1# RECORD a842ed7c689fc0985af5ca73c3f97f4279427db32# TITLE: Influence of a special tax-cooling measure on housing prices in Taiwan: a hedonic pricing model with consideration of the spillover effect3# YEAR: 2025.0 VENUE: Japanese Economic Review DOI: 10.1007/s42973-025-00217-54# AUTHORS: Yiwen Yang5# PAGES: 44 CANDIDATE TABLE PAGES (1-based): [18, 24, 26, 27, 30, 32, 33, 34, 35, 36]6# ABSTRACT: 7 To limit excessive real estate speculation and curb rising real estate property prices in Taiwan, in June 2011, the government imposed a transaction tax (i.e., the Specifically Selected Goods and Services Tax; SSGST) on short-term real estate transactions. The findings in the literature on the effects of transaction taxation on housing prices are inconsistent. To investigate the influence of the SSGST on real estate market prices in Taiwan, we adopted the hedonic pricing model, which can be used to determine the effects of houses’ intrinsic characteristics on their prices, and used merged administrative data. The results demonstrate that the special ad valorem tax was effective in reducing housing prices in short-term transactions, with larger declines observed at higher tax rates. Furthermore, its influence gradually extended to the broader housing market. However, housing prices in municipalities under the control of the central government were less sensitive to the SSGST policy.8910===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) =====1112Vol.:(0123456789)13The Japanese Economic Review14https://doi.org/10.1007/s42973-025-00217-515ARTICLE16Influence of a special tax‑cooling measure 17on housing prices in Taiwan: a hedonic pricing model 18with consideration of the spillover effect19Yiwen Yang1 20Received: 3 March 2024 / Revised: 4 May 2025 / Accepted: 12 June 2025 21© The Author(s) 202522Abstract23To limit excessive real estate speculation and curb rising real estate property prices 24in Taiwan, in June 2011, the government imposed a transaction tax (i.e., the Specifi-25cally Selected Goods and Services Tax; SSGST) on short-term real estate transac-26tions. The findings in the literature on the effects of transaction taxation on hous-27ing prices are inconsistent. To investigate the influence of the SSGST on real estate 28market prices in Taiwan, we adopted the hedonic pricing model, which can be used 29to determine the effects of houses’ intrinsic characteristics on their prices, and used 30merged administrative data. The results demonstrate that the special ad valorem 31tax was effective in reducing housing prices in short-term transactions, with larger 32declines observed at higher tax rates. Furthermore, its influence gradually extended 33to the broader housing market. However, housing prices in municipalities under the 34control of the central government were less sensitive to the SSGST policy.35Keywords Hedonic pricing model · Specifically Selected Goods and Services Tax · 36Real estate transaction tax · Price bubble · Spillover effect37JEL Classification H2 · H338 * Yiwen Yang 39 40ywyang@ntnu.edu.tw411 42Department of Civic Education and Leadership, College of Education, National Taiwan Normal 43University, Taipei City, Taiwan444546===== PAGE 18 (score 8, numeric-paren 0, table-words 2, var-words 2) =====4748 49The Japanese Economic Review50Table 1 (continued)51Variable52Definition53Obs54Mean55Std. Dev Min56Max57 Loan-to-value (LTV)58The loan to value at the time of the housing transaction59416,2636069.598610.4386268.736370.1764 Consumer Price Index (CPI) The consumer price index at the time of the housing transaction65416,2636698.169670.8376895.096999.7970 Quarter71The quarter of the year in which the housing unit was transacted72416,263732.664741.12275176477 Days_73078A binary variable equal to 1 if the transaction is classified as short term, and 0 otherwise7930,659800.919810.27382083184Spillover effect model85 N_tran_1 month86The total number of short-term housing transactions measured over a one-month window8728,168881022.121 297.778891901,49191 N_tran_2 months92The total number of short-term housing transactions measured over a two-month window9328,167942016.885 568.320952962,91097 N_tran_3 months98The total number of short-term housing transactions measured over a three-month 99window10028,1661012984.587 826.50910241034,067104 Tax proportion105The proportion of monthly short-term housing transactions that were subject to the 106SSGST107401080.1411090.21411001111112 LR_D_N_tran113The month-over-month change in the number of long-term housing transactions114281150.21411681.591117-306118130119 N_tran120The number of short-term housing transactions within a given month12140122704.200123479.18812411251,491126Note: Only observations included in the regression hereinafter are counted127128129===== PAGE 19 (score 10, numeric-paren 0, table-words 2, var-words 6) =====130131The Japanese Economic Review 132between the SSGST dummy and the effective tax rate (i.e., tax dummy × effective tax 133rate), as shown in Eq. (6). The individual terms, the tax dummy and the effective tax 134rate, are excluded from the model due to their high correlation with the interaction 135term. This specification allows us to examine whether higher tax burdens are associ-136ated with more pronounced reductions in housing prices.137To control for relevant covariates, in addition to housing characteristics such as 138locational and structural attributes, we also incorporate macroeconomic and macro-139prudential indicators, including LTV, CPI, GDP growth, and the unemployment rate, 140into the regression. To account for seasonality, we follow Osland (2013) and include 141quarterly dummy variables. Temporal and regional heterogeneity are addressed by 142including year fixed effects and county fixed effects (with “county” broadly refer-143ring to both cities and counties), respectively. The log-linear specification enables 144straightforward interpretation of the estimated coefficients as approximate percent-145age changes in housing prices. The hedonic pricing models are specified as follows.146In both Eqs. (5) and (6), Pit is the price of house i at the time (day) t of trans-147action. The vector Z = Z(z1, z2, ⋯, zj) denotes the housing characteristics, and 148G = G(g1, g2, ⋯, gk) represents government macroprudential policies and macro-149economic variables. q captures quarter dummies, while 휃y and 훿r are year and county 150fixed effects, respectively. The error term is denoted by 휀it . Additionally, we esti-151mate an alternative specification in which separate year and county fixed effects are 152replaced with county-by-year interaction dummies. This allows us to account for 153county-specific characteristics that vary over time, capturing unobserved time-var-154ying heterogeneity across counties. We replace, rather than add to, the original fixed 155effects to avoid multicollinearity and ensure that all county-level temporal variation 156is properly absorbed.157Osland (2013) demonstrated that unobserved attributes play a critical role in 158hedonic house price models. Because houses are extremely heterogeneous, includ-159ing all relevant attributes in such models would be impossible. To address this, 160Osland (2013) proposed panel data models as a means of removing the influence of 161unobserved housing attributes on housing prices. Although the merged data set com-162prises panel data, it is unbalanced. The proportion of houses that were transacted 163only once (93%) is high.12 To compromise on a few aforementioned data deficien-164cies and more precisely determine the influence of the SSGST on housing prices, 165we divide our data set into two categories, that is, houses transacted only once and 166houses transacted more than once, and employ different econometric models. It is 167(5)168ln(Pit169) = a + b ⋅tax dummyit +170∑171jcjzjit +172∑173kdkgkit +174∑175sesqsi + 휃y + 훿r + 휀it176(6)177ln(Pit178) = a + b ⋅(tax dummy × effective tax rate)it +179∑180jcjzjit +181∑182kdkgkit183+184∑185sesqsi + 휃y + 훿r + 휀it18612 The number of houses transacted only once is 386,978. The total number of transacted houses is 187416,263. Therefore, the proportion of houses transacted only once is 386,978/416,263 = 0.9296.188189190===== PAGE 24 (score 165, numeric-paren 135, table-words 4, var-words 8) =====191192 193The Japanese Economic Review194Table 2 Effects of the SSGST on Housing Prices -Transacted Only Once195Dependent variable: ln (price)196OLS: tax197OLS: tax dummy*effective tax rate198(1)199(2)200(3)201(4)202Tax203– 0.103***204– 0.108***205(0.033)206(0.033)207Tax dummy*effective tax rate208– 1.000***209– 1.039***210(0.269)211(0.270)212Municipality2130.236***2140.184**2150.236***2160.184**217(0.010)218(0.089)219(0.010)220(0.089)221Tax dummy*municipality2220.07262230.0816*2240.0870**2250.0961**226(0.045)227(0.045)228(0.042)229(0.042)230ln (house area)2310.904***2320.905***2330.904***2340.905***235(0.003)236(0.003)237(0.003)238(0.003)239Hall2400.0798***2410.0800***2420.0798***2430.0800***244(0.002)245(0.002)246(0.002)247(0.002)248wall249– 0.210***250– 0.210***251– 0.210***252– 0.211***253(0.007)254(0.007)255(0.007)256(0.007)257Age258– 0.00631***259– 0.00630***260– 0.00631***261– 0.00630***262(0.000)263(0.000)264(0.000)265(0.000)266Residential267– 0.156***268– 0.156***269– 0.156***270– 0.156***271(0.002)272(0.002)273(0.002)274(0.002)275Steel reinforced concrete2760.0978***2770.0967***2780.0978***2790.0967***280(0.012)281(0.012)282(0.012)283(0.012)284Apartment285– 0.209***286– 0.209***287– 0.209***288– 0.209***289(0.003)290(0.003)291(0.003)292(0.003)293Econ growth2940.0201***2950.0196***2960.0201***2970.0196***298(0.003)299(0.003)300(0.003)301(0.003)302CPI303– 0.00382304– 0.00348305– 0.00382306– 0.00347307(0.003)308(0.003)309(0.003)310(0.003)311Loan-to-value (LTV)3120.00733*3130.00807*3140.00733*3150.00807*316(0.004)317(0.004)318(0.004)319(0.004)320Unemployment rate3210.02193220.02273230.02183240.0226325(0.022)326(0.022)327(0.022)328(0.022)329Quarter 23300.0206***3310.0200***3320.0206***3330.0200***334(0.007)335(0.007)336(0.007)337(0.007)338Quarter 33390.0148***3400.0143**3410.0148***3420.0143**343(0.006)344(0.006)345(0.006)346(0.006)347Quarter 43480.0475***3490.0468***3500.0475***3510.0468***352(0.006)353(0.006)354(0.006)355(0.006)356Constant35710.63***35810.65***35910.63***36010.65***361(0.429)362(0.430)363(0.429)364(0.430)365County fixed effects366Yes367No368Yes369No370Year fixed effects371Yes372No373Yes374No375County-year paired fixed effects376No377Yes378No379Yes380N381386,978382386,978383386,978384386,978385Adj. R-squared3860.6003870.6003880.6003890.600390391392===== PAGE 25 (score 14, numeric-paren 0, table-words 1, var-words 12) =====393394The Japanese Economic Review 395that the housing prices of municipalities under the control of a central govern-396ment are 18.4% ~ 23.6% higher than those of nonmunicipalities. The coefficients 397on the interaction term between the SSGST tax dummy and the municipality indi-398cator (tax dummy × municipality) are positive across all specifications, ranging 399from 0.0726 to 0.0961 in Models (1) through (4). While the estimate in Model 400(1) is not statistically significant, the interaction terms in Models (2) through (4) 401are statistically significant at the 10% level. These results consistently suggest 402that the price-reducing effect of the SSGST is weaker in municipalities governed 403by the central government, indicating that housing prices in these areas are less 404responsive to the policy. This observed heterogeneity motivates further investi-405gation. In Sect. 6.2, we address this by estimating separate models for munici-406palities and nonmunicipalities to more clearly assess regional variation in respon-407siveness to the SSGST.408For the variables related to the structural attributes of houses, we analyze the 409total areas and the number of halls. The coefficients are all positive and signifi-410cant at the 1% level in all models, indicating that housing price increases with 411the square footage and number of halls in a house. Furthermore, the coefficients 412for wall and age are negative and significant at the 1% level in all models. This 413indicates that people generally dislike partition walls. Compared with those with-414out, houses with partition walls have lower prices. One possible reason for this 415is that people frequently renovate houses they have bought, and partition walls 416may force buyers to pay an additional cost to remove the walls. The estimated 417results for the age of houses indicate that housing prices decrease as the age of 418the houses increases. Furthermore, residential houses have lower prices compared 419with houses used for other purposes. Moreover, the results indicate that apartment-420type housing units are priced significantly lower than other building types. In con-421trast, the presence of reinforced concrete construction is positively associated with 422housing prices, highlighting the importance of building type and exterior quality 423in price determination.424Regarding macroprudential indicators, both economic growth and the LTV are 425positively and statistically significantly associated with housing prices, while the 426coefficients on the CPI and the unemployment rate are not statistically significant. 427Seasonal variation is also observed, with housing prices significantly higher in the 428second, third, and fourth quarters compared to the first quarter.4296.1.2 Houses transacted more than once430In this section, we begin by focusing on short-term transactions to examine the 431impact of the SSGST on housing prices. We then broaden the analysis to include all 432housing units that were transacted more than once, allowing us to assess the overall 433Table 2 (continued)434Note: 1. Standard errors in parentheses; 2. * p < 0.10, ** p < 0.05, *** p < 0.01; 3. Due to high collin-435earity between the tax dummy variable and its interaction with the effective tax rate, the tax dummy is 436excluded from Models (3) and (4) to mitigate multicollinearity concerns437438439===== PAGE 26 (score 102, numeric-paren 84, table-words 2, var-words 4) =====440441 442The Japanese Economic Review443Table 3 Effects of the SSGST on Housing Prices- Transacted More Than Once444Short-term transaction445Whole transaction446Tax dummy447Tax dummy*effective tax rate448Tax dummy449Tax dummy*effective tax rate450(1)451(2)452(3)453(4)454(5)455(6)456(7)457(8)458Dependent variable: ln (price)459Fixed460Effect461Fixed Effect462Fixed463Effect464Fixed Effect465Fixed Effect466Fixed467Effect468Fixed Effect469Fixed Effect470Tax471– 0.293***472– 0.294***473– 0.261***474– 0.267***475(0.068)476(0.071)477(0.062)478(0.066)479Tax dummy*effective tax rate480– 1.989***481– 1.940***482– 1.740***483– 1.735***484(0.588)485(0.608)486(0.523)487(0.560)488Municipality489 .490 .491 .492 .493 .494. 495 .496 .497Tax dummy*municipality4980.146*4990.147*5000.08625010.08015020.117*5030.1195040.06125050.0560506(0.078)507(0.081)508(0.081)509(0.083)510(0.071)511(0.076)512(0.072)513(0.078)514Days 7305150.232***5160.206***5170.232***5180.206***519(0.033)520(0.034)521(0.033)522(0.034)523ln (house area)524 Hall525– 0.00950526– 0.00931527– 0.00830528– 0.00824529– 0.00517530– 0.00443531– 0.00408532– 0.00347533(0.012)534(0.012)535(0.012)536(0.012)537(0.012)538(0.012)539(0.012)540(0.012)541 Wall5420.146***5430.120***5440.146***5450.121***5460.127***5470.101**5480.128***5490.102**550(0.043)551(0.044)552(0.043)553(0.044)554(0.041)555(0.041)556(0.041)557(0.041)558 Age559– 0.003535600.000662561– 0.002735620.00119563– 0.00685*564– 0.00368565– 0.00624*566– 0.00332567(0.007)568(0.006)569(0.006)570(0.006)571(0.004)572(0.004)573(0.003)574(0.003)575 Residential5760.2275770.277**5780.218*5790.266***5800.229**5810.264***5820.243**5830.265***584(0.144)585(0.114)586(0.125)587(0.100)588(0.105)589(0.099)590(0.104)591(0.096)592593594===== PAGE 27 (score 128, numeric-paren 106, table-words 1, var-words 6) =====595596The Japanese Economic Review 597Table 3 (continued)598Short-term transaction599Whole transaction600Tax dummy601Tax dummy*effective tax rate602Tax dummy603Tax dummy*effective tax rate604(1)605(2)606(3)607(4)608(5)609(6)610(7)611(8)612Steel reinforced concrete613 Apartment6140.02776150.03616160.02926170.03746180.04596190.04886200.04796210.0513622(0.058)623(0.058)624(0.059)625(0.059)626(0.056)627(0.058)628(0.056)629(0.058)630 Econ growth6310.0219*6320.0228*6330.0214*6340.0222*6350.0244**6360.0237**6370.0239**6380.0230**639(0.012)640(0.012)641(0.012)642(0.012)643(0.012)644(0.012)645(0.012)646(0.012)647 CPI6480.002306490.001536500.001726510.001056520.001996530.001136540.001206550.000158656(0.012)657(0.012)658(0.012)659(0.012)660(0.011)661(0.011)662(0.012)663(0.011)664 Loan-to-value (LTV)6650.02766660.02716670.03016680.02946690.02196700.01886710.02426720.0209673(0.019)674(0.019)675(0.019)676(0.019)677(0.019)678(0.019)679(0.019)680(0.019)681 Unemployment682– 0.195*683– 0.171684– 0.189*685– 0.164686– 0.175687– 0.134688– 0.170689– 0.127690(0.110)691(0.111)692(0.112)693(0.113)694(0.107)695(0.108)696(0.108)697(0.110)698 Quarter 2699– 0.0156700– 0.0126701– 0.0156702– 0.0124703– 0.0106704– 0.00312705– 0.0104706– 0.00231707(0.031)708(0.031)709(0.031)710(0.031)711(0.030)712(0.030)713(0.030)714(0.030)715 Quarter 37160.101***7170.0976***7180.100***7190.0967***7200.104***7210.102***7220.103***7230.102***724(0.025)725(0.025)726(0.025)727(0.025)728(0.024)729(0.024)730(0.025)731(0.024)732Quarter 47330.131***7340.131***7350.130***7360.131***7370.133***7380.139***7390.133***7400.140***741(0.027)742(0.027)743(0.028)744(0.027)745(0.026)746(0.026)747(0.027)748(0.026)749 Constant75013.48***75113.96***75213.32***75313.81***75413.66***75514.33***75613.53***75714.24***758(1.927)759(1.901)760(1.936)761(1.911)762(1.903)763(1.876)764(1.914)765(1.888)766County fixed effects767Yes768No769Yes770No771Yes772No773Yes774No775Year fixed effects776Yes777No778Yes779No780Yes781No782Yes783No784County-year paired fixed effects785No786Yes787No788Yes789No790Yes791No792Yes793794795===== PAGE 28 (score 9, numeric-paren 0, table-words 3, var-words 2) =====796797 798The Japanese Economic Review799Table 3 (continued)800Short-term transaction801Whole transaction802Tax dummy803Tax dummy*effective tax rate804Tax dummy805Tax dummy*effective tax rate806(1)807(2)808(3)809(4)810(5)811(6)812(7)813(8)814N81526,79481626,79481726,79481826,79481929,28582029,28582129,28582229,285823Adj. R-squared8240.4028250.4448260.3948270.4378280.4138290.4538300.4068310.446832Note: 1. Standard errors in parentheses; 2. * p < 0.10, ** p < 0.05, *** p < 0.01; 3. Due to high collinearity between the tax dummy variable and its interaction with the 833effective