Systematic exporters were another 21% more productive than latent exporters after controls.
Source
World Bank — https://blogs.worldbank.org/en/endpovertyinsouthasia/global-integration-can-spur-productivity-growth-pakistan. Verified in source set / retroactive normalization pending.
What it proves
Supports learning-by-exporting: integration can itself raise productivity.
What it does not prove: Controls remove observable differences, not unobserved ones: management quality, family capital and buyer relationships plausibly drive both sustained exporting and the 21% premium, so part of what looks like learning-by-exporting is still selection into staying an exporter. Note also that 'systematic' versus 'latent' classifies the regularity of exporting, not its volume — the premium attaches to consistency, so an occasional large shipment does not buy it.
Memory hook: 21% → staying in export markets keeps teaching, it doesn't only select
Rebuts the claim: 'Exporting only reflects productivity, it never creates it, so trade openness is a sideshow' — the premium that survives after controls says integration itself is a productivity channel, not merely a sorting mechanism.