1# RECORD 41a24683a4b966679145f8c822178e61a9a339a42# TITLE: A Comparative Hedonic Analysis of Condominium Asking Rents in Bangkok, Phuket, and Pattaya3# YEAR: 2026.0 VENUE: International Journal of Building, Urban, Interior and Landscape Technology DOI: 10.56261/built.v24.2665924# AUTHORS: Than Dendoung5# PAGES: 16 CANDIDATE TABLE PAGES (1-based): [8, 11]6# ABSTRACT: This study estimates condominium rental price differentials from online listings across Thailand’s three major tourist cities, using Bangkok as the benchmark for Phuket and Pattaya while controlling for observable unit characteristics. Intercity differentials are measured with a hedonic regression in which the dependent variable is the natural logarithm of monthly rent (in Thai baht). Model 1 controls for the number of bedrooms, while Model 2 adds unit size through log(size_sqm), allowing the effect of unit composition on estimated city effects to be assessed. Confidence intervals at 95% are based on heteroskedasticity-robust (HC1) standard errors. The main finding is the magnitude and persistence of city-level effects relative to Bangkok. In Model 1, Pattaya has a monthly rent 11.5% lower (95% CI -13.3% to -9.6%) and Phuket 23.8% higher (95% CI 21.4% to 26.3%). When apartment size is included in Model 2, the monthly rent in Pattaya is 16.8% lower (95% CI -18.3% to -15.2%) and in Phuket 14.3% higher (95% CI 12.5% to 16.2%), and the model fit improves as the adjusted R-squared rises from 0.491 to 0.663. These results indicate that factors beyond apartment size drive rental differences across cities, since the gaps remain statistically significant after adjusting for size. The study offers robust, transparent estimates of rental levels in Thailand’s major tourist cities that can support comparison and benchmarking.789===== PAGE 1 (score 16, numeric-paren 0, table-words 3, var-words 10) =====1011Dendoung, T. 112Research Article13A Comparative Hedonic Analysis 14of Condominium Asking Rents 15in Bangkok, Phuket, and Pattaya16Than Dendoung17Faculty of Architecture and Planning, Thammasat University, Phahonyothin Road, Khlong Nueng Subdistrict, 18Khlong Luang District, Pathum Thani Province 12121, Thailand19Corresponding author e-mail: dendoung.than@gmail.com20Received: 1 Jul 2026; Revised from: 31 Aug 2026 ; Accepted: 8 Sep 202621Print-ISSN: 2228-9135, Electronic-ISSN: 2258-9194, https://doi.org/10.56261/built.v24.26659222Abstract23 24In Thailand, rent comparisons between cities are based on unadjusted averages 25of listings, which confuses the mix of units advertised with the price a city sets on 26property. I ask how much of the rent difference between Bangkok, Phuket, and Pattaya is 27composition and how much survives it. I use a cross-section of 14,849 active condominium 28rental listings from FazWaz in January 2026 to estimate nested hedonic models. The 29dependent variable is the natural log of monthly asking rent in Thai baht; all inference 30uses heteroskedasticity-robust (HC1) standard errors. Model 1 controls bedroom count, 31Model 2 adds ln(size), Model 3 adds bathrooms and view type. The raw picture misleads: 32Model 1 places Phuket 22.8% above Bangkok (95% CI 20.3% to 25.4%) and Pattaya 3311.5% below (95% CI -13.4% to -9.6%). Controlling size moves both, Phuket to 14.4% and 34Pattaya to 16.8% below, and lifts adjusted R-squared from 0.492 to 0.662; about a third 35of the apparent Phuket premium is size composition, not location. The central result is 36asymmetric. Adding bathrooms and view type leaves Phuket essentially unchanged, 0.130 37to 0.134 log points, but widens the Pattaya discount from 15.6% to 22.4%: The listed 38stock in Pattaya leans toward sea-view and pool-view product and the city indicator had 39been absorbing that mix. City coefficients based on listing data are not location premiums 40or willingness to pay, but conditional associations. A third of the variation in rent is 41unexplained and the attributes that would close the gap are named.42Keywords: hedonic price model; condominium rental market; asking rents; online listings; 43Bangkok; Phuket; Pattaya44 45https://doi.org/10.56261/built.v24.266592464748===== PAGE 8 (score 17, numeric-paren 0, table-words 4, var-words 7) =====4950BUILT 24(2). 2026, 266592518524. Results534.1 Overview of analytic sample54 55The analytical sample consists of 14,849 56condominium listings: 9,774 in Bangkok, 572,835 in Phuket, and 2,240 in Pattaya. 