2023 Shortlist

Unhoused and Undercounted

Entry type: Single project

Country/area: United States

Publishing organisation: Center for Public Integrity in partnership with The Seattle Times, Street Sense Media and WAMU/DCist

Organisation size: Big

Publication date: 2022-11-15

Language: English

Authors: Reporters: Amy DiPierro and Corey Mitchell, Center for Public Integrity; Greg Kim and Anna Patrick, The Seattle Times; Amanda Michelle Gomez, WAMU/DCist; Kaela Roeder, Street Sense Media

Editors: Jamie Smith Hopkins and Jennifer LaFleur, Center for Public Integrity; Molly Harbarger, The Seattle Times; Eric Falquero, WAMU/DCist; Will Schick, Street Sense Media

Audience and partnership coordination: Lisa Yanick Litwiller, Janeen Jones, Ashley Clarke, Vanessa Lee and Charlie Hsing-Chuan Dodge, Center for Public Integrity

Fact-checking, data-checking, graphics: Peter Newbatt Smith, Joe Yerardi, Janelle O’Dea and Pratheek Rebala, Center for Public Integrity


Bios of the reporters:
Amy DiPierro is a data journalist at the Center for Public Integrity and previously reported for The Desert Sun in California.

Corey Mitchell, a Public Integrity senior reporter who covers education, was an associate editor at Education Week.

Greg Kim and Anna Patrick, both reporters at The Seattle Times, work on the Project Homeless team.

Amanda Michelle Gomez is a WAMU/DCist reporter who previously reported for Kaiser Health News and Washington City Paper.

Kaela Roeder, deputy editor of Street Sense Media, has also written for publications such as the Washington Blade and NBC Out.

Project description:

Many public schools across the country fail to provide required support to homeless students, decreasing the odds of reaching graduation — and increasing the risk of long-term housing instability. Our analysis suggests that roughly 300,000 students are falling through the cracks of a system meant to help. As one district found, the right support makes a big difference in graduation rates. But federal funding for this work is so low that it doesn’t come close to covering the true cost.

Impact reached:

Our investigation, launched in November, had an immediate impact.

U.S. Rep. Emanuel Cleaver of Missouri, who chairs the House subcommittee that oversees efforts on homelessness, [vowed to discuss undercounting with federal officials](https://publicintegrity.org/education/unhoused-and-undercounted/government-helping-homeless-students-mckinney-vento-funding/). Staff for U.S. Rep. Don Bacon of Nebraska contacted superintendents in his district for answers after Public Integrity’s analysis showed likely undercounts there. And in December, Congress increased funding for homeless-student support by 13%.

A robust reporting collaboration also put information about the little-known and often-misunderstood requirements to serve homeless students in the hands of more people.

The Center for Public Integrity partnered with The Seattle Times, Street Sense Media and WAMU/DCist, each newsroom producing stories that made use of Public Integrity’s data analysis.

We also made the data and other resources available to additional newsrooms pre-publication, part of an innovative national-local collaboration that included a detailed reporting kit and Zoom office hours for reporting help.

All told, nine newsrooms produced stories with the data.

The pieces were published in November and December. The potential for more impact is high.

Both in interviews and the later follow-up by staff for a member of Congress, some school officials responsible for serving homeless students showed confusion about who should be identified and helped. The federal definition includes children doubled-up with friends or extended family out of economic need after losing their home, for instance. Our stories laid out who qualifies and what’s required.

The December funding increase is a start, but Congress could increase identification and improve support by putting more money toward the effort.

Meanwhile, enforcement of the law is weak. The federal and state departments of education could use a benchmark like ours to flag districts likely undercounting and investigate further. Michigan has guidelines for that.

Techniques/technologies used:

We used federal education and census data, with acquisition and analysis in the R programming language. We used functions from packages that can be found on https://cran.r-project.org:

* tidycensus
* tidyverse
* educationdata
* broom
* DescTools
* optparse
* assertthat

Our analysis had to overcome a variety of difficulties. Districts that count zero homeless students are permitted to not report that number to the federal homeless enrollment dataset; they simply don’t appear in it. We created a dataset that adds the missing districts as long as they had actively enrolled students in a given school year.

We then had to gauge the consistency of homeless enrollment data. Education researchers and state departments of education often use eligibility for the National School Lunch Program as a benchmark: The state of Florida, for example, advises school districts that they should “consistently identify at least 5% of their [free- or reduced-price lunch] enrollment as homeless in a school year.” We derived this benchmark at the school district level using federal data and calculated correlations between the share of students who are homeless and the share of students who are eligible for free- or reduced-price lunch.

To address possible criticism of the limitation of using lunch eligibility as a proxy for poverty, we calculated correlations (using Spearman’s rank correlation coefficient) to evaluate the relationship between homelessness and two other poverty variables. In addition, we calculated separate simple linear regressions in each state and school year.

In nearly half of states, tallies of student homelessness bear no relationship with poverty, a sign of just how inconsistent the identification of kids with unstable housing can be.

Context about the project:

This investigation breaks new ground: Despite reporting about undercounts of homeless students over the years, no media outlet has tried to quantify the gap nationwide. It was quickly apparent why.

Quantifying a lack of data – the serious undercounting of homeless students, blocking children and youth from needed help – was possible in this case. But it’s challenging, requiring a deep understanding of the context of what or who is being counted and the process of the count itself.

What it took for us: months of interviews and re-interviews, a lot of internal conversations about methods, then writing a white paper long before publication so experts could review and offer feedback. (We made the [white paper](https://drive.google.com/file/d/141hZ3a6gYtBCT5IFk_dzkUklVw2zIcXr/view) public.)

Understanding how schools try to identify which of their students are homeless, as well as what support they provide to those they identified, should have been less challenging than the data – at least for gathering anecdotes. But this, too, was hard.

Rarely were school officials eager to discuss this, even though we didn’t only reach out to districts whose numbers suggested they were doing a poor job at step one. Reporters persisted, both by increasing the number of districts contacted and by not giving up on ones we thought really needed to explain themselves.

This is also where national-local newsroom partnerships shine. At Public Integrity, we had the benefit of time to work through the kinks of the data and keep calling districts; the two reporters from our newsroom worked full time on the project for months. The Seattle Times, Street Sense and WAMU/DCist had the benefit of deep expertise in the context of homelessness in their regions, deep local sourcing and communities who know their reporting well.

What can other journalists learn from this project?

Several of the lessons:

* Looking at the way homelessness intersects with something unrelated to housing, such as education, is a powerful way to see how the country ensures (or does not ensure) bedrock concepts like equal access.

* All datasets are missing information. You can potentially piece datasets together to fill in the blanks or create benchmarks to quantify the size of the gap. Additional detail regarding our analysis (including an extended discussion of our assumptions, statistical methods and results) is described in our white paper, linked above.

* Collaborative projects are more powerful than working alone. Reporters and other journalists at each of the partnering newsrooms brought ideas, sources and questions that made the project stronger.

* Housing insecurity is both traumatic and heavily stigmatized. Be especially mindful of that when interviewing sources who are currently experiencing or have experienced homelessness. Fully explain your story and aims. Affirmatively discuss with your source what they are comfortable having printed or aired. If you’re talking to a family, remember that some members might be OK with you naming them and others might not – or, for instance, a parent might be OK being named but not their child or children. Our investigation included a parent who decided, after mulling over it, that it was best not to name her son or show photos of him that showed his entire face. We made sure to accommodate that and – because this came up after the photos were taken – met with our design editor to figure out which should not be used.

Project links: