The source

This investigation draws from a single source: the Malta Planning Authority's public database. Every planning application submitted in Malta and Gozo since 1993 is published there — case number, location, description, applicant, architect, dates, recommendations, and decisions.

How the data was collected

The PA's public search interface was scraped and the results stored in a structured SQLite database. The result: 292,853 cases spanning three decades.

Each record contains the fields the PA publishes:

Case number and type (PA, DN, RG, PC)
Description of proposed works
Location and locality
Applicant and architect names
Filing, validation, and decision dates
Case officer recommendation (where available)
Board decision and vote counts
Geographic coordinates (where available)
Case category (e.g. Outside Development Zone)

Case numbers are automatically linked to the PA's website throughout this site. Click any case number (e.g. PA/00235/26) to see the original record, documents, and full decision history.

If you want to see the dataset directly, the interactive map plots every application with usable coordinates between 1993 and 2026.

How the analysis was done

Every finding was produced by querying the database directly. No models, no estimates, no projections. When a finding states "92.2% approval rate," that is a count of granted decisions divided by total decisions — not a statistical estimate.

Some categorisations require keyword analysis. Where the PA does not provide a structured field (e.g. "is this a demolition?"), the description of works was searched for relevant terms. This introduces some imprecision — not every demolition will contain the word "demolition," and some matches may be false positives. Findings note where this approach was used.

Board hearing minutes

Several stories analyse what happens inside the Planning Board — voting patterns, hearing duration, objector attendance, and commission composition. This comes from the PA's published board minutes: 69,254 hearing entries covering approximately 49,000 unique cases.

This is a large sample but not a complete record. Not every case goes to a board hearing — Development Notifications (DN) and Planning Commission (PC) cases are decided without one. For PA cases that do go to board, coverage ranges from 30–67% of decided cases in recent years (2014–2025), with earlier years less complete.

Is this sample biased? A comparison of approval rates shows it is not. Cases with board minutes: 88.2% approval. Cases without: 88.0%. The missing cases are not systematically different from the recorded ones — the sample is incomplete but unbiased.

Where stories cite aggregate board statistics, the sample size is noted. Individual case citations (specific vote counts, named attendees) reference verified records from specific hearings.

What this data cannot tell you

Approvals ≠ construction. A granted permission means the board said yes. It does not mean the building was built, or built as approved. A refusal may have been overturned on appeal or resubmitted successfully.
Categories shift over time. The PA's classification system has changed. The DN category did not exist before the early 2000s. ODZ boundaries have been redrawn through Local Plan revisions. Cross-era comparisons account for these changes where noted.
Keyword searches are imprecise. Searching descriptions for "pool" or "pre-1967" captures most relevant cases but not all, and may include some irrelevant ones.
Not every case has coordinates. Geographic data is more complete for recent applications. Older cases (pre-2000) are less likely to include location geometry.
Board minutes are partial. Hearing records exist for roughly a third of PA decisions in recent years. Findings based on board data state their sample size.

Verification

Every statistic was checked against the database. Where numerical claims in earlier drafts were found to be incorrect during verification, they were corrected. Case studies were checked against the PA's live website.

If you find an error, use the Report a correction link at the bottom of any story to flag it. Factual corrections are made promptly.