58Bedroom counts are derived from the 59title-based extraction rule in Section 3.2, 60which reduces missingness and enhances 61comparability between cities by classifying 62studios as zero and parsing stated counts 63as integers. Table 5 lists the number of 64bedrooms, unit size, and monthly asking 65rent by city. The log-linear specification 66is motivated by the widely dispersed and 67strongly right-skewed nature of both rent 68and size.694.2 Distributional diagnostics70 71All three cities’ asking rents have a long 72upper tail and a significant right skew. 73The rent distributions are plotted on a 74logarithmic horizontal axis in Figure 5, 75which allows the cross-city differences to 76be seen without distorting the majority 77of the observations. Because the log 78transformation stabilizes the variance and 79enables the coefficients to be interpreted 80as semi-elasticities and elasticities, the 81pattern supports modeling the dependent 82variable as ln(rent).83 84Figure 3. Rent per sqm density by city85Figure 4. Rent per sqm by bedrooms, faceted by city 86Table 5 | Descriptive Statistics by City | N = 14,849.87City88N89Minimum90Maximum91Mean92Std. Deviation93Bangkok94Monthly rent (THB)959,774963,500971,390,0009840,4039944,184100Unit size (sqm)1019,77410220.00103772.4910466.8210559.46106Bedrooms1079,7741080.001096.001101.421110.79112Phuket113Monthly rent (THB)1142,83511510,0001161,420,00011741,44411852,192119Unit size (sqm)1202,83512117.50122650.0012361.2012453.55125Bedrooms1262,8351270.001285.001291.121300.80131Pattaya132Monthly rent (THB)1332,2401346,000135290,00013627,63313722,032138Unit size (sqm)1392,24014021.00141450.0014256.2514340.74144Bedrooms1452,2401460.001475.001481.101490.80150Note. Summary statistics for the analytic sample (N = 14,849) reporting monthly asking rent (THB), unit size (sqm) and bedroom count by city. 151152153===== PAGE 9 (score 12, numeric-paren 0, table-words 4, var-words 3) =====154155Dendoung, T. 9156Figure 5. Rent distributions by city, log scale157Table 6 | Coefficients, Heteroskedasticity-Robust (HC1) Standard Errors | N = 14,849.158Model159Unstandardized 160Coefficients161Standardized Coefficients162t163Sig.16495% Confidence Interval for B165B166Robust Std. Error167Beta168Lower Bound169Upper Bound1701171(Intercept)1729.491173.0081741128.881175.0001769.4741779.507178Phuket179.205180.011181.11718219.402183.000184.185185.226186Pattaya187-.122188.011189-.064190-11.309191.000192-.143193-.101194Bedrooms195.589196.005197.693198114.160199.000200.579201.5992022203(Intercept)2046.902205.038206181.536207.0002086.8272096.977210Phuket211.135212.009213.07721415.625215.000216.118217.151218Pattaya219-.184220.009221-.096222-19.629223.000224-.202225-.165226Bedrooms227.104228.009229.12223012.202231.000232.087233.120234ln(size)235.822236.012237.69923867.554239.000240.798241.846242Note. Estimated coefficients for Models 1 and 2 with heteroskedasticity-robust (HC1) standard errors and associated test statistics. Robust inference is 243emphasized given cross-sectional listing data and potentially heteroskedastic residuals. The dependent variable is ln(rent) and Bangkok is the reference 244city category. N = 14,849245246247===== PAGE 11 (score 33, numeric-paren 0, table-words 9, var-words 13) =====248249Dendoung, T. 112504.5 Richer specification: bathrooms and 251view type252 253The city indicators in Models 1 and 2 254may absorb unmeasured attributes that 255systematically differ between markets. As a 256result, Model 3 adds the two characteristics 257that the listing records report at high 258coverage: the number of bathrooms and 259the advertised view type. Figure 7 shows 260the city coefficients for each of the three 261nested specifications, and Table 7 presents 262the estimates with robust standard errors.263 264Both added attributes are priced 265as expected. Keeping size, number of 266bedrooms, and city constant, an extra 267bathroom is linked to an 8.0% increase in 268rent. View type is economically substantial. 269Sea views are about 17.2% and lake views 270are about 20.0% more expensive than city 271views, which are the reference category; 272pool, garden, mountain, and river views 273are between 7% and 15% less expensive. 274The latter should not be interpreted as 275distaste for those outlooks because the 276coefficient shows that these outlooks are 277more common in less expensive inland and 278secondary regions.279Table 7 | Model 3 with Heteroskedasticity-Robust (HC1) Standard Errors | N = 13,224.280Unstandardized 281Coefficients282Standardized 283Coefficients284t285Sig.28695% Confidence 287Interval for B288%Δ289B290Robust Std. Error291Beta292Lower Bound293Upper Bound294(Intercept)2957.097296.046297152.770298.0002997.0063007.188301Phuket302.134303.012304.07730510.974306.000307.110308.158309+14.4310Pattaya311-.254312.013313-.135314-20.278315.000316-.279317-.230318-22.4319View: Garden320-.074321.016322-.024323-4.728324.000325-.105326-.043327-7.1328View: Lake329.183330.033331.0293325.490333.000334.117335.248336+20.0337View: Mountain338-.079339.017340-.027341-4.519342.000343-.113344-.045345-7.6346View: Pool347-.136348.013349-.054350-10.314351.000352-.162353-.110354-12.7355View: River356-.158357.049358-.013359-3.205360.001361-.255362-.061363-14.6364View: Sea365.159366.015367.07836810.491369.000370.129371.188372+17.2373View: Skyline374.044375.014376.0173773.206378.001379.017380.071381+4.5382View: Unblocked 383Open384-.092385.015386-.031387-6.075388.000389-.121390-.062391-8.8392Bedrooms393.088394.009395.1043969.521397.000398.070399.107400+9.2401ln(size)402.755403.016404.63340547.358406.000407.723408.786409Bathrooms410.077411.013412.0754135.997414.000415.052416.102417+8.0418Note. Model 3 adds bathroom count and view