Workshop on Integrating Geospatial and Statistical Standards

Session 3: Challenges and Solutions for Creating Geospatial Statistical Outputs

Location analytics in administrative data to produce House Price Statistics in

FRANCISCO VALA

INÊS FONTES | ELSA SOARES | FRANCISCO CALDEIRA

Stockholm, 6-8 November 2017 AGENDA

Motivation for disseminating House Prices 1 Statistics at Local Level

Administrative data sources in use and options 2 defined for location analytics

3 Indicators and products disseminated

4 Lessons learnt and critical issues Motivation

• Among the statistical gaps highlighted by the financial and economic crisis, real estate price statistics have been considered one of the areas to be urgently developed to provide appropriate indicators for both residential and commercial property prices. • Statistics Portugal is already addressing these two topics of real estate prices by taking advantage of administrative data: • House Price Index (HPI) is released quarterly since July 2014 (EC Regulation 93/2013) • Commercial Property Price Index (CPPI) is released annually since June 2017 Motivation

• However, there are significant regional differences in house price developments due to differentiating territorial assets:

• within EU countries (e.g. capital cities, touristic areas)

• within regions (e.g. city centres, coastal areas) • House prices statistics at local level aims to provide new quarterly official information to: • monitor territorial differentiation of house prices dynamics

• explore house prices within cities based on a Web GIS application Administrative sources • Statistics on house prices at local level are based on tax data provided by the Portuguese Tax and Customs Authority (AT) → agreement signed with Statistics Portugal (same framework as for HPI and CPPI) • The project is based on data linkage between information from: • Municipal Property Transfer Tax: from which the transaction prices are obtained → flow data • Municipal Property Tax: from which identifying characteristics of the transacted dwelling are obtained → stock data • The link between these two sources is made at the dwelling level Administrative sources Municipal STATISTICS PORTUGAL SDI Property Energy BGRI Blocks (polygons) Transfer Tax Certification BGE Buildings (points) (IMT) (…) FNA Household register (address) Transaction X,Y prices

Administrative boundaries Municipal Property Tax Statistical City (IMI) Characteristics of the dwelling Statistical GRID 500 x 500 m (…) LAU X,Y Statistical section (BGRI) Address Validation for location analytics X,Y Domain rules • With lack of information: 6.6% • Incorrect geographic coordinates (different reference systems and other errors): 0.01%

X,Y Consistency rules: Local Administrative Units (LAU) • LAU2 different but LAU1 equal: 1.11% → keep location based on X,Y • LAU1 different: 0.07% → keep location based on LAU (refuse X,Y)

LAU / City Complete transactions coverage Imputation

GRID and Statistical section Completeness above 90% Dissemination for location analytics • Average prices per square meter vs Median prices per square meter • While average prices are more common, median prices allow to remove the effect of extreme prices which may be particularly relevant at local level statistics. • Reference period for quarterly dissemination • Although the results are interpreted on a quarterly basis, they reflect the sales within a 12-month period, which reduces the impact of irregularities associated with the heterogeneity of the dwellings sold and eliminates the effect of potential seasonal fluctuations. • Dissemination thresholds • For each territorial unit, a minimum number of 15 transactions is considered for the four quarters ending in the reference quarter. In the case of data by subsection and grid the minimum number is 7. Indicators and products Indicators available at Statistics Portugal Official Website The quarterly results for the period between the first quarter of 2016 and the second quarter of 2017 are available at www.ine.pt • For the country, for the three NUTS levels and also for the municipalities.

Median value per m2 of dwellings sales (€) by Geographic localization (NUTS - 2013); Quarterly • For the country and up to the NUTS 3 level regions. In the case of Metropolitan Areas (Lisboa, Porto) and Algarve the values by LAU1 and 2 are also released.

Median value per m2 of dwellings sales (€) by Geographic localization and Category of housing unit; Quarterly Median value per m2 of dwellings sales in flats (€) by Geographic localization; Quarterly Median value per m2 of dwellings sales in existing flats (€) by Geographic localization; Quarterly • For the 7 cities with more than 100 thousand inhabitants (Lisboa, Porto, Vila Nova de Gaia, Amadora, Braga, Funchal and Coimbra). For Lisboa and Porto, data is also released by LAU 2.

Median value per m2 of dwellings sales (€) by Geographic localization and Category of housing unit; Quarterly Median value per m2 of dwellings sales in flats (€) by Geographic localization; Quarterly Median value per m2 of dwellings sales in existing flats (€) by Geographic localization; Quarterly Indicators and products Press release 31 October 2017 Indicators and products 41 out of the 308 Portuguese municipalities scored house prices above national value Median value per m² of dwellings sales, Portugal, NUTS III and municipality

Algarve Alcoutim Loulé A. M. Lisboa Moita Lisboa R. A. Madeira Santana Funchal PORTUGALPORTUGAL Fª .C.Rodrigo Lisboa A. M. Porto S. J. da Madeira Porto Alentejo Litoral Alcaçer do Sal Grândola Oeste Nazaré

