1# RECORD 323b30d53cb25830c8397eeb95e569a5dc5a30442# TITLE: Sub-National Spatial Price Indexes for Housing: Methodological Issues and Computation for Italy3# YEAR: 2022.0 VENUE: Journal of Official Statistics DOI: 10.2478/jos-2022-00044# AUTHORS: Ilaria Benedetti; L. Biggeri; T. Laureti5# PAGES: 26 CANDIDATE TABLE PAGES (1-based): [10, 13, 14, 15, 16, 18]6# ABSTRACT: Abstract It is essential to measure within-country differences in housing costs in order to evaluate costs of living, assessing and comparing poverty levels, quantifying salaries and disposable income of families and finally for designing housing policies at local level. To the authors knowledge, no studies have yet been carried out on the computation of Space Price Indexes for Housing Rents (SPIHRs). In this article we computed preliminary estimates of sub-national SPIHRs by using hedonic regression model, which is an extension of the Country Product Dummy method, for all the Italian regions. The hedonic regression is generally used to obtain multilateral spatial indexes, thus allowing us to obtain multilateral SPIHRs for the Italian regions. The estimates have been done using 2017 data from the Real Estate Market Observatory which is a part of the Italian Agency of Revenue and Tax. This data source is the most comprehensive source of information on Italian houses price rents with a wide geographical coverage, including data for each Italian municipality. The obtained results show significant differences across the Italian regions, thus highlighting the importance of calculating SPIHR in Italy on a regular basis and the need to continue researches in this field.789===== PAGE 1 (score 5, numeric-paren 0, table-words 0, var-words 5) =====1011Sub-National Spatial Price Indexes for Housing:12Methodological Issues and Computation for Italy13Ilaria Benedetti1, Luigi Biggeri2, and Tiziana Laureti114It is essential to measure within-country differences in housing costs in order to evaluate costs15of living, assessing and comparing poverty levels, quantifying salaries and disposable income16of families and finally for designing housing policies at local level. To the authors knowledge,17no studies have yet been carried out on the computation of Space Price Indexes for Housing18Rents (SPIHRs). In this article we computed preliminary estimates of sub-national SPIHRs by19using hedonic regression model, which is an extension of the Country Product Dummy20method, for all the Italian regions. The hedonic regression is generally used to obtain21multilateral spatial indexes, thus allowing us to obtain multilateral SPIHRs for the Italian22regions. The estimates have been done using 2017 data from the Real Estate Market23Observatory which is a part of the Italian Agency of Revenue and Tax. This data source is the24most comprehensive source of information on Italian houses price rents with a wide25geographical coverage, including data for each Italian municipality. The obtained results26show significant differences across the Italian regions, thus highlighting the importance of27calculating SPIHR in Italy on a regular basis and the need to continue researches in this field.28Key words: Spatial rental price indexes; hedonic regression; country product dummy method;29symbolic data analysis.301.31Introduction32It is essential to measure within-country differences in housing costs in order to evaluate33costs of living, assessing and comparing poverty levels, quantifying salaries and34disposable income of families and finally for designing housing policies at a local level35(Bishop et al. 2017).36It is well known that when making comparisons, the sub-national (regional) values of37the economic indicators should be adjusted for sub-national price