tax rate, the tax dummy is excluded from Models (3), (4), (7) and (8) to mitigate multicollinearity concerns834835836===== PAGE 30 (score 42, numeric-paren 32, table-words 1, var-words 4) =====837838 839The Japanese Economic Review840Table 4 Effects of the SSGST on Housing Prices in Municipalities and Nonmunicipalities841Municipality842Non-municipality843Short-term transaction844Tax dummy845Tax dummy*effective tax rate846Tax dummy847Tax dummy*effective 848tax rate849Tax850– 0.153***851– 0.152***852– 0.263***853– 0.266***854(0.039)855(0.039)856(0.065)857(0.068)858Tax dummy*effective tax rate859– 1.131***860– 1.121***861– 2.069***862– 2.096***863(0.306)864(0.311)865(0.593)866(0.632)867Housing Characterisitcs868Yes869Yes870Yes871Yes872Yes873Yes874Yes875Yes876Macro877Yes878Yes879Yes880Yes881Yes882Yes883Yes884Yes885Quarter dummies886Yes887Yes888Yes889Yes890Yes891Yes892Yes893Yes894Year fixed effects895Yes896No897Yes898No899Yes900No901Yes902No903County fixed effects904Yes905No906Yes907No908Yes909No910Yes911No912County-year paired fixed effects913No914Yes915No916Yes917No918Yes919No920Yes921N92220,69492320,69492420,69492520,6949266100927610092861009296100930Adjusted R squared9310.3819320.4069330.3779340.4019350.5069360.5629370.5019380.556939Whole transaction940Tax dummy941Tax dummy*effective tax rate942Tax dummy943Tax dummy*effective 944tax rate945Tax946– 0.148***947– 0.152***948– 0.240***949– 0.250***950(0.037)951(0.037)952(0.062)953(0.065)954Tax dummy*effective tax rate955– 1.083***956– 1.113***957– 1.882***958– 1.984***959(0.290)960(0.295)961(0.555)962(0.593)963Housing Characterisitcs964Yes965Yes966Yes967Yes968Yes969Yes970Yes971Yes972973974===== PAGE 31 (score 6, numeric-paren 0, table-words 1, var-words 3) =====975976The Japanese Economic Review 977Table 4 (continued)978Whole transaction979Tax dummy980Tax dummy*effective tax rate981Tax dummy982Tax dummy*effective 983tax rate984Macro985Yes986Yes987Yes988Yes989Yes990Yes991Yes992Yes993Quarter dummies994Yes995Yes996Yes997Yes998Yes999Yes1000Yes1001Yes1002Year fixed effects1003Yes1004No1005Yes1006No1007Yes1008No1009Yes1010No1011County fixed effects1012Yes1013No1014Yes1015No1016Yes1017No1018Yes1019No1020County-year paired fixed effects1021No1022Yes1023No1024Yes1025No1026Yes1027No1028Yes1029N103022,502103122,502103222,502103322,502103467831035678310366783103767831038Adjusted R squared10390.39310400.41610410.38810420.41110430.50810440.56210450.50310460.5581047Note: Standard errors in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01104810491050===== PAGE 32 (score 33, numeric-paren 16, table-words 3, var-words 3) =====10511052 1053The Japanese Economic Review1054Table 5 Propensity Score Matching Results for Housing Units Transacted Only Once and More Than 1055Once Under Different Tax Specifications1056Notes: 1. This table reports average treatment effects on the treated (ATT) estimated via propensity score 1057matching using 1-to-1 nearest neighbor matching with a caliper of 0.05. The treated group includes 1058housing units subject to each respective SSGST tax rate. Covariate balance diagnostics (available upon 1059request) confirm improved balance after matching. 2. *, **, and *** indicate significance at the 10%, 5%, 1060and 1% levels, respectively1061Transacted only once1062Tax Variable Covariates Included1063Treated Mean Control Mean ATT 1064(Treated – 1065Control)1066S.E1067T-stat1068Tax1069County and year dum-1070mies107115.187107215.2631073– 0.07610740.042 – 1.81*1075Tax1076County-year interaction 1077dummies107815.187107915.2601080– 0.07210810.042 – 1.7*1082Tax 10%1083County and year dum-1084mies108515.183108615.16710870.01510880.052 – 0.291089Tax 10%1090County-year interaction 1091dummies109215.183109315.18110940.00110950.052 0.031096Tax 15%1097County and year dum-1098mies109915.195110015.3461101– 0.15111020.074 – 2.04**1103Tax 15%1104County-year interaction 1105dummies110615.195110715.3601108– 0.16511090.074 – 2.22**1110Transacted more than once: Short-term transaction1111 Tax1112County and year dum-1113mies111415.195111515.3601116– 0.16511170.074 – 2.22**1118 Tax1119County-year interaction 1120dummies112114.943112215.1081123– 0.16511240.053 – 3.1***1125 Tax 10%1126County and year dum-1127mies112814.997112915.1721130– 0.17511310.064 – 2.71***1132 Tax 10%1133County-year interaction 1134dummies113514.997113615.1491137– 0.15211380.064 – 2.37**1139 Tax 15%1140County and year dum-1141mies114214.884114315.2691144– 0.38411450.080 – 4.82***1146 Tax 15%1147County-year interaction 1148dummies114914.884115015.1821151– 0.29811520.075 – 3.97***1153Transacted more than once: Whole transaction1154 Tax1155County and year dum-1156mies115714.943115815.1501159– 