character menu › codex › Revisiting: Brazil's airline network, before and after

◆  Revisiting: Brazil's airline network, before and after  ◆

2026-05-27 · 2 min read
#networks · #aviation · #retrospective

In August 2021 I published Brazilian Airline Network Evaluation - a short analysis using ANAC's open flight data that compared Brazil's domestic airline graph in 2019 (pre-pandemic) and 2020 (first pandemic year). I computed betweenness and closeness centrality for both years and found that 2020 wasn't really a contraction so much as a concentration: routes funneled into a smaller number of hubs while secondary airports lost their peripheral connections.

That finding was real. The post that contained it was incomplete in a way I want to flag now.

What the 2-year snapshot missed

Comparing 2019 to 2020 gives you a "before vs collapse" picture. It does not give you the part that turned out to be more interesting: the recovery. ANAC's open data portal still publishes the same monthly flight data, and the series now runs from 2019 through 2025. That's six years - pre-pandemic baseline, 2020 collapse, 2021-2022 partial recovery, 2023 return to scheduled capacity, 2024-2025 new normal.

The new normal does not match the old baseline. From what I've seen in the data:

  • The hub concentration that 2020 forced did not unwind. Secondary airports never quite regained their pre-2019 share of routes; the network is structurally more hub-and-spoke than it was.
  • Several low-volume routes that paused in 2020 never restarted. Some city pairs that had multiple daily flights in 2019 now have zero or one.
  • The relative importance of São Paulo as a connecting hub rose and stayed risen.

This isn't a pandemic story anymore. It's an industry-consolidation story that the pandemic just accelerated.

The shape of the lesson

The original Medium post is fine as a 2021 snapshot. The mistake was treating it as a finished answer. Two-year before/after is good for a class project; for a real claim about Brazilian aviation, the full series matters and the snapshot was always going to be misleading. That's the lesson I took away from re-reading it five years on - not the analysis itself, the framing.