Technical Notes From the Field: Pandemic Unemployment Assistance
Tens of millions of Americans have lost their jobs in the past month. Not only have local businesses closed, but countless others who work as part of the “gig economy” have lost their sources of income because of the COVID-19 crisis. Overall, over 26M people have made unemployment benefit claims since the crisis began, and those claims are largely from people who are eligible for traditional unemployment. It is expected that a wave of applications from gig workers is yet to come.
The CARES Act, passed March 27th, expands benefits for those who have lost work due to the pandemic, namely:
- An extra $600 per week for traditional Unemployment Insurance (UI)
- Pandemic Unemployment Assistance (PUA), also $600 per week, for those who were self-employed or worked in the gig-economy and now have lost their jobs or have reduced hours.
The unprecedented applicant load, as well as the new rules changes have sent state departments of labor nationwide scrambling, as they rush to handle the new applicants and implement the changes.
The human impact of the delays is immense. As of April 23rd 2020, 26M people have applied for unemployment insurance since mid-March. The latest data on accepted claims, from April 11, roughly 71% of claims are accepted. Assuming, approximately, that those rejected from UI are eligible for PUA, that’s around 7.5M individuals without benefits with an unknown number still left to apply.
The US Digital Response (USDR) has a unique perspective on the Unemployment Insurance rollout because we have placed technical volunteers on Department of Labor teams, helping states troubleshoot and extend their systems. USDR Volunteers are mostly technical ranging from Site Reliability Engineers (SRE) and COBOL engineers to Designers.
From their collective experiences, we have a unique vantage point to see technical trends as states extend their systems as fast as they can to meet the new requirements. Many states have similar technology stacks, and therefore similar issues. Most have a mainframe that stores and processes the applications and payouts, and have added a web application on top of it to support applications and status tracking.
Here are a few trends we see happening in the wild.
Breaking the Mainframe Monolith
Have you ever wondered why some states require the “apply twice” for unemployment insurance? Apply twice means people have to apply for Unemployment Insurance, get rejected and only then can apply for PUA. This seems inefficient, and double workload for staff! This doesn’t necessarily happen for nefarious reasons, like to apply friction to the application process. In fact, this can happen for a simpler reason: it’s a requirement for PUA that you be ineligible for normal Unemployment Insurance, and in some systems there is no way to check other than to submit an application to a legacy COBOL based system and see if it rejects it.
Obviously to many the “apply twice” is unacceptable, it’s slow, requires manual work, and so the changing environment yields positive technical change. The requirement to accept PUA is forcing local governments to decompose these legacy systems into smaller parts- breaking the monolith, so that they can, for example, check eligibility without submitting a whole application and waiting for a rejection. This allows for states to upgrade their systems in parts and modernize incrementally, similar to the workflow used by technology companies with legacy infrastructure.
Mainframe shift: Online to batch mode
Have you ever wondered why unemployment insurance websites shut down at night? Seems ridiculous that a website needs to “rest” at night, right? Well the origin of this strange phenomenon also has a basis in the change required of the mainframe that powers this system. During these high-traffic times, the mainframe often struggles to handle the traffic and shifts to an online mode (process applications as they come in) to a batch mode (take a big batch of applications and run them all at once). The outage is a way to allow the system to “catch up” a bit before accepting more applications.
Though the nightly outage is annoying, it does have a silver lining — by separating these processes, it allows states to choose from many types of software that can interface with the mainframe, including, but not limited to, open source. There are few modern applications that can interface with mainframes, due to the on-premise deployments and a host of other factors. Tons of applications can create data files and share them on a file system — giving governments more options and reducing vendor lock in.
First time for the cloud
Ever wondered who hosts their own web apps? Before the crisis, unemployment was at record lows, and unemployment assistance websites had low traffic. The sudden surge of traffic brought several state UI websites hard down. In response to a crisis, the first entry point for states was a Content Delivery Network (CDN) that reduced the load on their servers and served as an entry to the cloud.
After the CDN experience, and seeing the need for building up an alternate data pipeline to handle the new traffic, many states have started to experiment with cloud services like Amazon Web Services, Google Cloud Platform and Azure. The crisis is catalyzing movement to the cloud with some hybrid elements, for example web applications, chat bots, waiting rooms and call center software move piecewise to the cloud — but payout and mainframe remain on premise. Moving software to the cloud has been one of the most impactful transformations for
Login, login, login
Q: When is it more desirable to apply by fax than online? A: When you are locked out of your account and can’t get back in!
Login and auth is so fundamental to applications, and many states have hard-to-use account systems that you must register with to start or process an unemployment assistance application. Between lost passwords, lockouts, outages, accounts not synced and general usability problems login technical debt causes spikes in call center volume, fax/mail applications.
A huge spike in applications, which forces states to bring on 10x the number of call center agents has resulted in some states rethinking their websites logins. Some, like Rhode Island, have decided to skip the login all together. Others are upgrading and adding chatbots and password recovery flows to help ease the pain.
We applaud the hard work that dedicated staff at Departments of Labor nationwide have been doing to provide relief to the public. The speed and scale of the crisis make the environment extremely challenging. Overall we have the following recommendations for states to consider while working through their implementation:
- This crisis is a good time to do a real evaluation of how valuable your legacy infrastructure is to you. 10 times as many people will go through your system in one month than you have had in the past year. If there was ever a time to consider alternatives, it is now, as current systems will not support these new requirements.
- Simplify, simplify, simplify. A complicated form with tons of checks and fields will be hard to use and hard to maintain. We have a sample simplified form that can serve as inspiration.
- Have a backup plan. A vendor or IT consultant may have provided a project plan, but plan to do due diligence on at least one alternative, as projects have a high failure rate.
We have consulted on several PUA implementations and are happy to answer any questions. Email us at firstname.lastname@example.org