1# RECORD 784bd147866d1ec15bbfa3c42d3c2fc6a9ac31922# TITLE: The economic value of sustainability. Real estate market and energy performance of homes3# YEAR: 2023.0 VENUE: Valori e Valutazioni DOI: 10.48264/vvsiev-202334024# AUTHORS: Ezio Micelli; Giulia Gilberto; Eleonora Righetto; G. Tafuri5# PAGES: 28 CANDIDATE TABLE PAGES (1-based): [6, 7, 8]6# ABSTRACT: The energy transition and decarbonisation require a major transformation of the housing stock. Responsible for around 40% of fossil fuel consumption and 30% of CO2 emissions, the built heritage must be upgraded in terms of energy performance. The issue is particularly sensitive for Italy, where seven out of ten homes are in the worst energy classes according to the classification introduced by the EU. The research aimed at investigating whether, and to what extent, energy-efficient homes command a premium price on the market compared to less efficient homes. The study focused on three medium-sized cities in northern Italy and developed three hedonic models based on more than 900 ask prices. The results confirm the stratification of the market in terms of energy efficiency levels. Estimates converge in absolute and relative terms between different cities. Particularly significant is the average gap found between properties in the highest efficiency class (Class A) and those in the lowest efficiency class (Class G), with values of around thirty percentage points, while the gap is around fourteen percentage points if we consider properties in Class D compared to Class G. The conclusions are in line with European studies on the subject, but show a trend towards an increasing gap between the market values of properties with different energy performance classes. La transizione energetica e la decarbonizzazione impongono un’importante trasformazione del patrimonio abitativo. Responsabile di circa il 40% dei consumi di combustibili fossili e del 30% delle emissioni di CO2, il patrimonio costruito deve essere riqualificato sotto il profilo delle performance energetiche. Il tema è particolarmente delicato per l’Italia, dove sette abitazioni su dieci appartengono alle classi energetiche meno performanti secondo la classificazione introdotta dall’UE. La ricerca si è proposta di indagare se, e in quale misura, le abitazioni energeticamente performanti riscontrino sul mercato un premium price rispetto alle abitazioni meno efficienti. L’indagine si concentra su tre città medie del Nord Italia e ha sviluppato tre modelli edonici sulla base di oltre 900 prezzi d’offerta. I risultati confermano la stratificazione del mercato rispetto ai livelli di efficienza energetica. Le stime risultano convergenti in termini assoluti e relativi tra le diverse città. Particolarmente significativo è lo scarto medio rilevato tra immobili appartenenti alla classe di massima efficienza (class789===== PAGE 1 (score 5, numeric-paren 0, table-words 0, var-words 5) =====1011rivista valori e valutazioni n. 34 - 202312313Ezio Micelli*, Giulia Giliberto**, 14Key words: housing market, sustainable buildings, 15Eleonora Righetto***, Greta Tafuri**** 16energy transition, energy performance, hedonic prices 17The economic value 18of sustainability. 19Real estate market and 20energy performance 21of homes221. INTRODUCTION 23The energy transformation of the housing stock is one 24of the greatest challenges facing the international 25community. Building stock accounts for 40% of energy 26consumption and 36% of greenhouse gas emissions in 27Europe (European Commission, 2020). 28The energy transformation of the built heritage 29involves important actions in terms of energy savings 30and energy production from renewable sources. The 31challenge is to meet the carbon neutrality targets 32introduced by the European Green Deal.1 The housing 33market is therefore faced with an unavoidable 34challenge in the broader process of decarbonising 35cities.2 36The relationship between energy efficiency and 37property values is problematic. On the one hand, it can 38The energy transition and decarbonisation require a 39major transformation of the housing stock. Responsible 40for around 40% of fossil fuel consumption and 30% of 41CO2 emissions, the built heritage must be upgraded in 42terms of energy performance. The issue is particularly 43sensitive for Italy, where seven out of ten homes are in 44the worst energy classes according to the classification 45introduced by the EU. 46The research aimed at investigating whether, and to 47what extent, energy-efficient homes command a 48premium price on the market compared to less efficient 49homes. The study focused on three medium-sized cities 50in northern Italy and developed three hedonic models 51based on more than 900 ask prices. 