1# RECORD 8d5f6b09f9058fc5a131a22d58a856d18ce849c72# TITLE: Information value of property description: A Machine learning approach3# YEAR: 2021.0 VENUE: Journal of Urban Economics DOI: 10.1016/j.jue.2020.1032994# AUTHORS: Lily Shen; Stephen L. Ross5# PAGES: 42 CANDIDATE TABLE PAGES (1-based): [38, 39, 40, 42]6# ABSTRACT: Abstract This paper employs machine learning to quantify the value of “soft” information contained in real estate property descriptions. Textual descriptions contain information that traditional hedonic attributes cannot capture. A one standard deviation increase in the uniqueness of a property based on this “soft” information leads to a 15% increase in property sale price in a hedonic price model and a 10% increase in a repeat sales price model. The effects in the hedonic model appear to arise through two channels: the unobserved quality of the housing unit, and the market power of the housing unit relative to competing properties. The effects in the repeat sales model appear to be driven entirely by the market power of the unit. Further, an annual hedonic price index ignoring our measure of unobserved quality overstates real estate prices by between 10% to 23% and mistimes the stabilization of housing prices following the Great Recession. Similar, but smaller effects, are observed for the repeat sales price index.789===== PAGE 1 (score 1, numeric-paren 0, table-words 0, var-words 1) =====1011Department of Economics Working Paper Series 12365 Fairfield Way, Unit 106313Storrs, CT 06269-106314Phone: (860) 486-302215Fax: (860) 486-446316http://www.econ.uconn.edu/17This working paper is indexed in RePEc, http://repec.org 18Information Value of Property Description: 19A Machine Learning Approach20by 21Lily Shen22Clemson University23Stephen Ross 24University of Connecticut25Working Paper 2019-20R 26December 2019, Revised September 2020272829===== PAGE 38 (score 92, numeric-paren 78, table-words 5, var-words 4) =====3031Table 3: Compare Covariates with and without Unique Measures32Dependent Variable: Ln(price)33Full Sample34Repeat Sale35(1)36(2)37(3)38(4)39MLS Area-Year Unique Score (standard)400.149***410.099***42(0.034)43(0.020)44Age45-0.013***46-0.012***47-0.014***48-0.014***49(0.001)50(0.001)51(0.002)52(0.002)53AgexAge540.0001***550.0001***560.0001***570.0001***58(0.0000)59(0.0000)60(0.0000)61(0.0000)62Bed63-0.047**64-0.042**65-0.01466-0.01367(0.021)68(0.020)69(0.021)70(0.021)71Bath720.247***730.245***740.320***750.315***76(0.018)77(0.016)78(0.035)79(0.033)80Ranch81-0.260***82-0.244***83-0.03684-0.03885(0.021)86(0.020)87(0.042)88(0.041)89Renovated900.253***910.225***920.359***930.345***94(0.032)95(0.027)96(0.042)97(0.038)98Sqft990.009***1000.009***101-0.001102-0.001103(0.001)104(0.001)105(0.001)106(0.001)107Large Lot1080.0121090.0041100.0661110.077112(0.028)113(0.025)114(0.070)115(0.073)116Pool1170.049***1180.051***1190.0261200.010121(0.018)122(0.016)123(0.056)124(0.057)125Feature1260.126***1270.090***1280.023129-0.004130(0.031)131(0.030)132(0.083)133(0.078)134Constant13510.338***1368.610***13712.246***1388.696***139(0.081)140(0.359)141(0.265)142(0.521)143Observations14440,85114540,8511467,3171477,317148R-squared1490.7601500.7691510.9561520.958153House Characteristics154Yes155Yes156Yes157Yes158LocationxYear FE159Yes160Yes161Yes162Yes163Property FE164No165No166Yes167Yes168Cluster169Area-Year170Area-Year171Area-Year172Area-Year173Note: The results displayed in columns (1) and (2) are using the full sample, and results displayed in174columns (3) and (4) are based on the repeat sale sample.175Columns (2) and (4) include the control for176property uniqueness. The unique score coefficients are standardized to show the effect of the increase in one177standard deviation. All models include MLS Area by Year fixed effects. Robust standard errors are clustered178at the MLS Area by Year level, shown in parentheses (*** p<0.01, ** p<0.05, * p<0.1).17936180181182===== PAGE 39 (score 34, numeric-paren 17, table-words 8, var-words 1) =====183184Table 4: House Uniqueness vs. Market Power185Full Sample186Repeat