0.20711600.052 – 3.98***1161 Tax1162County-year interaction 1163dummies116414.943116515.2231166– 0.28011670.052 – 5.41***1168 Tax 10%1169County and year dum-1170mies117114.997117215.1851173– 0.18811740.063 – 2.97***1175 Tax 10%1176County-year interaction 1177dummies117814.999117915.2081180– 0.20911810.066 – 3.18***1182 Tax 15%1183County and year dum-1184mies118514.884118615.1891187– 0.30411880.077 – 3.96***1189 Tax 15%1190County-year interaction 1191dummies119214.884119315.1571194– 0.27311950.079 – 3.46***119611971198===== PAGE 33 (score 30, numeric-paren 15, table-words 2, var-words 2) =====11991200The Japanese Economic Review 1201Table 6 Propensity Score Matching Results Separately for Municipalities and Nonmunicipalities1202Tax Variable Covariates Included1203Treated Mean Control Mean ATT 1204(Treated–1205Control)1206S.E.1207T-stat1208Municipalities1209 Short-term transaction1210 tax1211County and year dum-1212mies121315.098121415.2781215– 0.18012160.065 – 2.75***1217 tax1218County-year interaction 1219dummies122015.098122115.2281222– 0.13012230.063 – 2.05**1224 tax 10%1225County and year dum-1226mies122715.108122815.2951229– 0.18812300.082 – 2.3**1231 tax 10%1232County-year interaction 1233dummies123415.108123515.2781236– 0.17012370.083 – 2.06**1238 tax 15%1239County and year dum-1240mies124115.087124215.2641243– 0.17812440.102 – 1.74*1245 tax 15%1246County-year interaction 1247dummies124815.098124915.2371250– 0.13912510.105 – 1.321252 Whole transaction1253 Tax1254County and year dum-1255mies125615.098125715.2191258– 0.12112590.069 – 1.75*1260 Tax1261County-year interaction 1262dummies126315.098126415.2851265– 0.18712660.063 – 2.97***1267 Tax 10%1268County and year dum-1269mies127015.108127115.2581272– 0.15012730.096 – 1.571274 Tax 10%1275County-year interaction 1276dummies127715.108127815.2281279– 0.12012800.078 – 1.531281 Tax 15%1282County and year dum-1283mies128415.087128515.1841286– 0.09812870.104 – 0.941288 Tax 15%1289County-year interaction 1290dummies129115.098129215.3121293-0.21412940.101 -2.13**1295Nonmunicipalities1296 Short-term transaction1297 Tax1298County and year dum-1299mies130014.670130114.9611302– 0.29013030.079 – 3.69***1304 Tax1305County-year interaction 1306dummies130714.670130814.9721309– 0.30213100.074 – 4.06***1311 Tax 10%1312County and year dum-1313mies131414.783131515.0591316– 0.27713170.094 – 2.94***1318 Tax 10%1319County-year interaction 1320dummies132114.783132215.0761323– 0.29313240.094 – 3.11***1325 Tax 15%1326County and year dum-1327mies132814.563132914.9371330– 0.37413310.122 – 3.08***1332 Tax 15%1333County-year interaction 1334dummies133514.563133614.8801337– 0.31613380.116 – 2.73***1339 Whole transaction1340 Tax1341County and year dum-1342mies134314.670134415.0591345– 0.38913460.077 – 5.03***134713481349===== PAGE 34 (score 19, numeric-paren 5, table-words 4, var-words 4) =====13501351 1352The Japanese Economic Review1353The coefficients on the interaction between the tax dummy and the municipality 1354indicator are positive and statistically significant at the 10% level in specifications 1355that include the tax dummy, except in the whole-market models with county-year 1356paired fixed effects. In contrast, the interaction terms are not statistically significant 1357in specifications that use the interaction between the tax dummy and the effective 1358tax rate. These findings suggest that the impact of the SSGST on housing prices 1359may differ between municipalities governed by the central government and nonmu-1360nicipalities. Accordingly, we further compare the effects of the SSGST on housing 1361prices between these two groups.13626.2 Heterogeneity between municipalities under the control of a central 1363government and nonmunicipalities1364To explore whether the impact of the SSGST differs across administrative regions, 1365we estimate panel data models separately for municipalities governed by the central 1366government and for nonmunicipalities. Based on the results of the Hausman test, we 1367adopt fixed effects specifications in both the short-term and full-sample transaction 1368models for houses transacted more than once. The results are presented in Table 4.1369In the short-term transaction models, the implementation of the SSGST is associ-1370ated with a housing price decrease of approximately 15.2%–15.3% in municipalities 1371under the control of the central government and 26.3%–26.6% in nonmunicipalities. 