differences in order to38avoid misleading regional analyses and the consequent policy implications and outcomes,39which is done by computing sub-national Spatial Price Indexes (SPIs) as illustrated by40Biggeri et al. (2017) and Laureti and Rao (2018).41In this context, the computation of sub-national SPIs for housing costs is particularly42important since housing costs can be a substantial financial burden to households,43especially for low-income families.44These sub-national SPIs have proven to be useful for assessing and comparing poverty45levels and for implementing housing policies for the poor at local level, however they can46also be used as proxies of general sub-national household consumption SPIs or in47q Statistics Sweden481 University of Tuscia, Via del Paradiso 47, 01100, Viterbo (VT), Italy. Emails: i.benedetti@unitus.it and49laureti@unitus.it502 The University of Florence, Viale Morgagni, 59, 50134 Firenze, 50134, Italy. Email: Luigi.biggeri@unifi.it51Journal of Official Statistics, Vol. 38, No. 1, 2022, pp. 57–82, http://dx.doi.org/10.2478/JOS-2022-0004525354===== PAGE 10 (score 20, numeric-paren 0, table-words 4, var-words 2) =====5556quotations for all of the Italian municipalities and on the average quotations of the for the5720 Italian regions according to type of dwellings and conservation status. Quotations for58poorly conserved dwellings are very few to compute adequate descriptive statistics.59Table 1 shows the variability of the rental quotations observed for the 8,04660municipalities and across the Italian regions and how type of dwelling and conservation61status affect the distribution of the minimum and maximum rental prices. As illustrated in62Table 1, the highest rental value is observed for well-finished apartments, for which the63rental price ranges from 3.99 to 5.37 EUR/m2. In particular, the mean rental price ranges64from 4.90 to 6.37 EUR/m2 for excellently conserved apartments (the maximum rental65price shows as well the highest level of variability). Contrastingly, the lowest rental value66is observed for economic apartments, for which the rental prices range from 2.73 to 3.7267EUR/m2.68From the second part of the Table 1, we can see that also the variability of the average69rental values across regions is very high and this surely justify the importance to compute70the SPIHRs for the Italian Regions.71Table 1.72Descriptive statistics based on the quotations available for all the municipalities. Italy, 2017.73Conservation status74Total75Excellent76Normal77Type of dwelling78Min79Max80Min81Max82Min83Max84Detached85Mean863.60874.86884.72896.12903.36914.5892Standard deviation932.04942.78952.42963.39971.86982.5599Coefficent of variation1000.571010.571020.511030.551040.551050.56106Skewness1072.411082.651092.501102.851112.361122.45113Well-finished114Mean1153.991165.371174.901186.371193.831205.19121Standard deviation1222.091232.941242.541253.761261.961272.73128Coefficent of variation1290.521300.551310.521320.591330.511340.53135Skewness1362.391372.941382.891393.891402.131412.32142Economic143Mean1442.741453.731463.501474.381482.681493.68150Standard deviation1511.591522.131531.971542.531551.541562.08157Coefficent of variation1580.581590.571600.561610.581620.581630.57164Skewness1652.411662.251673.431683.121692.191702.09171Total172Mean1733.461744.671754.581765.911773.271784.47179Standard deviation1802.001812.731822.441833.461841.851852.53186Coefficent of variation1870.581880.591890.531900.591910.571920.57193Skewness1942.361952.681962.661973.261982.201992.31200Journal of Official Statistics20166202203204===== PAGE 11 (score 12, numeric-paren 0, table-words 2, var-words 1) =====2052065.207The Specification of the Hedonic Models and the Estimation Procedure208As regards the