1# RECORD d3e27cac15a096168d22a145ab49342c924dc96e2# TITLE: Bridging the information gap: what are the impacts of EPCs on Prague’s residential market selling prices?3# YEAR: 2025.0 VENUE: Financial Internet Quarterly DOI: 10.2478/fiqf-2025-00264# AUTHORS: Tomáš Podškubka; Barbora Rýdlová; Josef Arlt5# PAGES: 12 CANDIDATE TABLE PAGES (1-based): [7, 8]6# ABSTRACT: Abstract It is assumed that energy performance certificates (EPC) can promote energy efficiency, increase property values, and improve market liquidity by providing trustworthy energy performance information. This article will provide a comprehensive study of the impact of EPCs on the residential selling prices in Prague, the capital of the Czech Republic. For the analysis, the original dataset was extracted from the publicly accessible sreality.cz portal. The hedonic regression model was applied to evaluate how the EPC influences the selling price of apartments in Prague. The regression analysis results are entirely in line with the analyzed current literature. EPC ratings have been proven to significantly affect the selling price in Prague and several other locations. In the most affluent locations of Prague, i.e., typically historical locations, the EPC ratings have not proved to be a statistically significant price factor because these locations suffer from excess demand and limited supply of properties.789===== PAGE 1 (score 8, numeric-paren 0, table-words 0, var-words 8) =====1011 39 12 13Financial Internet Quarterly 2025, vol. 21 / no. 4, p. 3914-50 1510.2478/fiqf-2025-0026 16BRIDGING THE INFORMA17TION GAP: WHAT ARE THE 18 19IMPACTS OF EPCS ON P20RAGUE ’S RESIDENTIAL MARKET21 22SELLING PRICES?23 24Tomáš Podškubka251, Barbora Rýdlová262, Josef Arlt273 28Abstract 29It is assumed that energy performance certificates (EPC) can promote energy efficiency, increase 30property values, and improve market liquidity by providing trustworthy energy performance 31information. This article will provide a comprehensive study of the impact of EPCs on the resi-32dential selling prices in Prague, the capital of the Czech Republic. For the analysis, the original 33dataset was extracted from the publicly accessible sreality.cz portal. The hedonic regression 34model was applied to evaluate how the EPC influences the selling price of apartments in Prague. 35The regression analysis results are entirely in line with the analyzed current literature. EPC 36ratings have been proven to significantly affect the selling price in Prague and several other loca-37tions. In the most affluent locations of Prague, i.e., typically historical locations, the EPC ratings 38have not proved to be a statistically significant price factor because these locations suffer from 39excess demand and limited supply of properties. 40JEL classification: C31, Q40, R21, R31 41Keywords: Energy Performance Certificate, Energy Efficiency, Czech Housing Market, Residential Properties, Hedonic Regression Model 42Received: 30.01.2025 Accepted: 28.04.2025 43Cite this: 44Podškubka, T., Rýdlová, B. & Arlt, J. (2025). Bridging the information gap: what are the impacts of EPCs on Prague’s residential market selling prices? 45Financial Internet Quarterly 21(4), pp. 39-50. 46© 2025 Tomáš Podškubka et al. published by Sciendo. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 474.0 License. 48www.finquarterly.com 49University of Information Technology and Management in Rzeszów 501 Department of Finance and Business Valuation, Faculty of Finance and Accounting, Prague University of Economics and Business, Czechia, email: 51tomas.podskubka@vse.cz, ORCID: https://orcid.org/0000-0002-8761-2661. 522 Department of Finance and Business Valuation, Faculty of Finance and Accounting, Prague University of Economics and Business, Czechia, email: 53barbora.rydlova@vse.cz, ORCID: https://orcid.org/0000-0003-0701-730X. 543 Department of Finance and Business Valuation, Faculty of Finance and Accounting, Prague University of Economics and Business, Czechia, email: 55josef.arlt@vse.cz, ORCID: https://orcid.org/0000-0001-8349-7419. 