Thread by @ValentinaBarca_

🧵 alert! "Enrollment and eligibility process of #Brazil’s #COVID19 Auxílio Emergencial ('AE'): #Data processing and use of administrative #registries" -- documents1.worldbank.org/curated/en/099… A few insights from @WBG_SPLabor @FCDOInclusive publication (1/16)

Authors' opinion in short, "in an experiment that involved the use of more than forty databases, Auxílio Emergencial makes it clear that trustworthy information is vital for identifying more vulnerable groups and their needs".. Down to some details: (2/N)

1. AE took different forms over its 3 phases, becoming increasingly more stringent in terms of eligibility and transfer values (NOTE: this is a different trajectory to other countries) - raising important doubts on 'impacts of restrictive alterations on vulnerable segments' (3/N)

2. The first 2 data flows for registration came from Cadastro Unico (Brazil's Social Registry): a) 14 Mill Beneficiaries of Bolsa Familia were automatially 'upgraded' to AE, b) 24 Mill people already registered but not receiving were assessed against AE criteria. (4/N)

The 3rd data flow came from digital (mobile phone app, 115 Mill downloads), simplified, registration, leading to the selection of further 38 Mill people (out of 57 Mill applications) --> Data added to new register 'ExtraCad' (5/N)

This is how those 3 flows compare against each-other: majority via ExtraCad new digital registration, then Bolsa Familia (PBF) bens, then those automatically enrolled via existing Cadastro data (with % varying for AE1-AE3) (6/N)

3. Did people struggle to access digital 'on-demand' registration? Yes! Especially 21% Brazilians with no internet access (up to 43% in low income brackets). Partly addressed via well-tested 'active outreach' strategies + the automatic enrolment using Cadastro data (above) (7/N)

...but not sufficiently supported by Brazil's (great) decentralised network of social welfare, who 'mobilised in improvised ways' with no interactions with Federal Government! (more analysis on that here openknowledge.worldbank.org/handle/10986/3…) (8/N)

..in fact, instead of leveraging tried and tested in-person capacity, an 'improvised' new channel of registration was created via Post Offices in June.. registering an under-whelming 14500 applications, of which 11,312 were eligible. (9/N)

4. Phases two and three of AE did NOT enable new applications (via app etc)! "Keeping a database of applicants closed to new registrations through both the mobile app and Cadastro Unico is one of the controversial points of the Programme", especially given shifts in needs.. 10/N

5. "All Auxílio Emergencial applications were run through a #verification process, conducted by cross-checking the information provided by applicants against government databases" - this built on previous Cadastro/Bolsa Familia practice - but pushed further.. (11/N)

..and crucially relied on the National Social Information Registry (CNIS) (run by Dataprev, government-run ICT company): information aggregator from > 20 admin registers logically interconnected, that gathers 35 billion records including personal and work data (12/N)

..this validation process (via an Analytical Database created in 5 days) ended up consolidating across 42 databases & led to cancellation of 5% of the 68.3 million bens in AE1 phase, most of them (51%) originating from the ExtraCad registry (13/N)

Individual Tax Ns from the Taxpayer Registry (CPF) ended up playing a crucial role in this process, even though it is not a 'national #ID' document & poses risks of exclusion... (with accompanying efforts to extend its coverage -more analysis on implications welcome, see Box!)..

..and key weakness of this process was lack of adequate and precise notification to those whose benefits were denied this way - clashing with right to #accountability, #privacy and personal data management (15/N)

6. Would advise reading Sections 4 and 5 on 'limits & possibilities' and 'lessons learned' if you have got this far in this digesting process :) - snippet here! (16/16)