estimation strategy, it is worth noting that, as already mentioned in209Section 4, the OMI database provides minimum–maximum range of rental prices210according to OMI zones, type of dwelling and conservation status. Therefore, for each211observation two rental prices are specified: the minimum and the maximum rental value.212As a consequence our dependent variable in model (1) is an interval-valued variable which213is defined for a given set of N entities S ¼ {s1, : : : , sN}, by an application Y: S ! B such214that si 7! Y si215ð Þ, Y si216ð Þ ¼ li; ui217218219, B is the set of intervals of an underlying set O # R, In220other words, the value of the interval-valued variable Y for each si [ S is usually defined221by the lower and upper bounds li and ui of Ii ¼ Y(si), In order to estimate the hedonic222regression model [1] with rij ¼ lij; uij223224225, we used the Symbolic Data Analysis approach226(SDA) which provides a framework for the representation and analysis of data that227comprehends inherent variability (Brito 2014).228Several approaches have been introduced for estimating regression models with229symbolic interval data, Billard and Diday (2000) proposed a first approach which230consisted in fitting a regression line to the centres of the intervals or midpoints defined as231cij ¼ (lij þuij)/2. Therefore, model (1) can be expressed as:232lncij ¼233X234M235j¼1236ajAjCþ237X238K239k¼1240bkCikC þ 1ijC241ð2Þ242Table 2.243Descriptive statistics based on averages computed for the 20 Italian regions. Italy, 2017.244Conservation status245Total246Excellent247Normal248Detached249Min250Max251Min252Max253Min254Max255Min2562.512573.442583.002593.002602.502613.42262Max2636.042648.1426510.0926613.002676.602688.97269Standard Deviation2702.042712.782722.422733.392741.862752.55276Well-finished277Min2781.862792.432802.462813.042821.682831.23284Max2855.592867.462879.4128812.582895.242907.00291Standard Deviation2922.092932.942942.542953.762961.962972.73298Economic299Min3001.253011.673021.623032.183041.233051.66306Max3077.5030811.653096.303107.973116.303127.97313Standard Deviation3141.593152.133161.973172.533181.543192.08320Total321Min3221.643232.163242.513253.103262.513273.10328Max3295.293307.223319.1833212.283339.1833412.28335Standard Deviation3362.003372.733382.443393.463401.853412.53342Benedetti et al.: Sub-National Spatial Price Indexes for Housing34367344345346===== PAGE 13 (score 113, numeric-paren 92, table-words 1, var-words 0) =====347348apartments. In this way useful information may be provided to economic and social policy349makers. The results of the estimates are reported in Tables 3A and 3B respectively. The350mid-point included in model estimation (2) is an arithmetic mean of the min and max351rental values. Estimation results of model (2) computed with the geometric mean are not352significantly different from the estimated results obtained using the arithmetic mean.353Table 3A.354Hedonic regional models for all dwelling types with different territorial classifications, first semester355(january–june), year 2017.356Hedonic model with districts357Hedonic model with358homogeneous zones359Model360[2]361Model362[3]363Model364[2]365Model366[3]367Midpoint368Min369Max370Midpoint371Min372Max373Regions (ref. Italy)374North375Piemonte37620.07937720.08237820.07537920.06538020.07038120.061382(0.005)383(0.005)384(0.005)385(0.005)386(0.005)387(0.005)388Valle d’Aosta3890.2163900.2403910.1973920.2273930.2513940.209395(0.022)396(0.022)3970.023398(0.022)399(0.022)400(0.022)401Lombardia4020.1734030.1914040.1594050.1744060.1924070.161408(0.005)409(0.004)410(0.005)411(0.004)412(0.004)413(0.004)414Liguria4150.4264160.4364170.4194180.4234190.4334200.416421(0.008)422(0.008)423(0.008)4240.0074250.007426(0.008)427Trentino-4280.4124290.4224300.4034310.4154320.4254330.406434Alto