565758===== PAGE 7 (score 38, numeric-paren 16, table-words 6, var-words 8) =====5960 45 61 62www.finquarterly.com 63University of Information Technology and Management in Rzeszów 64Tomáš Podškubka, Barbora Rýdlová, Josef Arlt 65Bridging the information gap: what are the impacts of EPCs on Prague’s residential 66market selling prices? 67Financial Internet Quarterly 2025, vol. 21 / no. 4 68further affect the F-test and can cause misleading inter-69pretation of the index of determination R2. Multicollin-70earity, the high correlation among the independent 71variables, can lead to inaccurate regression parameter 72estimates and overestimating their standard errors. It 73also causes problems with the interpretation of partial 74regression parameters. Namely, it isn't easy to under-75stand the individual impact of the particular independ-76ent variable, provided that the other independent vari-77able is constant when these variables are correlated. 78 79Results80 81The results of the regression analysis are presented 82in Table 3. All computations were performed using the 83software package EViews 13. The Jarque-Bera test 84(Jarque & Bera, 1980) rejects the hypothesis of error 85term normality in most cases. However, histograms, 86skewness and kurtosis of the error terms indicate that 87their distribution is not far from normal. Since the Cen-88tral Limit Theorem ensures that normality is not re-89quired if the sample size is large enough, it can be stat-90ed that non-normality does not distort the obtained 91results. The Breusch-Pagan-Godfrey (Breusch & Pagan, 921979; Godfrey, 1978) and Glejser tests (Glejser, 1969) 93reject homoscedasticity hypotheses in most cases. 94However, for heteroscedasticity to cause diagnostic 95test distortion , it must be extreme (as noted by Lumley 96et al., 2002), which is not the case in this study. The 97data do not contain extreme values, and the distribu-98tions of the error terms is close to normal, as men-99tioned above. The Breusch-Godfrey autocorrelation 100test (Breusch, 1978; Godfrey, 1978) does not reject the 101hypothesis of no autocorrelation in all cases. 102Methodology103 104To evaluate how the energy performance class 105influences the selling price of apartments in Prague, the 106following hedonic regression model is applied: 107(1) 108where the dependent variable pi, indexed by apartment 109(i) represents the price. The independent variables xij 110represent area_net, sep_kitchen, sqm_per_-room, 111state, location and variable of interest EPC. The param-112eter c is the intercept and βj, j = 1, …, k are the partial 113regression parameters. In this model, for a unit in-114crease in the independent variable xij, the expectation 115of the dependent variable changes by the percentage 116100(eβj - 1), provided that all other independent varia-117bles are held constant. When |βj| < 0.1 this percentage 118approaches to 100βj. The Ɛi are observation-specific 119zero-mean random error terms, which are a set of 120equally distributed independent random variables, i.e. 121Ɛi ~ IID(0, σ2). 122It is assumed that the error terms of the regression 123model (1) are normally distributed. This assumption is 124necessary to perform correct t-tests for the parameters 125of this model, when it is violated, the testing criterion: 126 127can be biased. Heteroscedasticity, the failure to satisfy 128the condition of constant variances of the error terms 129(homoskedasticity), i.e., D(Ɛi) = σ2130i can bias the standard 131error estimates sβj, resulting in incorrect t-tests. The 132presence of autocorrelation of the error terms, i.e. E133(ɛiɛj) ≠ 0, i ≠ j, can also bias the standard error esti-134mates. Both heteroscedasticity and autocorrelation can 135Number of 136Location 137Description 1384 139Bohnice, Čimice, Hloubětín, Hodkovičky, Hostivař, Kamýk, Košíře, Krč, Malešice, Motol, Prosek, 140Radlice, Řepy, Stodůlky, Střížkov, Štěrboholy, Veleslavín, Vokovice, Vysočany, Zličín 1415 142Běchovice, Čakovice, Černý Most, Dolní Chabry, Dolní Měcholupy, Dolní Počernice, Háje, 143Hlubočepy, Horní Měcholupy, Horní Počernice, Hostavice, Hrdlořezy, Chodov, Klánovice, Koloděje, 