52The results confirm the stratification of the market in 53terms of energy efficiency levels. Estimates converge in 54absolute and relative terms between different cities. 55Particularly significant is the average gap found between 56properties in the highest efficiency class (Class A) and 57those in the lowest efficiency class (Class G), with 58values of around thirty percentage points, while the gap 59is around fourteen percentage points if we consider 60properties in Class D compared to Class G. 61The conclusions are in line with European studies on 62the subject, but show a trend towards an increasing gap 63between the market values of properties with different 64energy performance classes.65Abstract661 Communication of the Commission of the European Parliament 67and Council to the European Economic and Social Committee 68and to the Committee of Regions COM (2019) 640 final. 692 Directive 2010/31/EU, 2010 of the European Parliament and of the 70Council on the energy performance of buildings. 717273===== PAGE 6 (score 31, numeric-paren 0, table-words 9, var-words 8) =====7475analyses are given in the appendix (Tables A, B and C). 76The sample analysis shows the preponderance of 77homes belonging to the lowest energy efficiency 78classes. In particular, the percentages of the least 79energy-efficient homes (E, F and G) reach percentages 80of 79.1%, 79.2% and 66.5% for Padua, Mestre and 81Bergamo, respectively (Fig. 1). The characteristics of the 82sample are consistent with the national building stock 83data, which show a building stock percentage for these 84three classes, amounting to 70.3%. 85The descriptive statistics show a homogeneous picture 86in terms of typological characteristics and state of 87maintenance. In fact, the majority are multi-family 88buildings in a good state of maintenance, despite the 89low energy efficiency highlighted (Tables 2, 3 and 4). 90The average unit value is € 2,034.85/sqm for the city of 91Padua, 1,905.43/sqm for Bergamo and 1,767.02 €/sqm 92for Mestre (Fig. 2). 934. METHODS AND ANALYSIS. THE HEDONIC 94MODELS FOR THE THREE CITIES 95The regression model used to estimate hedonic prices 96uses the natural logarithm of the unit’s market value as 97the unknown variable. The known variables are 98the 99locational, 100typological 101and 102technological 103characteristics of the housing unit. 104Formally, the model employed is as follows (Fuerst et 105al., 2016) Eq. (1): 106 107 108 109(1) 110 111 112where 113– Pi is the natural logarithm of the price of a home 114expressed in €/sqm; 115– β0 is the model’s constant; 116– βi is the marginal price of the characteristic; 117– Xi is the numerical value of the variables studied, 118including the EPC; 119– ei represents the random error. 120The regression models for Padua and Mestre show an 121R2 of 51.9% and 52.2% and an adjusted and corrected 122R2 of 51.3% and 51.2%, respectively. On the other hand, 123the model for the city of Bergamo shows lower values, 124with an R2 of 35.9% and an adjusted R2 of 34.8%. 125The reliability of the models was tested using the F-126statistic, whose p-value is less than 0.05 in all three 127cities (less than 0.001 for Padua, 0.034 for Mestre and 1280.024 for Bergamo). The results of the variance analysis 129(ANOVA) are presented in Table 5. 130journal valori e valutazioni No. 34 - 20231318132Table 2 - Padua: descriptive statistics 133Variable134n135mean136s.d.137min138max139Unit value [€/sqm]1403541412,034.85142805.451437901446,341145Zone (OMI)1463541471.721480.8914911503151Proximity to local 152transport 153infrastructure1543541551.991560.8515711583159Energy class 16035416110.251622.36163116412165Type1663541670.941680.2316901701171No. of bathrooms1723541731.851740.6817511763177Gross floor area 178[sqm]179354180143.3518167.7518226183450184Maintenance 185status1863371872.671880.7618911904191Table 3 - Mestre: descriptive statistics192Variable193n194mean195s.d.196min197max198Unit value [€/sqm]1992542001,767.02201538.952029502035,376204Zone (OMI)2052542061.592070.8120812093210Proximity to local 211transport 212infrastructure2132542141.842150.8121612173218Energy class 