Adige435(0.008)436(0.008)437(0.008)438(0.008)439(0.008)440(0.008)441Veneto4420.10244320.17644420.17444520.17644620.17644720.174448(0.009)449(0.009)4500.010451(0.009)452(0.009)453(0.009)454Friuli-Venezia4550.02045620.0224570.0484580.02845920.0144600.056461Giulia462(0.009)463(0.009)464(0.009)465(0.009)466(0.009)467(0.009)468Emilia-4690.2224700.2354710.2134720.2184730.2314740.209475Romagna476(0.007)477(0.007)478(0.007)479(0.007)480(0.007)481(0.007)482Centre483Toscana4840.4354850.4364860.4354870.4334880.4344890.432490(0.007)491(0.007)492(0.007)493(0.007)494(0.007)495(0.007)496Umbria49720.07149820.07449920.06850020.08050120.08350220.078503(0.013)504(0.013)505(0.013)506(0.013)507(0.013)508(0.013)509Marche5100.0245110.0215120.0265130.0245140.0225150.027516(0.008)517(0.008)518(0.008)519(0.008)520(0.008)521(0.008)522Lazio5230.3875240.3885250.3865260.3865270.3885280.385529(0.008)530(0.008)531(0.008)532(0.008)533(0.008)534(0.008)535Abruzzo53620.17953720.21053820.15753920.18254020.21254120.159542(0.008)543(0.008)544(0.008)545(0.008)546(0.008)547(0.008)548South and Islands549Molise55020.30855120.35155220.27655320.28455420.32855520.251556(0.023)557(0.023)558(0.023)559(0.023)560(0.023)561(0.023)562Campania56320.17456420.16856520.17956620.19556720.18856820.201569(0.005)570(0.005)571(0.005)572(0.005)573(0.005)574(0.005)575Puglia57620.12257720.10457820.13457920.15058020.13158120.163582(0.008)583(0.008)584(0.008)585(0.008)586(0.008)587(0.008)588Benedetti et al.: Sub-National Spatial Price Indexes for Housing58969590591592===== PAGE 14 (score 81, numeric-paren 60, table-words 3, var-words 1) =====593594Table 3A.595Continued596Hedonic model with districts597Hedonic model with598homogeneous zones599Model600[2]601Model602[3]603Model604[2]605Model606[3]607Basilicata60820.63960920.62361020.65161120.62661220.61161320.637614(0.012)615(0.012)616(0.012)617(0.012)618(0.012)619(0.012)620Calabria62120.36262220.35962320.36462420.37262520.36962620.374627(0.008)628(0.008)629(0.008)630(0.007)631(0.007)632(0.008)633Sicilia63420.33363520.34663620.32363720.33663820.34963920.326640(0.007)641(0.007)642(0.007)643(0.007)644(0.007)645(0.007)646Sardegna64720.14964820.14764920.15165020.13565120.13265220.136653(0.010)654(0.010)655(0.010)656(0.010)657(0.010)658(0.010)659Type of dwelling660(ref. well-finished apartment)661Economic66220.17566320.17666420.17466520.17566620.17666720.174668apartment669(0.004)670(0.004)671(0.004)672(0.004)673(0.004)674(0.004)675Detached6760.1256770.1306780.1216790.1306800.1366810.126682(0.004)683(0.004)684(0.004)685(0.004)686(0.004)687(0.004)688Status of the dwelling689(ref. Normal )690Excellent6910.1946920.2266930.1706940.1966950.2286960.172697(0.005)698(0.005)699(0.005)700(0.005)701(0.005)702(0.005)703District (ref. Central )704No705No706No707Semi-central7080.3267090.3167100.333711(0.007)712(0.007)713(0.007)714Outskirts7150.1167160.1147170.118718(0.005)719(0.005)720(0.005)721Suburban7220.0317230.0337240.030725(0.005)726(0.005)727(0.005)728Rural and72920.25573020.25573120.255732extra-urban733(0.005)734(0.005)735(0.005)736Homogeneous737zones738No739No740No741Yes742Yes743Yes744Intercept7451.0587461.0587471.3747481.1897491.0207501.332751(0.004)752(0.004)753(0.004)754(0.004)755(0.004)756(0.004)757Correlation7580.9757590.974760Breusch-Pagan test7615156576251467763Adjusted R27640.4637650.4787660.4787670.4857680.4717690.471770RMSE7710.3757720.3747730.3817740.3687750.3677760.372777AIC77847.719779267.210780267.21078145.462782269.667783269.667784BIC78547.977786266.644787266.64478845.969789268.564790268.564791N79254,21479354,21479454,21479554,21479654,21479754,214798Source: our elaboration from OMI data 2017.799Note: standard error in brackets; all the