144Kolovraty, Kunratice, Kyje, Letňany, Lhotka, Modřany, Radotín, Ruzyně, Řeporyje, Slivenec, Such-145dol, Šeberov, Velká Chuchle, Vinoř, Zbraslav 1466 (least 147attractive) 148Dubeč, Cholupice, Jinonice, Kbely, Královice, Křeslice, Liboc, Libuš, Lipence, Lipenice, Lysolaje, 149Miškovice, Petrovice, Písnice, Pitkovice, Satalice, Sedlec, Sobín, Točná, Třebonice, Třeboradice, 150Uhříněves, Újezd nad Lesy, Újezd u Průhonic 151Source: Authors’ own work. 1521153ln154,1551,...,156k157i158j159ij160i161j162p163c164x165i166N167168169=170=171+172+173=174175176177j178j179t180s181182=183Table 3: Regression results of log price per square meter in Czech crowns in Prague 184 185Prague 186Location 187 1881 1892 1903 1914 1925 1936 194c 19511.619*** 19611.485*** 19711.691*** 19811.436*** 19911.435*** 20011.398*** 20111.292*** 202area_net 2031.26E-5 2047.83E-5 2050.000 2060.001** 2070.000 208-0.001*** 209-0.000 210sep_kitch 211-0.019 212-0.002 213-0.069 214-0.017 215-0.035 2160.022 2170.036 218sqm_room 2190.015*** 2200.013*** 2210.009*** 2220.013*** 2230.015*** 2240.018*** 2250.016*** 226227228===== PAGE 8 (score 91, numeric-paren 47, table-words 13, var-words 5) =====229230 46 231 232www.finquarterly.com 233University of Information Technology and Management in Rzeszów 234Tomáš Podškubka, Barbora Rýdlová, Josef Arlt 235Bridging the information gap: what are the impacts of EPCs on Prague’s residential 236market selling prices? 237Financial Internet Quarterly 2025, vol. 21 / no. 4 238EPC (-0.5355) distorts the results, increasing the param-239eter estimates for EPC (decreasing them in absolute 240value) and increasing the estimates of their standard 241errors. The solution is to exclude state from the model 242and analyze only the apartments for which state takes 243the value 1 (after reconstruction, project, under con-244struction, new construction). The results of this regres-245sion analysis are presented in Table 4. 246There is a certain risk of multicollinearity. The ex-247planatory variable sep_kitchen is correlated with the 248variable state (-0.4068) and with the EPC (0.4108), and 249the variable area_net is correlated with sqm_per_room 250(0.6532). However, the parameter estimates for both 251variables are, with two exceptions, statistically insignifi-252cant, and their exclusion has no effect on the model 253parameter estimates. The correlation of the state with 254 255Prague 256Location 257 2581 2592 2603 2614 2625 2636 264state 2650.180*** 2660.266*** 2670.073* 2680.180*** 2690.178*** 2700.128*** 2710.178*** 272location 273-0.056*** 274 275 276 277 278 279 280EPC 281-0.024*** 2820.021 283-0.001 284-0.016 285-0.502 286-0.064*** 287-0.026 288Mean dep 28912.024 29012.227 29112.154 29211.994 29312.034 29411.967 29511.942 296SD dep 2970.459 2980.506 2990.381 3000.421 3010.464 3020.493 3030.394 304R2 3050.491 3060.383 3070.359 3080.482 3090.548 3100.498 3110.487 312F-test 313672.913*** 31432.042*** 31540.147*** 316169.256*** 317309.155*** 318210.649*** 31957.127*** 320D-W 3211.995 3222.981 3231.886 3242.079 3252.115 3261.828 3272.398 328N 3294189 330264 331364 332914 3331281 3341059 335307 336Notes: 1) The LS point estimates of parameters of regression model (1). 2) Asterisks denote statistical significance: 337*** p < 0.01, ** p < 0.05, * p < 0.1. 3) Mean dep – mean of the dependent variable, SD dep – standard deviation of 338the dependent variable, R2 – index of determination, D-W – Durbin Watson test, N – number of observations 339Source: Authors’ own work. 340 341Prague 342Location 3431 3442 3453 3464 3475 3486 349 350c 35111.972*** 35212.080*** 35311.881*** 35411.814*** 35511.815*** 35611.728*** 35711.536*** 358area_net 3593.71E-05 3600.001** 3610.001 3620.001 3630.001 364-0.001*** 365-0.001 366sep_kitch 367-0.055* 368-0.189 369-0.022 3700.034 371-0.084 3720.013 3730.145 374sqm_room 3750.014*** 3760.006** 3770.008*** 3780.010*** 3790.015*** 3800.018*** 3810.016*** 382location 383-0.056*** 384 385 386 387 388 389 390EPC 391-0.080*** 392-0.039 393-0.028 394-0.065*** 395-0.139*** 396-0.142*** 397-0.055 398Mean dep 39912.156 40012.417 40112.201 40212.139 40312.186 40412.107 40512.030 406SD dep 4070.446 4080.499 4090.366 4100.385 4110.439 4120.502 4130.368 414R2 4150.393 4160.253 4170.301 4180.387 4190.467 4200.408 4210.423 422F-test 423329.703*** 42410.649*** 42524.976*** 42673.744*** 427178.399*** 428115.832*** 42939.508 430D-W 4311.960 4323.969 4331.265 4342.339 4351.886 4361.495 4371.989 438N 4392557 440131 441237 442472 443818 444678 445221 446Table 4: Regression results of log price per square meter in Czech crowns in Prague 447Notes: 1) The point estimates of parameters of regression model (1). 