Sale187Dependent Variables: Ln(Price)188(1)189(2)190(3)191(4)192(5)193(6)194(7)195(8)196MLS Area-Year Unique Score (standard)1970.149***1980.099***199(0.034)200(0.020)201MLS Area Competitor Unique Score (standard)2020.125***2030.0692040.090***2050.084*206(0.032)207(0.071)208(0.020)209(0.047)210MLS Area Non-Competitor Unique Score (standard)2110.112***2120.0592130.068***2140.006215(0.028)216(0.061)217(0.021)218(0.048)219Observations22040,85122140,85122240,85122340,8512247,3172257,3172267,3172277,317228R-squared2290.7692300.7672310.7672320.7672330.9582340.9582350.9572360.957237House Characteristics238Yes239Yes240Yes241Yes242Yes243Yes244Yes245Yes246LocationxYear FE247Yes248Yes249Yes250Yes251Yes252Yes253Yes254Yes255Property FE256No257No258No259No260Yes261Yes262Yes263Yes264Cluster265Area-Year266Area-Year267Area-Year268Area-Year269Area-Year270Area-Year271Area-Year272Area-Year273Note: The dependent variable this table is Ln(sale price). Columns (1) and (5) repeat the estimates from the previous table for the hedonic and274repeat sales models respectively. Columns (2) and (6) present estimates using the new MLS area measure of uniqueness based on competing units,275and Columns (3) and (7) present estimates using uniqueness based on non-competing units. Finally, in Columns (4) and (8), we run a horse race276between the uniqueness measure based on competing units and based on non-competing units. All of the unique score coefficients are standardized277to show the effect of the increase in one standard deviation. All models include MLS area by year fixed effects. Robust standard errors are clustered278at the MLS Area by Year level, shown in parentheses (*** p<0.01, ** p<0.05, * p<0.1).27937280281282===== PAGE 40 (score 44, numeric-paren 18, table-words 12, var-words 2) =====283284Table 5: Are the Uniqueness Effects Sensitive to Alternative Models?285Panel A: Full Sample Analysis286Dependent Variable:287Ln(Price)288Price Spread289(1)290(2)291(3)292(4)293(5)294MLS Area-Year Unique Score (standard)2950.149***2960.149***2970.136***2980.122***299-0.761300(0.034)301(0.032)302(0.026)303(0.019)304(0.616)305Observations30640,85130740,85130840,85130940,85131040,851311R-squared3120.7693130.7843140.8113150.8863160.238317House Characteristics318Yes319Yes320Yes321Yes322Yes323Keywords324No325No326Yes327Yes328No329Transaction Characteristics330No331Yes332Yes333Yes334No335LocationxYear FE336Yes337Yes338Yes339Yes340Yes341Agent FE342No343No344No345Yes346No347Property FE348No349No350No351No352No353Cluster354Area-Year355Area-Year356Area-Year357Area-Year358Area-Year359Panel B: Repeat Sale Analysis360Dependent Variable:361Ln(Price)362Price Spread363(1)364(2)365(3)366(4)367(5)368MLS Area-Year Unique Score (standard)3690.099***3700.098***3710.080***3720.050***373-0.297374(0.020)375(0.020)376(0.015)377(0.018)378(0.592)379Observations3807,3173817,3173827,3173834,1173847,317385R-squared3860.9583870.9613880.9683890.9893900.684391House Characteristics392Yes393Yes394Yes395Yes396Yes397Keywords398No399No400Yes401Yes402No403Transaction Characteristics404No405Yes406Yes407Yes408No409LocationxYear FE410Yes411Yes412Yes413Yes414Yes415Agent FE416No417No418No419Yes420No421Property FE422Yes423Yes424Yes425Yes426Yes427Cluster428Area-Year429Area-Year430Area-Year431Area-Year432Area-Year433Note: Panel A shows hedonic estimates using the full sample, and Penal B displays repeat sale results based434on the repeat sales sample. The dependent variable in all specifications in Column (1)–(4) is Ln(sale price),435and the dependent variable in Column (5) is Listing-Sale Spread in thousand dollars. Column (1) presents436the standard hedonic model with just the uniqueness control added, Column (2) presents estimates for a437model that also adds transaction characteristics, Column (3) also adds the dummy variables for the presence438of specific keywords, and Column (4) adds agent fixed effects. Column (5) presents the model on the listing-439sales price spread. The unique score coefficients are standardized to show the effect of the increase in one440standard deviation. Robust