coefficients are significant at p , 0.01.800Journal of Official Statistics80170802803804===== PAGE 15 (score 139, numeric-paren 114, table-words 1, var-words 0) =====805806Table 3B.807Hedonic regional models for economic apartments with different territorial classifications. First808semester (january–june), year 2017.809Hedonic model with districts810Hedonic model with811homogeneous zones812Model813[2]814Model815[3]816Model817[3]818Model819[2]820Model821[3]822Model823[3]824Midpoint825Min826Max827Midpoint828Min829Max830Regions (ref. Italy)831North832Piemonte83320.04483420.06483520.02883620.03383720.05483820.016839(0.013)840(0.013)841(0.013)842(0.013)843(0.013)844(0.013)845Valle d’Aosta8460.1468470.1748480.1258490.1618500.1898510.141852(0.038)853(0.038)854(0.038)855(0.038)856(0.038)857(0.038)858Lombardia8590.1968600.2138610.1838620.1908630.2078640.178865(0.013)866(0.013)867(0.013)868(0.013)869(0.013)870(0.013)871Liguria8720.4528730.4458740.4578750.4508760.4438770.455878(0.016)879(0.016)880(0.016)881(0.016)882(0.016)883(0.016)884Trentino8850.6738860.6918870.6598880.6418890.6668900.623891Alto Adige892(0.177)893(0.179)894(0.178)895(0.175)896(0.175)897(0.175)898Veneto8990.2389000.2469010.1539020.2379030.2509040.158905(0.026)906(0.026)907(0.026)908(0.026)909(0.026)910(0.026)911Friuli Venezia9120.02091320.0279140.0529150.03191620.0179170.064918Giulia919(0.019)920(0.019)921(0.019)922(0.019)923(0.019)924(0.019)925Centre926Emilia-9270.2139280.2239290.2059300.2119310.2219320.203933Romagna934(0.016)935(0.016)936(0.016)937(0.016)938(0.016)939(0.016)940Toscana9410.3909420.3999430.3849440.3899450.3979460.382947(0.017)948(0.017)949(0.017)950(0.017)951(0.017)952(0.017)953Umbria95420.07095520.06595620.07495720.07695820.07195920.079960(0.027)961(0.027)962(0.027)963(0.027)964(0.027)965(0.027)966Marche9670.0149680.0129690.0169700.0169710.0139720.018973(0.016)974(0.016)975(0.016)976(0.016)977(0.016)978(0.016)979Lazio9800.4339810.4279820.4389830.4369840.4309850.441986(0.018)987(0.018)988(0.018)989(0.018)990(0.018)991(0.018)992Abruzzo99320.21799420.24999520.19399620.21799720.24999820.193999(0.016)1000(0.016)1001(0.016)1002(0.016)1003(0.016)1004(0.016)1005South and Islands1006Molise100720.360100820.402100920.329101020.339101120.382101220.3071013(0.035)1014(0.035)1015(0.035)1016(0.035)1017(0.035)1018(0.035)1019Campania102020.212102120.214102220.211102320.230102420.231102520.2301026(0.013)1027(0.013)1028(0.013)1029(0.013)1030(0.013)1031(0.013)1032Puglia103320.140103420.117103520.156103620.158103720.135103820.1751039(0.016)1040(0.016)1041(0.016)1042(0.016)1043(0.016)1044(0.016)1045Basilicata104620.734104720.720104820.743104920.724105020.712105120.7331052(0.020)1053(0.020)1054(0.020)1055(0.020)1056(0.020)1057(0.020)1058Calabria105920.368106020.363106120.371106220.373106320.368106420.3751065(0.015)1066(0.015)1067(0.015)1068(0.015)1069(0.015)1070(0.015)1071Sicilia107220.415107320.427107420.405107520.419107620.431107720.4091078(0.014)1079(0.014)1080(0.014)1081(0.014)1082(0.014)1083(0.014)1084Sardegna108520.215108620.212108720.218108820.193108920.190109020.1971091Benedetti et al.: Sub-National Spatial Price Indexes for Housing109271109310941095===== PAGE 16 (score 48, numeric-paren 30, table-words 5, var-words 0) =====10961097No evidence of functional form misspecification was observed from the Ramsey test1098(1969). The estimation results of models (2) and (3) at regional level for all the dwelling types1099are reported in Table 3A by considering two model specifications. The first three columns1100refer to districts, while the last three columns refer to homogeneous territorial zones.1101As regards the estimation results, the best fitting-models are obtained when homogeneous1102zones are included. Moreover, model (3) is