Perc. 769Perc. 770Density 771Transit 772Income 773Educated 774Share 775English 776Born 777Panel A: First Wave of Infections (03/01-07/31, 2003-2020) 778 779 780 781 782 783 784 785Cases per 100,000 (‘000) 786-0.706* 787-6.459*** 788-0.854* 789-0.848** 790-0.891** 791-0.591* 792-0.745** 793-0.972** 794-1.012*** 795-0.998*** 796 797 798 799(0.358) 800(1.364) 801(0.469) 802(0.396) 803(0.346) 804(0.339) 805(0.355) 806(0.407) 807(0.380) 808(0.365) 809Cases X Neigh. Attributes 810 811 812 8130.867*** 8140.739*** 815-0.693*** 816-0.348 8170.999*** 8180.531** 8190.501** 820 821 822 823 824 825 826(0.252) 827(0.247) 828(0.234) 829(0.212) 830(0.251) 831(0.241) 832(0.219) 833Unemployment Rate (%) 834-0.039 835-0.032 8360.019 837-0.005 838-0.020 839-0.053** 840-0.042* 8410.014 842-0.038 843-0.005 844 845 846 847(0.024) 848(0.027) 849(0.064) 850(0.025) 851(0.024) 852(0.026) 853(0.024) 854(0.026) 855(0.024) 856(0.032) 857Unemp. X Neigh. Attributes 858 859 860 861-0.080*** 862-0.064** 8630.044* 8640.008 865-0.086*** 866-0.015 867-0.055* 868 869 870 871 872 873 874(0.025) 875(0.028) 876(0.023) 877(0.027) 878(0.023) 879(0.025) 880(0.025) 881N 882 883 88415,540 8852,915 8864,360 88715,540 88815,540 88915,540 89015,540 89115,540 89215,540 89315,540 894adj. R2 895 896 8970.146 8980.168 8990.093 9000.147 9010.147 9020.147 9030.146 9040.148 9050.146 9060.146 907F 908 909 91014.2*** 9117.8*** 91216.3*** 91310.8*** 91411.3*** 91510.7*** 91610.6*** 91711.7*** 91810.3*** 91910.9*** 920 921 922 923 924 925 926 927 928 929 930 931 932 933Panel B: Second Wave of Infections (08/01-12/31, 2003-2020) 934 935 936 937 938 939 940 941Cases per 100,000 (‘000) 942-0.633*** 9430.200 944-1.242*** 945-0.694*** 946-0.806*** 947-0.549** 948-0.639** 949-0.701*** 950-0.712*** 951-0.677*** 952 953 954 955(0.241) 956(0.724) 957(0.353) 958(0.243) 959(0.236) 960(0.242) 961(0.246) 962(0.243) 963(0.232) 964(0.231) 965Case X Neigh. Attributes 966 967 968 9690.550*** 9700.636*** 971-0.276*** 972-0.095 9730.482*** 9740.253** 9750.204* 976 977 978 979 980 981 982(0.119) 983(0.127) 984(0.096) 985(0.105) 986(0.124) 987(0.109) 988(0.105) 989Unemployment Rate (%) 990-0.060** 991-0.005 992-0.070 9930.006 994-0.008 995-0.074*** 996-0.060** 9970.033 998-0.056** 999-0.044 1000 1001 1002 1003(0.024) 1004(0.027) 1005(0.053) 1006(0.027) 1007(0.025) 1008(0.027) 1009(0.024) 1010(0.032) 1011(0.024) 1012(0.032) 1013Unemp. X Neigh. Attributes 1014 1015 1016 1017-0.142*** 1018-0.158*** 10190.046 1020-0.011 1021-0.145*** 1022-0.031 1023-0.030 1024 1025 1026 1027 1028 1029 1030(0.028) 1031(0.030) 1032(0.029) 1033(0.034) 1034(0.030) 1035(0.028) 1036(0.031) 1037N 1038 1039 104015,733 10413,208 10424,856 104315,733 104415,733 104515,733 104615,733 104715,733 104815,733 104915,733 1050adj. R2 1051 1052 10530.131 10540.182 10550.070 10560.135 10570.135 10580.132 10590.131 10600.135 10610.131 10620.131 1063F 1064 1065 106610.4*** 10670.2 106813.0*** 106911.1*** 107011.4*** 10718.7*** 10727.8*** 10739.9*** 10747.9*** 10758.2*** 1076Month-of-sale FE 1077Y 1078Y 1079Y 1080Y 1081Y 1082Y 1083Y 1084Y 1085Y 1086Y 1087Month-of-previous-sale FE 1088Y 1089Y 1090Y 1091Y 1092Y 1093Y 1094Y 1095Y 1096Y 1097Y 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111Notes: Standard errors are reported in parenthesis. Across all specifications, we cluster standard errors across MODZCTAs. Panel A shows the results focused on the pandemic-induced effects 1112during first wave of infections between March and July for each year from 2003 through 2020, whereas Panel B shows results focused on the second wave between August and December for 1113each year from 2003 through 2020. The estimation samples underlying the results are restricted to repeatedly-sold non-investor owned one or two-family properties. Accordingly, all