Blog: Five Questions Housing Analysts Should Ask About "By-Name Data"

Built for Zero (BFZ) is a data-driven allocation scheme that is spreading across the Anglo world. The administrative staff of BFZ communities— a local network of service providers (i.e., homeless system) who collaboratively use BFZ to disseminate housing assistance to homeless households—use “by-name data” to help administrators monitor and manipulate system flow.

By-name data is reliable, real-time, population-level data with personalised information about each service recipient. BFZ advocates argue by-name data is a resource that administrators need to monitor and manipulate entries into (inflow) and exits from (outflow) their homeless system. The ability to “see” system flow is said to help administrators identify and fix bottlenecks that either facilitate or prolong homelessness.

Although BFZ has spread from the USA to homeless systems in Britain, Canada, and Australia over the past ten years, housing scholars have just started to analyse its implementation and impact. Until recently, none of that research had examined the way administrators generate by-name data. That knowledge gap has impeded efforts to evaluate BFZ performance and inform policymaking.

I recently published an article “Cookin’ the Books: How Waitlist Management Biases Data Production in Built for Zero Communities” in the International Journal on Homelessness to fill that void by analysing the way local administrators from 28 BFZ communities in the USA produce by-name data. This essay advances that study by identifying five questions that housing analysts should answer whilst evaluating BFZ performance and advising policymakers.

1.        How does a BFZ community define “active status?”

Active status is the primary metric that BFZ communities use to manage their housing prioritisation list. That metric is a temporal criteria that specifies the length of time since someone last contacted a participating agencies within the BFZ community. Once someone crosses that threshold, they are no longer considered homeless and thus removed from the by-name dataset.

BFZ communities vary in the temporal limit that they place on active status. My research showed this time limit can vary from 1-week to 24 months. If someone reconnected with an agency in the BFZ community after they were labelled inactive, then the administrator readded them to waitlist. However, the extent to which administrators documented uncounted time on the street as homelessness was constrained by federal policy.

A narrow definition of active status can let a BFZ community quickly declare an end to homelessness and vice versa. This creates a problem for analysts that are trying to evaluate BFZ performance because a low or high count might reflect the data policy of a BFZ community rather than a reduction in or increase of homelessness. Analysts should therefore determine how a BFZ community defines active status whilst interpreting its data and/or evaluating its performance.

2.        How do local administrators get information about unsheltered homelessness?

When someone gets estranged from participating agencies in a BFZ community, administrative staff try to verify their active status before removing them from the waitlist. The resources that administrators have to do verify estranged clients influence whether they can determine someone’s active status and vary across BFZ communities.

The US Department of Housing and Urban Development estimates 39.2% of people who experienced homelessness in 2023 were unsheltered. Whereas emergency shelter providers generate daily counts of residents, street outreach workers have to find people experiencing unsheltered homelessness on the street. This task is complicated by social forces that push homeless individuals across urban spaces and made (nearly) impossible by austerity budgets that have defunded street outreach teams.

With inadequate support from street outreach, the administrators of a BFZ community struggle to access by-name data about roughly 40% of their homeless population. A gap like that in the by-name dataset would produce counts that underestimate the scale of homelessness in a BFZ community. Analysts should therefore consider the efficacy of a BFZ community’s outreach system whilst evaluating its performance.  

3.        What multi-sector partnerships have administrators formed?

Participating agencies in a BFZ community may lack information about the status of estranged clients. In those situations, administrators often contact external agencies like hospitals, jails, and charities that regularly engage people experiencing homelessness. Multi-sector partnerships are therefore vital to by-name data production.

My recent article shows BFZ communities vary in the breadth and depth of multi-sector partnerships that administrators have to verify estranged clients. Weak multi-sector ties hinder administrators from accessing information about those clients and vice versa. This structural deficit may get expressed in the by-name data as lower counts if system leaders narrowly defined active status. Alternatively, that deficit may inflate by-name data if system leaders broadly define active status by including service recipients who are no longer homeless.

Analysts should consequently request information about the multi-sector partnerships that a BFZ community has and way that administrators use those relationships to verify active status. With that information, analysts can judge the accuracy of by-name data that is reported by a BFZ community.

4.        What data quality issues do administrators confront?

Frontline workers collaboratively produce by-name data in a shared database that includes information uploaded by emergency shelter providers and street outreach workers. The leadership of BFZ communities establish data protocols to ensure “standardisation.” However, the administrators who participated in my study said frontline workers often violated these protocols, either intentionally or unintentionally.

Data errors could cause someone to be erroneously labelled inactive. If multiple agencies and frontline workers in a BFZ community create data errors, then the population counts for that system will be inaccurate. Administrators tried to fix this problem by cleaning their data, but this task could take a lot of time if frontline workers throughout the system habitually violated data protocols.

Extensive data cleaning reduced the contemporaneousness of by-name data because it delayed the accuracy of records. It is therefore important for analysts to consider the data protocols, compliance practices, and quality assurance of BFZ communities whilst interpreting claims that its by-name data is “real-time.”

5.        How do administrators exercise discretion whilst producing by-name data?

Lastly, administrators exercised bounded discretion whilst producing by-name data. My study showed limited oversight of their daily workflow let administrators bend formal polices like active status that were adopted by institutional authorities.

The tolerance of discretion impacted by-name data because administrators could choose which estranged clients they tried to verify and how they went about doing so. If an administrator was biased against a particular individual or group, then they might not put a lot of effort into verify their status once they become estranged and vice versa. Furthermore, administrators may feel political pressure from local authorities if homelessness in their community rises up on the public agenda. An expectation by elected officials that administrators “produce results or else” could motivate the latter to deflate homeless counts by shirking active status verification.

It is therefore imperative that analysts consider the subjective and contextual factors that influence the way administrators (mis)use their discretion whilst producing by-name data.

Conclusion

A general finding from my recent article is by-name data is socially constructed within nested contexts that constrain and enable its production. The variability of these contexts across time and place limits the comparability of by-name data both within and between BFZ communities. This complexity makes it difficult for analysts to interpret by-name data, evaluate BFZ performance, and give policymakers sound advice about the merits of BFZ.

As BFZ spreads around the Anglo world, this essay sheds light on the black box that is by-name data production. To enhance transparency, I recommend the administrators in BFZ communities publicly release information on each of the topics that were discussed above. This will help analysts evaluate BFZ performance and institutional authorities decide if their homeless system should adopt this methodology.