preferable to model (2) as it allows for the1103correlation between 1ijL and 1ijU and the estimation of the full variance–covariance matrix1104of the coefficients. As shown in the last rows of Table 3, the correlation of the residuals in the1105minimum and maximum rent equations is equal to 0.975. We can reject the hypothesis that1106Table 3B.1107Continued1108Hedonic model with districts1109Hedonic model with1110homogeneous zones1111Model1112[2]1113Model1114[3]1115Model1116[3]1117Model1118[2]1119Model1120[3]1121Model1122[3]1123(0.020)1124(0.020)1125(0.020)1126(0.020)1127(0.020)1128(0.020)1129Status of the dwelling1130(ref. Normal)1131Excellent11320.19311330.24611340.15311350.19811360.25011370.1581138(0.012)1139(0.012)1140(0.012)1141(0.012)1142(0.012)1143(0.012)1144District (ref. Central)1145Semi-central11460.31211470.30511480.3171149(0.012)1150(0.012)1151(0.012)1152Outskirts11530.11711540.11611550.1171156(0.008)1157(0.008)1158(0.008)1159Suburban11600.00911610.01011620.0081163(0.009)1164(0.009)1165(0.009)1166Rural and116720.294116820.295116920.2931170extra-urban1171(0.008)1172(0.008)1173(0.008)1174Homogeneous1175zones1176No1177No1178No1179yes1180yes1181yes1182Intercept11831.07811840.90611851.22211861.04611870.87611881.1891189(0.011)1190(0.011)1191(0.011)1192(0.011)1193(0.011)1194(0.011)1195Correlation11960.97811970.97811980.97711990.9771200Breusch-Pagan1201test120215,345120315,345120415,322120515,3221206Adjusted R212070.45712080.46312090.44912100.47712110.48212120.4701213RMSE12140.37312150.37612160.37412170.36512180.36912190.3661220AIC122113.8731222222.2421223222.242122413.2751225222.8881226222.8881227BIC122814.0801229221.7811230221.781123113.6981232221.9581233221.9581234N123516,040123616,040123716,040123816,040123916,040124016,0401241Source: our elaboration from OMI data 2017.1242Note: standard error in brackets; all the coefficients are significant at p , 0.01.1243Journal of Official Statistics124472124512461247===== PAGE 17 (score 18, numeric-paren 0, table-words 7, var-words 4) =====12481249this correlation is zero on the basis of the Breush-Pagan test of independence results. The1250correlation is strong, so the efficiency gains with the SUR estimation is great (Cameron and1251Trivedi 2009).1252Table 3A shows different distributional patterns as rental prices vary greatly across the1253Italian regions, thus confirming territorial heterogeneity. The lowest rental prices were1254observed for Basilicata and Molise while the highest rental prices were observed in1255Trentino Alto-Adige and Liguria (both for midpoint and minimum- maximum models).1256As expected, the localization factor plays an important role in determining rental prices.1257However, it is interesting to note that the rents of houses located in central areas of the1258municipalities cost 11.4% – 11.8% less than the rents of dwellings on the outskirts as shown in1259Table 3A while houseslocatedinrural areascost 25.5% less thanhouseslocatedin centralareas1260of the municipalities. This effect may be due to the heterogeneity of houses included in the1261district classification which does not take into account the socio-economic characteristics of the1262territorial area. More specifically, the district variable captures an average effect of the1263locational characteristics since it is obtained by considering rental prices related to all the1264municipalities included in each region. Yet, Italy is characterized by the presence of numerous1265small municipalities in which the rent values are smaller than those observed in the big1266municipality and the more expensive houses (well-finished dwelling and detached houses) are1267located in semi-central and outskirts districts. Contrastingly, wehave found that Italian regional1268capitals, such