Resources
Blog
Four Recommendations to Address Race and Inequity in Identity Fraud Detection

Four Recommendations to Address Race and Inequity in Identity Fraud Detection

States across the country have seen dramatic increases in fraudulent unemployment insurance claims being filed using stolen identities.

Partner:

States across the country have seen dramatic increases in fraudulent unemployment insurance claims being filed using stolen identities. Since the Coronavirus Aid, Relief, and Economic Security (CARES) Act was passed last spring, more than $36 billion of unemployment benefits has been lost to improper payments, with the majority being from fraud. To improve fraud detection going forward, the Continued Assistance for Unemployment Workers Act, passed in December, includes a new requirement for states to implement a process to verify the applicant’s identity.

Data shows that people of color are more likely to be flagged for identity fraud, even though there is no evidence that they are more likely to commit identity fraud. As states work to make sure assistance ends up in the hands of those who need it most, we must also pay attention to areas where we can reduce — not broaden — racial inequities. USDR’s Unemployment Insurance team works closely with states to ensure the successful delivery of these critical benefits. Here are USDR’s recommendations based on this work for updating fraud detection processes to support vulnerable and minority communities.

1. Make sure the system works for all names.

Many benefit systems were built with English names in mind, which does not reflect the linguistic and cultural diversity of the United States. Systems need to be able to handle:

  • Short first and family names without error
  • Long first and family names without truncation
  • Non-Roman characters (spaces, hyphens, apostrophes, accents) without error

The identity fraud detection process also needs flexibility in how it interprets and cross-references names across parts of the application. This includes accounting for mistakes that the system made with someone’s name (e.g., truncated after 10 characters in part A but truncated after 15 characters in part B), as well as for cultural considerations (e.g., someone might have “Graciela” on their birth certificate but “Grace” on their driver’s license).

2. Update the threshold for claims filed from the same address.

One common fraud flag is multiple people filing for benefits from the same address. People of color and families with immigrant members are more likely to live in larger, multigenerational households, and the addresses of homeless shelters and other social service agencies can be used in the absence of permanent housing. The systems must take into account these types of factors when setting thresholds for flagging claims.

Agencies can go further to improve the address verification process by:

  • Looking at all parts of the address (e.g., including apartment or floor number) when comparing addresses to avoid disproportionately targeting multi-family housing units
  • Proactively “clearing” addresses known to be used by those who are housing insecure and may be likely to file multiple claims

3. Provide a clear alternative to digital document verification.

Many identity verification solutions require applicants to provide photo IDs to prove their identity. This can cause many issues and, at times, a complete break in the process because someone doesn’t have the right (or any) documentation. People who are poor, formerly incarcerated, or Black are less likely to have a valid ID. The document verification process must provide a clear way for an applicant to get in touch with an agency representative to find an alternative method to verify they are who they say they are.

4. Rethink how risk is determined using IP addresses.

IP addresses are used to geolocate web traffic and often used to detect fraud in online unemployment insurance claims, as traffic coming from out of state can be suspicious. However, IP addresses are not particularly trustworthy as IP address databases can quickly become out of date and cell phone traffic can resolve to multiple locations at the same time, which appears suspicious if you don’t know how to look for it. Vulnerable populations, including those that are housing insecure, are more likely to use a computer borrowed from a friend or a library, or to use a cell phone — all of which means your location may change from week to week, and your apparent location may change even more. Rather than using IP address information as a yes/no fraud indicator, agencies should instead use it as part of a larger risk score.

U.S. Digital Response has published additional recommendations for addressing race and inequity in identity fraud detection, including how to increase accountability for ensuring rightful claimants make it through the system.

Governments have expanded access to public benefits for millions of Americans in response to COVID-19, but that expanded access has also scaled the challenges governments face to determine the identity and eligibility of applicants. USDR can help streamline processes for applicants and case workers to deliver benefits quickly to those who need it most. If you are a government seeking assistance with unemployment insurance modernization, please connect with our team by filling out this brief intake form.