1372In the whole transaction models, which include both short- and long-term transac-1373tions, the SSGST corresponds to a price decline of approximately 14.8%–15.2% in 1374Notes: 1. This table reports average treatment effects on the treated (ATT) estimated via propensity score 1375matching using 1-to-1 nearest neighbor matching with a caliper of 0.05. The treated group includes 1376housing units subject to each respective SSGST tax rate. Covariate balance diagnostics (available upon 1377request) confirm improved balance after matching. 2. *, **, and *** indicate significance at the 10%, 5%, 1378and 1% levels, respectively1379Table 6 (continued)1380Tax Variable Covariates Included1381Treated Mean Control Mean ATT 1382(Treated–1383Control)1384S.E.1385T-stat1386 Tax1387County-year interaction 1388dummies138914.670139015.0511391– 0.38113920.077 – 4.97***1393 Tax 10%1394County and year dum-1395mies139614.783139714.9961398– 0.21413990.098 – 2.18**1400 Tax 10%1401County-year interaction 1402dummies140314.783140415.0421405– 0.25914060.100 – 2.61***1407 Tax 15%1408County and year dum-1409mies141014.563141114.8911412– 0.32814130.121 – 2.71***1414 Tax 15%1415County-year interaction 1416dummies141714.563141814.9111419– 0.34714200.117 – 2.96***142114221423===== PAGE 35 (score 59, numeric-paren 46, table-words 2, var-words 3) =====14241425The Japanese Economic Review 1426Table 7 Effect of the Number of Transactions on Housing Prices1427Note: Standard errors in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.011428Holding period1429Short-term transactions (within 2 years, including units transacted only once and more than once)1430Tax dummy1431Tax dummy*effective 1432tax rate1433Tax dummy1434Tax dummy*effective tax rate1435Tax dummy1436Tax dummy*effective tax rate1437FE1438FE1439FE1440FE1441FE1442FE1443FE1444FE1445FE1446FE1447FE1448FE1449N_tran_1 month14500.000050914510.000050214520.0000533*14530.0000541*1454(0.000)1455(0.000)1456(0.000)1457(0.000)1458N_tran_2 months14590.0000535**14600.0000497**14610.0000553**14620.0000521**1463(0.000)1464(0.000)1465(0.000)1466(0.000)1467N_tran_3 months14680.0000491**14690.0000484***14700.0000512***14710.0000505***1472(0.000)1473(0.000)1474(0.000)1475(0.000)1476Tax1477-0.291***1478-0.292***1479-0.291***1480-0.292***1481-0.290***1482-0.292***1483(0.068)1484(0.071)1485(0.068)1486(0.071)1487(0.068)1488(0.070)1489tax dummy*effective 1490tax rate1491-1.975***1492-1.923***1493-1.994***1494-1.945***1495-1.992***1496-1.944***1497(0.585)1498(0.604)1499(0.584)1500(0.605)1501(0.582)1502(0.602)1503Housing Character-1504isitcs1505Yes1506Yes1507Yes1508Yes1509Yes1510Yes1511Yes1512Yes1513Yes1514Yes1515Yes1516Yes1517Macro1518Yes1519Yes1520Yes1521Yes1522Yes1523Yes1524Yes1525Yes1526Yes1527Yes1528Yes1529Yes1530Quarter dummies1531Yes1532Yes1533Yes1534Yes1535Yes1536Yes1537Yes1538Yes1539Yes1540Yes1541Yes1542Yes1543Year fixed effects1544Yes1545No1546Yes1547No1548Yes1549No1550Yes1551No1552Yes1553No1554Yes1555No1556County fixed effects Yes1557No1558Yes1559No1560Yes1561No1562Yes1563No1564Yes1565No1566Yes1567No1568County-year paired 1569fixed effects1570No1571Yes1572No1573Yes1574No1575Yes1576No1577Yes1578No1579Yes1580No1581Yes1582N158328,168158428,168158528,168158628,168158728,167158828,167158928,167159028,167159128,166159228,166159328,166159428,1661595Adj.R squared15960.40315970.44515980.39515990.43816000.40416010.44716020.39716030.43916040.40616050.44816060.39916070.441160816091610===== PAGE 36 (score 24, numeric-paren 7, table-words 5, var-words 5) =====16111612 1613The Japanese Economic Review1614municipalities and 24.0%–25.0% in nonmunicipalities. While the policy exerts a 1615price-depressing effect in both regions, the estimated magnitudes are consistently 1616larger in nonmunicipalities. These findings suggest that housing prices in munici-1617palities under the control of the central government are less responsive to the tax, 1618potentially due to stronger housing demand driven by urbanization.1619When incorporating the interaction between the tax dummy and the effective tax 1620rate, the coefficients on the interaction term are consistently negative and statisti-1621cally significant across all specifications. In the short-term transaction models, the 1622coefficients range from –1.121 to –1.131 for municipalities and –2.069 to –2.096 1623for nonmunicipalities. In the whole transaction models, the coefficients range from 1624–1.083 to –1.113 for municipalities and –1.882 to –1.984 for nonmunicipalities. 