as Florence (Tuscany), Rome (Lazio) or Milan (Lombardy), show a different1269behaviour of housing rental prices with more expensive houses located in the central districts.1270When considering a detailed specification for the localization factors (OMI zones), the1271effects of type-of-dwelling and property status do not vary: detached houses cost 13.6%1272and 12.6% more than rented well-finished apartments, while excellently conserved houses1273cost 22.8% and 17.2% more than those with normal conservation status.1274Table 3B reports hedonic regional model estimation results for economic apartments by1275using models (2) and (3) specifications. As regards the estimation results, the best fitting-1276models are obtained when homogeneous zones are included, and model (3) is preferable to1277model (2). The number of observations is equal to 16,040 and estimated results confirm1278high territorial heterogeneity across the Italian regions, highlighting disparities between1279the Northern and Southern regions. As regards the localization factors, the coefficients1280associated to the districts dummy continue to assume the same signs but are lower than the1281coefficients estimated in Table 3A for all the dwelling type thus confirming the influences1282of houses rental prices in small municipalities.1283Table 4 reports the SPHIRs obtained when homogeneous territorial zones are included1284in the models (2) and (3) if we consider the 20 Italian regions and compare every region to1285the overall mean (Italy ¼ 100). The SPHIRs are calculated both for all dwelling type and1286the economic apartments.1287On observing regional models reported in Table 4, our results highlight high territorial1288heterogeneity and confirming significant rental price differences across the Italian regions1289thus demonstrating that rents are higher in the Northern-Central regions than in the South in1290respect to the whole of Italy. Basilicata proved to be the less expensive region for rental1291properties while Tuscany, Liguria while the highest rents were found in Trentino-Alto1292Adige. Significant differences were observed in the estimated SPHIRs, when all of the1293dwelling type were included with the exception of the estimation for economic apartments.1294Benedetti et al.: Sub-National Spatial Price Indexes for Housing129573129612971298===== PAGE 18 (score 17, numeric-paren 0, table-words 2, var-words 2) =====12991300As regards economic apartments, the SPHIRs estimated for Trentino-Alto Adige are not1301significant due to the reduced number of observations. In this case, the SPHIRs estimated1302for Veneto and Lazio are higher than the SPHIRs estimated for all types of dwelling, while1303the SPHIRs estimated for Toscana, Abruzzo, Molise, Sicilia and Sardegna are lower than1304the SPHIRs estimated for all dwelling types. These findings may be influenced by two1305contrasting effects which emerge when estimating SPHIRs at regional level: the inclusion1306of tourist destinations characterized by higher rents and small municipalities with lower1307rents. Taking into account the various housing characteristics provided by the OMI data1308set, it is clear that the role played by tourism is stronger than the effect of the prevalence of1309low-cost houses with economic apartments.1310In order to explore the territorial heterogeneity of rents within each Italian region,1311Figures 1–4 show sub-national SPHIRs estimate with model (3) if we consider1312homogeneous territorial zones for the Italian regions (Italy ¼ 100).1313Figures 3 and 4 show sub-national SPHIRs with the corresponding intervals at 95%.1314Although most regions have small confidence intervals, few regions have overlapping1315confidence intervals. Overlapping