Barometer methodology

What we count, what we exclude, where the data comes from and how we revise past figures.

Where the method comes from

We did not invent the indicator. Building permits are a recognised leading indicator: the U.S. Census Bureau has published the Building Permits Survey monthly since 1959, and the Conference Board uses it in its index of leading economic indicators.

In Europe, Eurostat publishes an index of building permits (tables sts_cobp_m and sts_cobp_q), with an index of dwellings and an index of useful floor area. We follow the same shape: an index on a base period, parallel series, a publication calendar and a revision rule.

The difference is the lag and the detail, not novelty. At country level the figure already exists, at INE and at Eurostat; and INE does publish, by municipality and every month, the dwellings licensed in new construction, with a lag of about 40 days. By type of works INE also publishes per municipality, but only once a year and with a lag of about six and a half months. What does not exist is that detail every month, nor with a lag of a few days. That is where this barometer adds something.

Publication calendar

New figures on the 8th of each month, covering through the previous month. The date is fixed and lives in the code, not in a promise: a series without a calendar does not get cited, because the writer does not know whether the number arrives in time.

Unit: the case

We count building-works cases opened at municipal councils, each case once, in the month of its reference event.

The reference event is declared council by council: where the dataset carries the date of submission, submission is what we use; where it carries only the year, that municipality does not enter the monthly series. Mixing submission with the decision in a single series would join two points of the process that are months apart.

Taxonomy: INE’s, not ours

Every council uses its own vocabulary. We map everything onto the three classes used by INE: new construction; extensions, alterations and reconstructions; and the total of building works.

Where the council does not state the type of works, the case counts towards the total and also appears as “unclassified” — the share is published next to the figure.

What is left out

Use permits and use notifications, legalisations of existing buildings, standalone demolitions, maintenance works, alterations during construction and licence renewals: either they produce no building, or they count the same thing twice.

Land subdivisions and plot detachments are left out too: they are acts concerning land, and the INE survey counts building works.

Prior information requests are counted separately: they are a question, not a commitment.

Seasonality

August and January drop every year. We apply no seasonal adjustment: with few municipalities and five years, estimating a model would mean fitting it, not measuring it.

Instead we publish three series: the raw month, the 12-month total and the change against the same month of the previous year.

Revisions: the series is kept as snapshots

Municipal datasets are rewritten retroactively: a case submitted in March may appear in the dataset in May, and a closed case may disappear.

Each collection therefore records a dated snapshot and does not correct the previous one. The figure published in September stays accessible in November, and the difference is seen as a difference between two snapshots.

The three most recent months are marked provisional. Older values that change between snapshots are declared as a revision.

Fixed panel and growing coverage

The number of municipalities will grow as more councils open their data. If the index were calculated “over all those there are today”, the arrival of a new municipality would show up as economic growth.

The index is therefore calculated over a fixed panel: only municipalities with a complete base year enter it. A new municipality publishes its level and its year-on-year change straight away, and enters the index at the next revision of the base, which is announced.

Checking against INE

INE publishes, by municipality, the buildings and dwellings licensed. It is the only external check this series has, and it has its own section on the barometer page.

The levels do not coincide, and should not: we count submissions, INE counts licences granted. What is compared is the direction and the size of the movement.

Known limitations

A licence is not a building. Coverage is partial: only the councils with usable open data.

Digital archives fill up over time: in the first years of a dataset there are fewer cases because there is less archive, not because less was being built.

Datasets that show only the current state — and not the history — do not enter the series, because the past would shrink on its own.

Sources and licence

In the downloadable CSV, every row carries the address of the municipal dataset it was counted from, and every official figure carries the exact address of the request to INE.

Our derived data is published under CC BY 4.0. The INE data belongs to INE, under CC BY 4.0.

The leading signal: ADENE energy pre-certificates

An energy pre-certificate is issued by ADENE (Energy Certification System, SCE) for every dwelling at the design stage. It is not a building licence and not a construction start: many projects reach a licence months later, some never. We use the context 'Projeto Edifícios Novos', type 'Habitação'.

Coverage: mainland Portugal and Madeira. ADENE does not publish the Azores in these statistics; we do not count them as zero. A municipality with no row in a covered month had zero pre-certificates that month.

'12 months' always means 12 consecutive months covered by ADENE files; if one is missing, the number is not published. If the latest month is more than 9 months old, the whole section disappears from the page.

The fastest growth is only announced among municipalities with over 50 thousand residents and at least 50 pre-certificates in the previous 12 months. The map uses its own steps per thousand residents: 1.5, 3, 6 and 12.

This series is never added to INE's licensed dwellings or divided by them: they are different stages of the same path. The origin of every row — the ADENE file and its sha256 — is in our CSV /api/v1/barometer/sce.csv. Source: ADENE — SCE.

Building-permit barometer by municipality

You may copy and reuse these figures as long as you credit the source with a link to flatin.pt. Better still, link to this page: that way the reader can check the number. Terms of use