Região de Coimbra P. da Serra Coimbra Região de Aveiro Anadia Aveiro Alto Minho Melgaço Caminha Alentejo Central Portel Évora Ave M. de Basto V.N. Famalicão Cávado T. de Bouro Esposende R. A. Açores Sª C. Graciosa Ponta Delgada Região de Leiria C. Pêra Leiria Viseu Dão Lafões P. Castelo Viseu 2 Lezíria do Tejo Alpiarça Benavente €/m Frequencies ] 896 ; 2 231 ] Municipalities Tâmega e Sousa Cinfães Penafiel PT ] 600 ; 896 ] Médio Tejo Mação Ourém ] 400 ; 600 ] Douro Sernancelhe Vila Real ] 106 ; 400 ] Alto Tâmega Boticas Chaves Not applicable 41 79 98 85 T.Trás-os-Montes Vimioso Bragança Territorial Limits Municipality Beira Baixa Penamacor Castelo Branco NUTS III Baixo Alentejo Mértola Beja Alto Alentejo Crato Portalegre Beiras e S. Fª .C.Rodrigo Guarda

- 500 0 0 500 1 000 1 500 2 000 €2 /500 m² 0 50 km Highest municipal value Lowest municipal value NUTS III Indicators and products

The price of existing dwellings in Lisboa (2 146 €/m2) surpassed the prices of new dwellings in the remaining municipalities of Área Metropolitana de Lisboa Median value per m² of dwellings sales by category of housing unit of Área Metropolitana de Lisboa, by municipality

Category of housing unit Total

Lisboa

Cascais

Oeiras

A.M.A.M.LISBOA Lisboa Mafra V. F. Xira

Almada Loures Alcochete Sesimbra Odivelas

Mafra Amadora Lisboa Alcochete Montijo Amadora Oeiras Montijo V. F. Xira

Montijo Almada Moita

Sintra Barreiro Palmela Seixal Seixal

Setúbal €/m2 ] 1 228 ; 2 231 ] Setúbal Frequencies Palmela AML Municipalities ] 1 000 ; 1 228 ] Sesimbra Barreiro ] 800 ; 1 000 ] [ 588 ; 800 ] Moita Territorial Limits Municipality 0 500 1 000 1 500 2 000 2 500 3 000 € / m² 3 6 5 4 NUTS III New Existing Total 0 5 km Indicators and products

The parishes of Santo António (3 294 €/m2) and Misericórdia (3 244 €/m2) scored the highest dwellings prices and the highest year-on-year rates of change in the city of Lisboa

Median value and year-on-year rate of change of median value per m² of dwellings sales, Lisboa and Median value per m² of dwellings sales, Lisboa and parishes parishes 50 LISBOA = 2 231 € / m² Sto. António

Sta. 40 Clara Misericórdia Parque S. Vicente das

30 Nações Q2017 (%) Q2017

nd Penha de França Belém S. Alcântara Domingos Estrela de 20 Alvalade LISBOA= 15.1% Benfica Penha Benfica Beato Olivais S. Domingos de Arroios de Areeiro Campo de Sto. França

dwellings sales, 2 sales, dwellings Benfica Sta. Maria Maior

10 Ourique António year growth rate of median value per m² of perm² value medianof rate growth year - Avenidas Novas S. Vicente Sta. Clara Carnide Ajuda Alcântara Sta. on Parque das Nações - Lumiar Misericórdia Maria Campolide Estrela Maior

Year Marvila Frequencies 0 2 Belém Parishes €/m 6 6 6 6 0 500 1 000 1 500 2 000 2 500 3 000 3 500 ] 2 500 ; 3 294 ] nd ] 2 231 ; 2 500 ] Median value per m² of dwellings sales, 2 Q2017 (€ / m²) Lisboa ] 1 800 ; 2 231 ] [ 1 500 ; 1 800 ] Territorial Limits 6 6 6 6

Parish 0 1 km Municipality Indicators and products

Web application allows the interactive search of median price on dwellings sales (€/m2) for the cities of Lisboa and Porto, for territorial units defined by the user based on the statistical section or a 500m x 500m grid, facilitating the analysis of selling prices charged in the different areas of each one of the cities. Indicators and products

Web application allows the interactive 5 250 No. 5 250 No. 5 250 No.

4 500 4 500 4 500 4 781 4 781 4 781 search of median 4 559 4 559 4 559 price on dwellings sales (€/m2) for the 3 750 3 750 3 750 cities of Lisboa and 3 000 3 000 3 000 Porto, for territorial units defined by the 2 250 2 250 2 250 user based on the 1 500 1 500 1 500 statistical section or a Total Total Total 500m x 500m grid, 750 750 750 facilitating the 0 0 0 analysis of selling Unique visitors Unique visitorsUniqueNumber visitors of visits Number of visitsNumber of visits prices charged in the different areas of each one of the cities. Impact: social media and news

Numero de Dia visitas Páginas Hits Bytes 31-Oct-17 561 8.459 12.665 665.19 MB 01-Nov-17 3.304 58.398 84.093 3.90 GB 02-Nov-17 114 2.253 3.297 166.39 MB 03-Nov-17 522 15.844 22.672 1.10 GB 04-Nov-17 271 5.759 7.965 362.27 MB Lessons learnt and critical issues

Territorial statistics and spatial data allows for new dimensions of analysis and insights, that tend to be relevant for business, public policies and the general public. 1. Institutional arrangements and technical cooperation? 2. Maintenance of quality standards of (external) input data and data availability? 3. Share infrastructural (geospatial) data and agree on a single system of codification across Public Administration? 4. Stand for differentiated ‘fit for purpose’ quality assurance or a parallel set of experimental statistics? Workshop on Integrating Geospatial and Statistical Standards

Session 3: Challenges and Solutions for Creating Geospatial Statistical Outputs

Location analytics in administrative data to produce House Price Statistics in Portugal

FRANCISCO VALA

INÊS FONTES | ELSA SOARES | FRANCISCO CALDEIRA

Stockholm, 6-8 November 2017