confidence intervals were observed between Lombardia1316and Liguria SPIHRs and Calabria and Sicilia SPIHRs for the model estimated for all1317dwelling types and for the economic apartments. With the aim of making multiple1318Table 4.1319Regional SPIHRs (Italy ¼ 100), year 2017.1320All dwelling types1321Economic apartments1322Model [2]1323Model [3]1324Model [2]1325Model [3]1326Region1327Midpoint1328Min1329Max1330Midpoint1331Min1332Max1333North1334Piemonte133593.72133693.28133794.09133896.76133994.70134098.371341Valle d’Aosta1342125.481343128.581344123.261345117.521346120.781347115.181348Lombardia1349119.001350121.121351117.451352120.981353122.991354119.481355Liguria1356152.731357154.181358151.581359156.871360155.811361157.661362Trentino-Alto Adige1363151.391364152.951365150.061366Veneto1367110.09136883.82136984.021370126.691371128.451372117.141373Friuli-Venezia Giulia1374102.83137598.611376105.811377103.15137898.281379106.591380Emilia-Romagna1381124.371382126.001383123.231384123.431385124.761386122.521387Centre1388Toscana1389154.131390154.311391154.081392147.481393148.781394146.581395Umbria139692.29139792.01139892.53139992.66140093.14140192.371402Marche1403102.451404102.191405102.701406101.591407101.351408101.821409Lazio1410147.111411147.341412147.011413154.651414153.661415155.471416Abruzzo141783.36141880.89141985.26142080.49142177.94142282.451423South1424Molise142575.31142672.05142777.80142871.27142968.28143073.551431Campania143282.29143382.90143481.80143579.43143679.36143779.461438Puglia143986.09144087.72144184.94144285.38144387.33144483.981445Basilicata144653.47144754.30144852.89144948.48145049.09145148.061452Calabria145368.96145469.16145568.83145668.90145769.20145868.701459Sicilia146071.44146170.53146272.15146365.79146464.96146566.451466Sardegna146787.41146887.62146987.26147082.41147182.73147282.151473Source: our elaboration from OMI data 2017.1474Journal of Official Statistics147574147614771478===== PAGE 19 (score 0, numeric-paren 0, table-words 0, var-words 0) =====14791480Source: our elaboration from OMI data 2017 1481SPIHRs on minimum values1482Piemonte1483ta/Vallée d'Aoste1484Lombardia1485Trentino−Alto Adige/Südtirol1486Veneto1487Friuli−Venezia Giulia1488LiguriaEmilia−Romagna1489Toscana1490Umbria1491Marche1492Lazio1493Abruzzo1494Molise1495Campania Puglia1496Basilicata1497Calabria1498Sicilia1499Sardegna150040150160150280150310015041201505140150616015071801508200 401509601510801511100151212015131401514160151518015162001517SPIHRs on maximum values1518Piemonte1519ta/Vallée d'Aoste1520Lombardia1521Trentino−Alto Adige/Südtirol1522Veneto1523Friuli−Venezia Giulia1524LiguriaEmilia−Romagna1525Toscana1526Umbria1527Marche1528Lazio1529Abruzzo1530Molise1531Campania Puglia1532Basilicata1533Calabria1534Sicilia1535Sardegna1536Fg. 1.1537SPIHRs for Italian regions for all dwellings types, year 2017.1538Source: our elaboration from OMI data 2017 1539SPIHRs on minimum values154040154160154280154310015441201545140154616015471801548200 401549601550801551100155212015531401554160155518015562001557SPIHRs on maximum values1558Piemonte1559ta/Vallée d'Aoste1560Lombardia1561Trentino−Alto Adige/Südtirol1562Veneto1563Friuli−Venezia Giulia1564LiguriaEmilia−Romagna1565Toscana1566Umbria1567Marche1568Lazio1569Abruzzo1570Molise1571Campania Puglia1572Basilicata1573Calabria1574Sicilia1575Sardegna1576Piemonte1577ta/Vallée d'Aoste1578Lombardia1579Trentino−Alto Adige/Südtirol1580Veneto1581Friuli−Venezia Giulia1582LiguriaEmilia−Romagna1583Toscana1584Umbria1585Marche1586Lazio1587Abruzzo1588Molise1589Campania Puglia1590Basilicata1591Calabria1592Sicilia1593Sardegna1594Fig. 2.1595SPIHRs for Italian regions for economic apartments, year 2017.1596Benedetti et al.: Sub-National Spatial Price Indexes for Housing1597751598