21925422010.212212.75222122312224Type2252542260.992270.1122802291230No. of bathrooms2312522321.372330.5023412353236Gross floor area 237[sqm]23825423996.4024028.3424138242205243Maintenance 244status2452462462.502470.7024812494250Table 4 - Bergamo: descriptive statistics251Variable252n253mean254s.d.255min256max257Unit value [€/sqm]2582932591,905.43260714.892616652624,793263Zone (OMI)2642932651.832660.9226712683269Proximity to local 270transport 271infrastructure2722932732.172740.7427512763277Energy class 2782932799.852802.41281128212283Type2842932850.972860.1628702881289No. of bathrooms2902932911.442920.5829312943295Gross floor area 296[sqm]297293298108.8629945.8530035301300302Maintenance 303status3042913052,413060.8030713084309 310311312 313314315===== PAGE 7 (score 30, numeric-paren 0, table-words 6, var-words 11) =====316317The economic value of sustainability. 318Real estate market and energy performance of homes319The hedonic prices of the characteristics all have a p-320value of less than 0.05, so their reliability can be 321considered significant. Table 6 highlights the model 322values and reports the implied marginal prices of the 323identified characteristics. 324It reports the Variance Inflation Factor (VIF), which 325provides a measure of multicollinearity between the 326independent variables. The factor allows the detection 327of any interaction effects present in the model (Bottero 328et al., 2018). In the three cases considered, a VIF close 329to one indicates that the variables are weakly 330correlated and therefore independent. 331For two of the three cities studied, the locational 332characteristics condition the articulation of the values. 333For Padua and Bergamo, the traditional centre-334periphery articulation marks the decline in property 335values. The case of Mestre is a significant exception, in 336that the market does not distinguish the traditional 337hierarchy between the centre and the periphery. 338Proximity to urban public transport infrastructure is a 339positional advantage that the models reported in two 340out of three cases, Padua and Mestre. One city is an 341exception: in Bergamo, no significant relationship was 342found 343between 344proximity 345to 346public 347transport 348infrastructure and residential market values. 349Property types that are adequately equipped record 350higher values, as indicated by the positive sign for the 351variable of the number of bathrooms in the cases of 352Mestre and Bergamo. As maintenance features 353increase, the unit value of the property also increases 354in all cities. 355Finally, in line with the law of diminishing marginal 356utility, the larger the gross floor area of the property, 357the lower the unit value. 358The EPC energy rating is a critical element of value 359creation in all three urban markets. The model 360confirms the empirical research cited above on the 361relevance of superior energy efficiency for value 362creation. The developed hedonic prices show similar 363and 364comparable 365magnitudes, 366confirming 367the 368robustness of the models. 369The models development the building energy 370performance variable in an ordinal form. To reach more 371robust conclusions regarding the contribution to the 372value of this intrinsic characteristic of the home, a 373second model considered energy class membership in 374the form of multiple dummy variables. 375The development of models based on variables of the 376latter type makes the estimation methodology more 377robust on account of the two alternative classifications 378of the variable sought and its weight in value creation. 379journal valori e valutazioni No. 34 - 20233809381Table 5 - Variance analysis of the three models382City383Sum of 384squares385df386Quadra-387tic 388mean389F390p value391Padua392Regression39325.11839453955.02439675.223 <0.001397Residual39823.2403993484000.067401Total40248.358403353404Mestre405Regression4069.14140754081.82840954.111 <0.001410Residual4118.3794122484130.034414Total41517,521416253417Bergamo418Regression41913.95942054212.7924222.200423<0.001424Residual42524.8834262874270.087428Total42938.842430292431Table 6 - Regression models432Padua433Mestre434Bergamo435Predictors Xi436βi437p value438VIF439βi440p value441VIF442βi443p value444VIF445Costant4469.190447<0.001448-4498.233450<0.001451-4527.970453<0.001454-455Zone (OMI) 456-0.244457<0.0014581.141459-460-461-462-0.0614630.0024641.015465Proximity to local 466transport 