2) Asterisks denote statistical significance: 448*** p < 0.01, ** p < 0.05, * p < 0.1. 3) Mean dep – mean of the dependent variable, SD dep – standard deviation of 449the dependent variable, R2 – index of determination, D-W – Durbin Watson test, N – number of observations. 4) Sta-450te: new construction, after reconstruction, project, under construction 451Source: Authors’ own work. 452supply is limited, and property is scarce, which dimin-453ishes the signaling and information asymmetry power 454of EPCs and boosts the importance of location itself. 455These findings are directly supported by several au-456thors (e.g. Marmolejo-Duarte & Chen, 2019, Wilhelms-457son, 2019 and Taruttis et al., 2022). It is worth noting 458that the mentioned studies were performed in Spain, 459Norway and especially Germany, which is a neighboring 460market for the Czech Republic and the most important 461Results of the adjusted regression analysis con-462tained in Table 4 are entirely in line with the analyzed 463current literature. This paper has proven that EPC 464ratings affect the selling price in Prague, taken as one 465location, and in Locations 3, 4, and 5 with the parame-466ters t-test p-values lower than 0.01. On the other hand, 467affluent Prague Locations 1 and 2 achieved statistically 468immaterial results for EPC influence, which can be in-469terpreted such that in prominent locations, real estate 470471472===== PAGE 9 (score 13, numeric-paren 0, table-words 1, var-words 11) =====473474 47 475 476www.finquarterly.com 477University of Information Technology and Management in Rzeszów 478Tomáš Podškubka, Barbora Rýdlová, Josef Arlt 479Bridging the information gap: what are the impacts of EPCs on Prague’s residential 480market selling prices? 481Financial Internet Quarterly 2025, vol. 21 / no. 4 482a sophisticated analysis of the unique data from Pra-483gue, Czech Republic. Prague’s real estate market is 484quite specific as it is characterized by high prices and 485a significant affordability gap compared to other major 486EU cities. This analysis is framed by the theory of infor-487mation asymmetry and the signaling theory, as the 488EPCs should mitigate information asymmetry in the real 489estate market and signal the energy performance of 490buildings. Using the most frequent hedonic regression 491analysis, the relationship between energy efficiency 492measured by EPCs and property prices on data from 493Prague while focusing on several micro locations has 494been tested. 495Our results can be summarized in the following 496three points: 4971. EPCs have an impact on prices - EPC in Prague apart-498ments statistically significantly influences the price. 499This finding holds in 3 out of 6 micro-locations and 500the Prague region as a whole. 5012. The degree of influence differs by micro-location - In 502the most affluent locations of Prague, i.e., typically 503historical (central) locations, the EPC ratings have 504not proved to be a statistically significant price fac-505tor. This can be easily explained: these locations 506suffer from excess demand and limited supply of 507properties and are usually viewed as seller markets. 508Buyers have no option but to disregard most typical 509price drivers and focus only on location. In addition, 510historical locations are composed of old dwellings, 511and meeting state-of-the-art building standards 512through complete refurbishment is a cost and tech-513nical issue due to heritage protection. 