type to Model 2. The dependent variable is ln(rent), Bangkok is the reference city category and a city 419view is the reference view category. N = 13,224. Adjusted R squared = 0.668. The %Δ column converts a coefficient to a percentage effect using Equation 4204 and is reported only where a one-unit change in the regressor is meaningful, that is, for the indicator and count variables. It is left blank for ln(size), 421whose coefficient is an elasticity and is read as the percentage change in rent associated with a one percent increase in floor area. 422 423The main outcome of the extended 424specification is that the effect on the 425city coefficients is asymmetric. Since the 426Phuket premium, as estimated on the 427common sample, is essentially unchanged, 428changing from 0.130 to 0.134 log points, 429the conclusion that Phuket commands a 430genuine premium over Bangkok survives 431the addition of these variables. In contrast, 432the Pattaya discount increases from -0.170 433to -0.254 log points, or around 15.6% 434to 22.4% less than Bangkok. Pattaya’s 435listed stock is weighted toward sea-view 436and pool-view product, and once those 437attributes are priced individually the city’s 438underlying position turns out to be weaker 439than the size-adjusted comparison suggests. 440This is direct evidence for H3, and evidence 441that city indicators in parsimonious hedonic 442specifications can absorb compositional 443differences in amenity mix as well as in 444unit size. It is the principal reason all city 445coefficients reported here are described 446as conditional associations rather than as 447location premiums. 448449450===== PAGE 12 (score 43, numeric-paren 0, table-words 11, var-words 21) =====451452BUILT 24(2). 2026, 26659245312454BUILT 24(2). 2026, 26659245512456Figure 7. City coefficients relative to Bangkok across nested specifications4574.6 Robustness checks458 459Two further attributes are reported for 460only part of the sample, so I examine them 461in subsamples rather than in the pooled 462specification. Table 8 reports both. 463For Bangkok, the listing records report 464distance to the nearest BTS or MRT station 465for 83% of listings. Added to the Model 3 466specification for Bangkok alone, it returns 467a coefficient of -0.082, so each additional 468kilometer from a station is associated 469with rent about 7.9% lower, holding unit 470attributes constant. This is the accessibility 471control that the pooled specification lacks, 472and its magnitude is consistent with the 473accessibility literature discussed in Section 4742.4. The comparable field is absent for 475Phuket and Pattaya, where urban rail does 476not operate. 477 478More than 95% of listings in Phuket 479and Pattaya provide the floor level, while 480only 4% of listings in Bangkok do so. I add 481floor level to the Model 3 specification for 482the two coastal cities and get a coefficient 483of 0.0017 per story that is not statistically 484significant (p = 0.18). Once view type 485is controlled, height per se does not 486command a measurable premium in these 487markets: it is the outlook that is priced, 488for which floor level is only a proxy. In this 489subsample a sea view is worth about 31.5% 490and the Pattaya discount relative to Phuket 491about 34.3%.492 493Additionally, I created and tested 494an index counting advertised amenities, 495covering furnishing, balcony, washing 496machine, wifi, pool access, gym, microwave, 497television and recent renovation, and 498excluded it from the reported specifications. 499It entered negative and significant, 500an implausible sign that diagnostic 501examination attributes to the source data’s 502structure rather than preferences. There 503are only a few attribute slots available in 504the listing design, which are filled with basic 505appliances in lower-priced ads and various 506descriptions in higher-priced ads.507Mean amenity count accordingly falls 508from 2.35 in the smallest size quartile 509to 1.82 in the largest, and the individual 510indicators price a television at -5.0% and 511Table 8 | Subsample Robustness Checks | N = 7,195 and 4,422.512Specification513B514Robust Std. 515Error516Sig.517%Δ518Bangkok 519(+ transit distance)520Distance to BTS/521MRT (km)522-.082523.004524.000525-7.9526ln(size)527.735528.024529.000530Bathrooms531.086532.015533.000534+9.0535Bedrooms536.070537.014538.000539+7.2540Phuket & Pattaya 541(+ floor level)542Floor level543.002544.001545.183546+0.2547ln(size)548.629549.021550.000551Bathrooms552.073553.026554.005555+7.6556Bedrooms557.157558.014559.000560+17.0561Pattaya562-.420563.016564.000565-34.3566Note. Both specifications add the stated attribute to the Model 3 covariate set. The 567Bangkok specification is estimated on 7,195 listings with adjusted R squared 0.674. The 568Phuket and Pattaya specification is estimated on 4,422 listings with adjusted R squared 5690.703, with Phuket as the reference city. The dependent variable is ln(rent) in both cases. As 570in Table 7, the %Δ column is left blank for ln(size), whose coefficient is an elasticity. 571