467infrastructure468-0.115469<0.0014701.047471-0.0324720.0344731.122474-475-476-477Energy class (EPC)478-0.049479<0.0014801.667481-0.039482<0.0014831.537484-0.043485<0.0014861.767487Type488-489-490-491-492-493-494-495-496-497No. of bathrooms498-499-500-5010.161502<0.0015031.6675040.275505<0.0015061.834507Gross floor area 508-0.001509<0.0015101.073511-0.003512<0.0015131.530514-0.002515<0.0015161.757517Maintenance 518status519-0.114520<0.0015211.682522-0.097523<0.0015241.462525-0.0665260.0245271.789528R2 Padua: 0.519; R2 Mestre: 0.522; R2 Bergamo: 0.359529530531===== PAGE 8 (score 26, numeric-paren 0, table-words 7, var-words 8) =====532533Following the buyer’s heuristic (Kahneman & Egan, 5342011), the homes were grouped into three domains. 535The first includes homes belonging to the efficient 536classes (EPC A and B). The second includes the 537intermediate classes (EPC C and D), while the last 538includes the lowest energy-efficient classes (EPC E, F 539and G). Membership of each class determines a 540different hedonic price according to a scheme that is 541no longer constrained, as before, by the ordinal nature 542of the variable. 543Increasing energy class corresponds to increasing 544statistically significant hedonic prices (p-value always 545<0.05): the patterns shown in Table 7 confirm the 546results of the patterns previously described in Table 6. 5475. DISCUSSION. VALUE DEVIATIONS AND 548COMPARISON WITH THE LITERATURE 549The analysis shows that there are significant 550differences in value according to the energy quality of 551the properties, confirming the theoretical premises 552outlined in the first section and the empirical research 553developed 554nationally 555and 556internationally. 557The 558economic advantage associated with lower operating 559costs is capitalised in the market value. Added to this 560is the prospect of marketing the property without 561constraints and limitations, a characteristic associated 562with properties of superior energy efficiency. 563The difference in value between the energy classes 564deserves further study to compare the results with other 565researches carried out in Italy and internationally. 566Starting from the models for the three cities, the unit 567value is calculated for a property in Classes A, D and G. 568In the second step, the gap between the identified 569prices is evaluated in relative and absolute terms. 570 571The market value of a home with the most common 572characteristics in the sample in Padua is 2,347 €/sqm for 573energy Class G, 2,719 €/sqm for Class D and 3,150 €/sqm 574for Class A. The values of homes with the most 575common characteristics in Mestre are 1,533 €/sqm for 576a home in Class G, 1,723 €/sqm for a home in Class D 577and 1,937 €/sqm if the home is classified in Class A. The 578values of homes in Bergamo are 1,408 €/sqm, 1,602 579€/sqm and 1,823 €/sqm for EPC energy performance 580Classes G, D and A, respectively (Fig. 3). 581The absolute difference between a property in Class G 582and a property in Class D is 372 €/sqm, 190 €/sqm and 583194 €/sqm for the cities of Padua, Mestre and Bergamo, 584respectively. The absolute differences correspond to 585very similar percentage increases: 15.84% for Padua, 58612.41% for Mestre and 13.77% for Bergamo (Fig. 4). 587On the other hand, if we consider the gap between the 588least energy-efficient Class G and the most energy-589journal valori e valutazioni No. 34 - 202359010591Table 7 - Regression models with dummy variables for energy characteristics592Regression models593Padua594Mestre595Bergamo596F: 57,018 597p value: <0,001 598R2: 0,509599F: 50,478 600p value: <0,001 601R2: 0,513602F: 25,365 603p value: <0,001 604R2: 0,349 605Predictors Xi606βi607p value608βi609p value610βi611p value612Costant6138.675614<0.0016157.708616<0.0016177.502618<0.001619Zone (OMI)620-0.235621<0.001622-623-624-0.0616250.002626Proximity to local transport infrastructure627-0.119628<0.001629-630-631-632-633High energy efficiency (Classes A and B)6340.316635<0.0016360.317637<0.0016380.231639<0.001640Medium energy efficiency (Classes C and D)6410.179642<0.0016430.179644<0.0016450.1116460.024647Type648-649-650-651-652-653-654Number of bathrooms655-656-6570.172658<0.0016590.286660<0.001661Gross floor area662-0.001663<0.001664-0.003665<0.001666-0.0026670.002668Maintenance status669-0.131670<0.001671-0.103672<0.001673-0.0826740.006675Figure 3 - Unit values of homes in Padua, Mestre and Bergamo 676according to energy class.677678679===== PAGE 9 (score 6, numeric-paren 0, table-words 0, var-words 5) =====680681The economic value of sustainability. 