variables 1114are given in changes between sales, i.e., the dependent variable is given by the change in sale prices between sales in $100,000. Column (1) repeats the baseline results from Table 3. Columns 1115(2) and (3) show the effects for homes priced in the bottom and top quartiles of pre-pandemic sale prices. Columns (4) through (11) show case and unemployment effects and their interactions 1116with various neighborhood attributes. The interaction variables are indicators that equal 1 if a home is sold in a neighborhood with a above median characteristic. The indicators for ``BBL 1117Density", for example, marks MODZCTAs with above median population per BBL. Indicators for ``Public Transit", ``Long Commute", ``Limited English", ``College Educated", and ``Foreign 1118Born" represent MODZCTAs with above median population shares for residents that, for example, use public transit, have a commute longer than 30 minutes, are college educated, etc. 1119Statistical significance at the conventional 10%, 5%, and 1% thresholds is given by *, **, ***, respectively. 1120 1121 112211231124===== PAGE 29 (score 19, numeric-paren 0, table-words 6, var-words 7) =====1125112629 1127 1128case rates. The estimates given in columns (9) and (10) lend support to this hypothesis and show that the 1129contagion effect is roughly half for properties located in these neighborhoods. 1130Overall, these price effect dynamics across various neighborhood characteristics are similar during the 1131first and second wave of the pandemic. Interestingly, the differences in contagion price effects are less 1132pronounced during the second wave of the pandemic, while the differences in income effects intensify 1133with in neighborhoods with above median density, public transit dependence and rental shares. 11345.3 Robustness Analysis 1135We conduct a host of robustness analyses to test the sensitivity of our primary COVID-19 price effects 1136(as initially reported in columns (3) and (7) of Table 3). We begin by adjusting the restrictions on our 1137repeat-sales sample. The primary results are based on transactions of properties repeatedly sold either 1138between March and July 2003-2020 (i.e., first infection wave) or between August and December 2003-11392020 (i.e., second wave of infections). In column (1) of Table 5, we report the COVID-19 and 1140unemployment price effects when we relax these restrictions and estimate the average point estimates for 1141the full sample instead. The coefficient estimates on case and unemployment rates are consistent in 1142magnitude and are statistically significant at 1% level. 1143Next, we consider the sensitivity of our result against the presence of influential outliers. One criticism 1144against the use of repeat-sales samples has arisen from the literature of price indexes, which have been 1145shown to be more sensitive to outliers when relying on a repeat-sales sample (see, for example, Wallace 1146and Meese (1997)). Although we are not using the repeat-sales sample to construct a price index, we 1147investigate the potential impact of outliers. Excluding transactions that are more or less than three standard 1148deviations removed from the average change in sale prices between repeat sales reduces the number of 1149observations by 187 transactions for the first-wave sample and 158 transactions for the second-wave 1150sample. The coefficient estimates of interest remain stable and are statistically significant at the 1151conventional levels (see column (2) of Table 5). 