1625These results provide further evidence that higher tax rates are associated with 1626greater reductions in housing prices, with the effect being more pronounced in non-1627municipal areas.16286.3 Robustness check using propensity score matching1629To further assess the robustness of our findings, we employ PSM to estimate the 1630average treatment effect on the treated (ATT), comparing treated transactions to 1631observationally similar controls. Specifically, we estimate ATT using one-to-one 1632nearest neighbor matching with a caliper of 0.05. The treated group consists of hous-1633ing units subject to one of the SSGST tax treatments, represented by a general tax 1634dummy, a 10% tax rate, or a 15% tax rate. Propensity scores are derived from logis-1635tic regressions that include a comprehensive set of covariates, such as housing char-1636acteristics, macroeconomic indicators, quarterly dummies, and either county and 1637year dummies or county-year interaction dummies (Caliendo & Kopeinig, 2005).1638Table 8 Effect of the SSGST on Number of Transactions1639Note:1. Standard errors in parentheses; 2.* p < 0.10, ** p < 0.05, *** p < 0.01; 3. We observed that the 1640tax proportion variable takes on extreme values of 1 or 0 in five instances. These values arise from a 1641limited number of transactions that could be classified as short-term or long-term during the initial and 1642final months of the dataset. To account for these extreme cases, we exclude the five observations and 1643re-estimate the models in Table 8. The coefficients across all four specifications remain negative and 1644statistically significant with 35 observations. Due to space limitations, the results are not reported but are 1645available upon request1646N_tran1647N_tran1648N_tran1649N_tran1650OLS1651OLS1652OLS1653OLS1654Tax proportion1655− 1075.0***1656− 1138.1***1657− 1059.9***1658− 159.01659(210.498)1660(304.587)1661(307.125)1662(146.022)1663Macro1664No1665No1666Yes1667Yes1668Quarter dummies1669No1670Yes1671Yes1672Yes1673Year dummies1674No1675No1676No1677Yes1678N1679401680401681401682401683adj. R-sq16840.21116850.24916860.60516870.875168816891690===== PAGE 37 (score 21, numeric-paren 2, table-words 7, var-words 5) =====16911692The Japanese Economic Review 1693Table 9 Test of Transfer 1694of Holding Periods from 1695Short-Term to Long-Term 1696Transactions1697Note: Standard errors in parentheses; * p < 0.101698LR_ D _N_tran1699OLS1700L.N_tran1701− 0.164*1702(0.081)1703Macro1704Yes1705Quarter dummies1706Yes1707Year dummies1708Yes1709N1710281711adj. R-sq17120.4331713Table 5 presents the ATT estimates for housing units transacted only once and 1714for those transacted more than once, while Table 6 displays the results separately for 1715municipalities and nonmunicipalities. Across both tables, the ATT estimates, meas-1716ured as the differences between treated and control group means, associated with 1717the general SSGST tax dummy are consistently negative and statistically significant 1718at the 10% level. For example, in the specification for housing units transacted only 1719once using county and year dummies, the difference is − 0.076, significant at the 172010% level. These results consistently indicate that the implementation of the SSGST 1721led to a reduction in housing prices.1722When comparing the ATT estimates for the 10% and 15% tax rates, the results 1723suggest that the 15% tax generally resulted in larger price reductions than the 10% 1724tax. Although some of the estimates, particularly for housing units transacted only 1725once and for whole-market transactions in municipalities, are not statistically sig-1726nificant, the overall pattern still points to a stronger effect associated with the higher 1727tax rate. In addition, the ATT estimates are consistently larger in magnitude in non-1728municipalities than in municipalities. For example, in the case of short-term trans-1729actions, the ATT estimate for the general tax dummy using county-year interaction 1730dummies is − 0.302 in nonmunicipalities and − 0.130 in municipalities, both statisti-1731cally significant. This suggests that housing prices in nonmunicipal areas are more 1732sensitive to the SSGST.1733Overall, the propensity score matching results reinforce the main findings. They 1734support the robustness of the conclusion that the SSGST has a price-depressing 1735effect on housing markets, with greater reductions observed under higher tax rates 1736and in nonmunicipal regions.17376.4 Spillover effect generated by speculation suppression1738We further investigate the mechanism underlying the effectiveness of the SSGST. 1739We investigate potential price spillover effect from the tax by incorporating aspects 1740of the short-term transactions included in Table 2 into the data presented in Table 3. 1741As a result, the number of short-term transaction observations increases from 26,794 1742to 28,168, while that of long-term transactions remains at 2,491. We use fixed effects 1743models. The results regarding the association between aggregate trade volumes and 1744