Thread by @donmoyn

Quick thoughts on automation as a means to reduce #administrativeburden in public services. Automation using administrative data can make life easier for most, but those with eligibility characteristics not captured in admin data must apply manually. Creates a two-track system.

This paper examines provision of child benefits in Norway, which automate awards to those whose eligibility can be confirmed via administrative data (or what is sometimes called registry data in Europe). Great example of value of descriptive analysis. sciencedirect.com/science/articl…

Some observations:
*few systems can be fully automated, meaning there will be a need for manual application processes even if a govt wants to go toward automation
*this creates a source of inequality of treatment, even if most people are better off

*policymakers who want to automate more services should design simple programs with few eligibility conditions. Information about whether someone satisfies multiple eligibility requirements is less likely to be included in administrative data, preventing automation.

An example of the above point: US pandemic relief, a simple program where those under a certain $ level get checks. IRS automates based on tax records, so no application forms. (Exception is for those w v low income who don't submit tax returns, and must apply manually).

*Inequality between manual and automated processes can be compounded by correlation between conditions that push ppl into the manual lane and other forms of vulnerability e.g. in the Norwegian case, are those w non-traditional family structures also more likely to be low income?

*the efficiency gains of automation will be limited as long as the state has to maintain both an automated and manual system
*with more automation, fewer street-level bureaucrats, more people observing automated and reviewing processes

*Emergence of two distinct means of processing claims, automated and manual, creates an obligation for the state to track and transparently compare how people fare depending on which system they experience. Not sure that is on anyone's radar.

Automation of state services is a big topic, and can means lots of different things (e.g., automation of algorithms, decision processes) and lots of prior evidence it can increase inequality and be used to exclude people, e.g.,

https://twitter.com/HannahLebovits/status/1343628057916092423?s=20

Automation can also be used in ways that increases public trust in outcomes. See, e.g. blacks are more confident in automated traffic violation decisions. academic.oup.com/jpart/advance-…

Mostly, I'm just making the point that the benefits and risks of automation but are intrinsically tied to the quality of administrative data, the nature of the population served, and program design, which are ultimately all political decisions.

Nods

https://twitter.com/mkfeeneywyo/status/1343649377353535488?s=21

https://twitter.com/mkfeeneyWYO/status/1343649377353535488