1152Another potential concern with repeat sales is potential changes in unobservable property 1153characteristics between sales. While the PLUTOTM database indicates whether a home has been altered 1154from the perspective of the assessed value for taxation purposes, and we control for this alteration, it is 1155possible that some properties undergo changes that are not necessarily captured by the tax authority and 115611571158===== PAGE 31 (score 82, numeric-paren 69, table-words 5, var-words 3) =====1159116031 1161Table 5: Robustness Tests of Repeat-Sales COVID-19 Price Effects 1162(1) 1163(2) 1164(3) 1165(4) 1166(5) 1167(6) 1168(7) 1169(8) 1170Sample Restrictions 1171Alternative Pandemic Measures 1172All 1173No 1174Repeat 1175Out- 1176Sales 1177liers 1178> 3 yrs. 1179between 1180Sales 1181# of 1182Sales 1183< 4 1184NYC 1185Shutdown 118630-Day Lag 60-Day Lag Borough 1187Level 1188Rates 1189Panel A: First Wave of Infections (03/01-07/31, 2003-2020) 1190Cases per 100,000 (’000) 1191-0.594*** 1192-0.669* 1193-0.893 1194-0.703* 1195-0.719** 1196-0.631* 1197-0.422 1198-1.005* 1199(0.211) 1200(0.351) 1201(1.856) 1202(0.359) 1203(0.358) 1204(0.344) 1205(0.394) 1206(0.580) 1207Unemployment Rate (%) 1208 1209-0.053*** 1210-0.068*** 1211-0.074*** 1212-0.041 1213-0.039 1214-0.037 1215-0.043* 1216-0.232* 1217(0.018) 1218(0.018) 1219(0.026) 1220(0.026) 1221(0.024) 1222(0.024) 1223(0.023) 1224(0.118) 1225NYC Shutdown 1226 1227 1228 1229 12300.238 1231 1232 1233 1234 1235 1236 1237 1238(0.443) 1239 1240 1241 1242N 124336,051 124415,353 12456,174 124613,757 124715,540 124815,323 124915,094 125015,540 1251adj. R2 12520.145 12530.228 12540.101 12550.136 12560.146 12570.146 12580.146 12590.146 1260F 126119.7*** 126221.4*** 126311.2*** 126415.0*** 126514.1*** 126614.3*** 126713.1*** 126813.8*** 1269Panel B: Second Wave of Infections (08/01-12/31,2003-2020) 1270Cases per 100,000 (’000) 1271-0.594*** 1272-0.481** 1273-1.060** 1274-0.633*** 1275- 1276-0.622** 1277-0.500 1278-1.119*** 1279 1280(0.211) 1281(0.198) 1282(0.441) 1283(0.242) 1284- 1285(0.297) 1286(0.318) 1287(0.344) 1288Unemployment Rate (%) 1289-0.053*** 1290-0.070*** 1291-0.064*** 1292-0.064*** 1293- 1294-0.067*** 1295-0.061*** 1296-0.166 1297 1298(0.018) 1299(0.019) 1300(0.024) 1301(0.024) 1302- 1303(0.022) 1304(0.021) 1305(0.105) 1306N 130736,051 130815,575 13097,037 131014,014 1311- 131215,570 131315,410 131415,733 1315adj. R2 13160.145 13170.229 13180.107 13190.120 1320- 13210.134 13220.134 13230.131 1324F 132519.7*** 132621.5*** 13277.3*** 132810.1*** 1329- 133010.2*** 13319.5*** 13328.8*** 1333Home & Neigh. Charac. 1334Y 1335Y 1336Y 1337Y 1338Y 1339Y 1340Y 1341Y 1342Month-of-Sale FE 1343Y 1344Y 1345Y 1346Y 1347Y 1348Y 1349Y 1350Y 1351Month-of-Pre.-Sale FE 1352Y 1353Y 1354Y 1355Y 1356Y 1357Y 1358Y 1359Y 1360Notes: Standard errors are reported in parenthesis. Across all specifications, we cluster standard errors across 1361MODZCTAs. Panel A shows the results focused on the pandemic-induced effects during first wave of infections 1362between March and July for each year from 2003 through 2020, whereas Panel B shows results focused on the 1363second wave between August and December for each year from 2003 through 2020. Depending on the imposed 1364sample restrictions, we observe sales in 135 to 155 MODZCTAs anywhere from as early as January 2003 through 1365December 2020. Columns (1) through (4) show the COVID-19 cumulative case effect per 100,000 local 1366MODZCTA residents (measured in 1,000 cases) as well as the unemployment effect on the change in sale prices 1367between repeat transactions under varying sample restrictions. In column (5), we estimate the price effects of 1368cumulative Coronavirus infections and unemployment while controlling for the signing of the New York State 1369on PAUSE executive order. In columns (6) and (7), we report the 30-day and 60-day lagged effect of cumulative 1370MODZCTA infection and unemployment rates. Finally, in column (8) we show the effect of changes in borough, 1371rather than MODZCTA, cumulative infection and unemployment rates. Across all regressions, we control for 1372time-varying home and neighborhood characteristics (i.e., whether a home was altered between sales and the 1373share of investment-owned properties in a given neighborhood), month-of-sale and month-of-previous-sale fixed 1374effects. Statistical significance at the conventional 10%, 5%, and 1% thresholds is given by *, **, ***, 1375respectively. 