682Real estate market and energy performance of homes683efficient Class A (conservatively the first subclass A), 684the hedonic prices in absolute value are 802 €/sqm for 685Padua, 404 €/sqmfor Mestre and 414 €/sqm for 686Bergamo. In percentage terms, the deviations are 68734.18%, 26.36% and 29.43%, respectively for the three 688locations (Fig. 4). 689In the three cities, the shift from Class D 690to Class A shows a significant convergence in 691percentages, between 12.4% and 15.8%. There is a 692more pronounced convergence in the premium price 693between the values of Class G and those of Class A, but 694this convergence is around thirty percentage points. 695 696 697The results converge even when the energy variable is 698reported as a dummy variable. The value gaps 699highlighted in Figure 5 show the increasing value of the 700properties compared to the less efficient classes. 701More specifically, the gap between inefficient classes 702E, F and G and moderately-efficient properties classes 703C and D is 19.60% for the cities of Padua and Mestre 704and 11.74% for Bergamo. The results also converge for 705the value gap between the least energy-efficient 706properties and those in the highest classes (A and B), 707with percentages of 37.16% for Padua, 37.30% for 708Mestre and 25.99% for Bergamo. 709The results of the survey are in line with the academic 710research on the subject. In particular, the estimated 711premium prices are in line with the results of empirical 712studies on the Dutch (Brounen & Kok, 2011), Welsh 713(Fuerst et al., 2016) and Irish (Hyland et al., 2013) 714housing markets. The three studies report an average 715premium of 14% for properties in energy class A 716compared to D. 717Compared to the Italian studies, the research 718highlights empirical evidence that is consistent with 719some 720studies 721and 722contradictory 723with 724others. 725Manganelli et al. (2019) found similar results to those 726of this study. In the study of the Bari housing market, 727the analysis of values is carried out for both central and 728suburban areas. In the former case, the spread 729between Class A and Class G homes is 29.41%, while in 730suburban areas it rises to 41.2%. 731Less consistent, however, are the values obtained in 732studies carried out in other cities. Dell’Anna et al. 733(2019) and Bisello et al. (2020) studied the markets of 734Turin and Bolzano. In both cases, the value deviation 735is estimated in relation to Class G, the minimum energy 736class. In Turin, the value deviation from Class A is 7376.33%, analogous to the 6.3% found for the city of 738Bolzano. 739The difference between the deviations found can be 740attributed to two factors. The first is related to the 741dynamism of the property markets considered. Turin is 742a metropolitan city and Bolzano, despite being a 743medium-sized city, has a very active market with values 744of absolute importance in the country. As pointed out 745by Eichholtz et al. (2010), value gaps increase where the 746real estate market is characterised by less active trading 747and values. In addition, recent EU positions may have 748had an effect by further widening the gap of the 749property values. 750In other segments of the real estate market, a similar 751trend of value divergence seems to be confirmed. In 752the tertiary market, JLL (2023) finds value gaps of more 753than twenty percentage points between green and 754traditional buildings, exceeding previous international 755(Eichholtz et al., 2010) and national (Mangialardo et al., 7562019) estimates. Thus, even in segments other than 757residential, the trend appears to be one of increasing 758market polarisation between different levels of energy 759performance. 7606. CONCLUSIONS 761The energy transition and decarbonisation require a 762major transformation of the housing stock. The issue 763journal valori e valutazioni No. 34 - 202376411765Figure 5 - Change in value between properties with different 766energy performance in the cities of Padua, Mestre and Ber-767gamo (energy classification measured with dummy variables).768Figure 4 - Change in value between properties in energy 769 classes G and D and between properties in energy classes G 770and A in the cities of Padua, Mestre and Bergamo.771