5143. Apartment size also impacts prices - The size of the 515apartment, measured by square meters per room, 516was the other statistically significant explanatory 517variable of real estate prices in Prague. This aligns 518with theory and market practice, as smaller flats sell 519at a premium per unit price. 520The findings of this study imply that policymakers 521should continue to promote energy efficiency through 522EPC regulations and strive to increase public awareness 523of their long-term benefits, as higher EPC ratings are 524associated with increased property values. Incentives 525for property owners to improve their EPC ratings could 526further enhance these benefits. However, the results of 527this study suggest that the policy makers should design 528the EPC related programs taking into account regional 529market specifics. In Prague’s historical center we ob-530served no market driven motivation of the real estate 531owners for energy performance improvement, since 532the transaction motives lies in the premium location, 533the price effect of energy inefficiency proving negligi-534ble. Given the unique challenges faced by historical 535locations, support measures should be directed to-536wards these areas to help property owners compen-537trade partner encompassing almost 80% of Czech inter-538national trade within the EU. Considering this, compa-539rable regulation applies, and similar real estate 540patterns can be expected. Correspondingly, the model 541without separated locations, which was constructed for 542the whole Prague region, demonstrated the high statis-543tical significance of the parameter estimate of the loca-544tion factor, which is the most well-known value driver 545in real estate pricing (Fuerst et al., 2015 or Olaussen et 546al., 2019). 547The notion of asymmetrical effect of EPC is also 548presented by Fuerst et al. (2016), McCord et al. (2020), 549and Cespedes-Lopez et al. (2019). Their findings are, 550however, somewhat different because they claim that 551only properties with higher EPC scores display positive 552significant effects at the higher end of the price distri-553bution, but there are also brown discount effects evi-554dent for lower-rated properties within F and G-rated 555EPC properties, which contradicts findings in this paper. 556The other influential real estate price driver proved 557to be the size of the apartment as measured by square 558meters per room (sqm_room). This outcome is also 559logical and expected as smaller flats are typically sold at 560a premium to larger ones measured by unit price. The 561parameter estimates of this factor are statistically sig-562nificant in all regressions having p-value lower than 5630.01. 564The parameter estimates of the remaining two 565explanatory variables (area_net, sep_kitch) showed 566predominantly no statistical significance. 567 568Conclusions569 570The EPC regulation imposed in the EU aims to miti-571gate climate change and global warming issues. It is 572assumed that EPCs can promote energy efficiency, in-573crease property values, and improve market liquidity 574by providing trustworthy energy performance infor-575mation. However, increased energy efficiency is also 576associated with higher costs, so the economic rationale 577behind this idea must be sound. In any case, the energy 578efficiency of buildings will remain the central focus of 579EU policymakers, practitioners, and scholars, and the 580importance of EPCs will continue to increase. 581The literature review showed that many authors 582have already studied this issue. Nevertheless, the re-583sults remain inconclusive, and research gaps have been 584identified. The analyzed papers proved robust evidence 585that higher EPCs accompany higher real estate prices, 586though the extent of this impact varies by region, prop-587erty type, and methodological approach and is some-588what uneven. This paper will address the gap identified 589in the current literature, which is the need for more 590granular data from a different market and climate envi-591ronment (e.g., Marmolejo-Duarte & Chen, 2019, Wil-592helmsson, 2019; Taruttis et al., 2022) and provide 593