1376 1377occurred after much of the activities in NYC had been restricted. If the shutdown of NYC, rather than the 1378intensity of the local outbreak, is the explanation for the estimated price effect, then the inclusion of this 137913801381===== PAGE 32 (score 8, numeric-paren 0, table-words 3, var-words 2) =====1382138332 1384control variable should render our estimated case effect on home sale prices insignificant. If, however, we 1385find that homes sold in neighborhoods with greater infection rates experience significantly larger price 1386discounts even when controlling for the shutdown executive order, our estimate provides evidence of the 1387contagion price effect (rather than the shutdown effect). Indeed, column (5) of Table 5 shows a statistically 1388and economically significant price discount for homes located in neighborhoods with higher Coronavirus 1389case rates, while the price effect of the shut-down is statistically insignificant.18 1390Combined, these robustness tests based on various sample restrictions produce convincing evidence 1391in support of our primary estimates of the pandemic-induced price effects and the proper identification 1392thereof. A final set of robustness analyses considers the timing of the evidenced pandemic-induced price 1393discount and sensitivity of our results against more aggregated measures of the intensity of the COVID-139419 pandemic and rise in unemployment. Similar to the sample restrictions, these tests produce estimates 1395that reiterate and tend to support our primary findings. The results reported in columns (6) and (7) of 1396Table 5, for example, show the estimated COVID-19 price discount with respect to 30- and 60-day lagged 1397case and unemployment rates. While these lagged effects are quantitatively consistent with the primary 1398results, the effects of 60-day lagged cumulative case rates are found to be statistically insignificant. Lastly, 1399we consider case and unemployment rates at the more aggregated borough, rather than MODZCTA, level. 1400Reassuringly, the estimates in column (8) of Table 5 show that our findings are generally robust against 1401the extrapolation of MODZCTA case and unemployment rates. 1402 14036 Discussion and Potential Limitations 1404What is the economic significance of these estimates? Focusing on our repeat-sales sample results, our 1405price estimate suggests that during the second wave of the pandemic the average sale price of a repeat-1406sale NYC home drops by close to $63,000 for every 1,000 novel Coronavirus infections added to the 1407cumulative case rate per 100,000 residents. Between August and December, 2020 (the end of our sample), 1408the change in MODZCTA case rates in NYC ranged anywhere from 100 to 7,100 and averaged 1,776 1409infections per 100,000 residents. Combining these statistics with our estimate suggests that the expected 1410price effects may range from as little as $6,000 on the lower end to nearly $450,000 in neighborhoods 1411with the largest increases in case rates and average around $100,000. A comparison of these figures to the 1412 141318 We note that the gradual reopening was largely concluded by August 2020 and therefore we cannot estimate the effect of the 1414NYC shut down during the second wave of the pandemic. 1415