standard errors are clustered at the Area-Year level, shown in parentheses (***441p<0.01, ** p<0.05, * p<0.1).44238443444445===== PAGE 41 (score 13, numeric-paren 0, table-words 4, var-words 5) =====446447Table 6: Summary Statistics of Hedonic Controls by Year448Year:4492010450201145120124522013453VARIABLES454Mean455Std. Dev.456Mean457Std. Dev.458Mean459Std. Dev.460Mean461Std. Dev.462MLS Area-Year Unique Score4630.754640.074650.784660.064670.784680.054690.784700.05471MLS Area-Year Unique Score (Mean Diff)4720.00134730.02694740.00344750.0262476-0.00034770.0262478-0.00114790.0242480Ln(Sale Price)48111.844821.3948311.854841.4448512.014861.3248712.344881.17489Age49042.6249130.3249242.8549330.2049445.0349530.0949645.3349730.02498Fireplace (d)4990.995001.055011.035021.065031.055041.055051.075060.99507Pool (d)5080.055090.215100.065110.235120.055130.225140.055150.23516Bed5173.615181.025193.625201.045213.615221.035233.635241.06525Ranch (d)5260.475270.505280.445290.505300.435310.505320.425330.49534Bath5352.645361.305372.705381.335392.725401.295412.755421.26543Renovated (d)5440.005450.015460.035470.185480.065490.245500.085510.27552Sqft (100)55320.9555411.6155522.5655612.9955722.9555813.7155923.4456013.58561Year:562201456320155642016565VARIABLES566Mean567Std. Dev.568Mean569Std. Dev.570Mean571Std. Dev.572MLS Area-Year Unique Score5730.795740.045750.785760.055770.785780.06579MLS Area-Year Unique Score (Mean Diff)580-0.00215810.0254582-0.00165830.0267584-0.00085850.0272586Ln(Sale Price)58712.455881.1258912.475901.1059112.495921.05593Age59446.8559531.3559647.6059732.0259848.5359931.94600Fireplace (d)6011.106021.066031.086041.066051.036061.06607Pool (d)6080.056090.226100.056110.226120.056130.21614Bed6153.616161.056173.596181.026193.606201.02621Ranch (d)6220.446230.506240.426250.496260.446270.50628Bath6292.736301.296312.716321.286332.686341.26635Renovated (d)6360.086370.276380.096390.286400.106410.30642Sqft (100)64323.7564414.5164523.5364613.7264723.0664813.86649Notes: This table displays the summary statistics of the control variables by transaction year. The means650for each variable are shown in separate rows, and means and standard deviation for each year are shown in651separate columns.65239653654655===== PAGE 42 (score 53, numeric-paren 45, table-words 4, var-words 0) =====656657Table 7: Coefficient of Year FE658Dependent Variable: Ln (Price)659Model:660Hedonic Index661Repeat Sale Index662Model 1663Model 2664Diff. M1- M2665Model 1666Model 2667Diff. M1- M2668(1)669(2)670(3)671(4)672(5)673(6)6742011 (d)675-0.057*676-0.177***6770.120678-0.012679-0.068**6800.056681(0.030)682(0.026)683(0.033)684(0.028)685P(Chi2)6860.0006870.0096882012 (d)6890.068690-0.060**6910.0086920.063*6930.0206940.043695(0.044)696(0.025)697(0.035)698(0.033)699P(Chi2)7000.0047010.0207022013 (d)7030.362***7040.220***7050.1427060.287***7070.250***7080.037709(0.054)710(0.029)711(0.047)712(0.039)713P(Chi2)7140.0037150.0577162014 (d)7170.463***7180.304***7190.1597200.360***7210.323***7220.037723(0.066)724(0.034)725(0.046)726(0.037)727P(Chi2)7280.0067290.0667302015 (d)7310.494***7320.352***7330.1427340.484***7350.453***7360.031737(0.080)738(0.057)739(0.049)740(0.048)741P(Chi2)7420.0197430.1307442016 (d)7450.553***7460.444***7470.1097480.604***7490.575***7500.029751(0.091)752(0.065)753(0.072)754(0.063)755P(Chi2)7560.1157570.206758Control Unique759No760Yes761No762Yes763R-squared7640.5097650.5717660.9477670.949768Notes: This table displays the year indicator coefficient estimates from the indexing models first for the769hedonic models and then for the repeat sale models. The estimates for each year are presented in separate770rows, and each specification is shown in a set of three columns: the first presenting estimates for a model that771does not control for the uniqueness, the second from a model that includes uniqueness, and the third column772presenting the differences between these two model estimates and the corresponding P–values.773Robust774standard errors are clustered at the Zip Code level, shown in parentheses (*** p<0.01, ** p<0.05, * p<0.1).77540776