Boil water alerts and their impact on the unexcused absence rate in public schools in Jackson, Mississippi

The study investigates whether and how boil water alerts (BWAs) in Jackson, Mississippi, disrupt student learning by increasing daily unexcused absence rates in public schools. Against the backdrop of widespread drinking water quality violations in the United States and environmental justice concerns disproportionately affecting low-income and non-white communities, Jackson represents a critical case study. Longstanding infrastructure disinvestment, demographic shifts, and repeated EPA enforcement actions have produced frequent BWAs. The authors focus on Jackson Public School District (JPSD) students and ask: do localized and city-wide BWAs increase unexcused absences, and to what extent after accounting for community-level socioeconomic vulnerabilities and COVID-19-related changes in instructional delivery?

Prior work documents widespread violations of the Safe Drinking Water Act and Clean Water Act across the United States, affecting tens to hundreds of millions of residents. Contaminated drinking water is linked to adverse health outcomes in adults and children, including cancers, gastrointestinal illness, and cognitive decline. Environmental justice research shows that low-income and non-white communities are more likely to experience unsafe water. Studies on educational outcomes indicate environmental exposures (e.g., temperature, air pollution) can increase school absenteeism, with both immediate and lagged effects, and that chronic absenteeism is associated with poor academic performance, elevated risk of dropout, social and behavioral issues, reduced adult economic wellbeing, and decreased life expectancy. This study extends the literature by examining BWAs as an environmental stressor influencing unexcused school absences, while controlling for neighborhood-level socioeconomic vulnerabilities.

Design and setting: Observational study of Jackson, Mississippi, linking BWA events to JPSD daily unexcused absence rates from 2015–2021, with adjustments for neighborhood-level vulnerabilities and COVID-19 instructional delivery changes. Data sources: (1) BWA data (2015–2021) obtained via FOIA from the City of Jackson’s Water/Sewer Business Administration Office; 2,547 notices total, including nine city-wide alerts. Street-level and address-range alerts were matched to residential parcels using Hinds County 2021 tax assessor parcel data and OpenStreetMap (OSM) building footprints. Of the original set, 2,146 alerts were matched to parcel addresses; 401 (15.7%) were excluded, yielding 191,748 address-specific BWAs plus 360,983 address notices for the nine city-wide BWAs. Population per address was estimated using dasymetric mapping of 2010 census block counts proportional to building footprint area. (2) JPSD daily attendance data for 52 schools (31 elementary, 10 middle, 7 high; four special program schools excluded) from Jan 2015 to Oct 2021. Unexcused absence rate defined as daily unexcused absences divided by daily enrollment. (3) Instructional delivery during COVID-19: periods of virtual (all schools: 16 Mar–22 May 2020; 10 Aug–18 Dec 2020) and from 19 Jan–26 May 2021 onsite or virtual for elementary and hybrid or virtual for middle/high. (4) School zoning: JPSD-provided street lists and ZIP codes were parsed and standardized to match census road shapefiles (via R packages tigris and postmastr); 98% of listed streets matched; unmatched or multiple assignments were resolved by nearest school. BWA parcels were matched to zoned elementary, middle, and high schools using the same approach; 95% matched on street and ZIP code. (5) Community-level vulnerability variables from ACS 5-year estimates (2016–2020) at the census block group level: poverty, race/ethnicity (percent identifying as Black, Asian, Native American, Hispanic, or two or more races), single-parent households, elderly (65+), young (<18), limited English proficiency, computer/internet access, disability, and elderly living alone (social isolation). A composite vulnerability score was computed by summing the percentages of these variables per block group. Time periods and modeling: The analysis used generalized additive models (GAMs) with identity link and Gaussian distribution, modeling daily unexcused absence rates with a non-parametric smooth function of time and covariates. Two time periods were analyzed: Period 1 (5 Aug 2015–31 Dec 2019) and Period 2 (1 Jan 2020–26 May 2021) to account for increased BWA exposure in 2020–2021 and COVID-19 instructional changes. Covariates (standardized unless time/dummy variables) included BWA indicator, city-wide BWA indicator, COVID-19 instructional mode dummies (Period 2 only), and the vulnerability variables. Lag effects: Because duration data were available only for 2015 (mean ≈2 days), three lag definitions were considered: (1) notice date only; (2) notice date + 1 day; (3) notice date + 2 days. For each model, the lag window was selected using BIC; the BWA indicator equals 1 if the attendance date falls within the selected window and 0 otherwise. Modeling was implemented in R using mgcv (v1.8-40). Analyses were conducted for all schools combined, by school level (elementary, middle, high), and for each individual school. Sensitivity analyses examined effects when both households and schools were under a BWA.

• Exposure and vulnerability context: 2,547 BWA notices (2015–2021), including nine city-wide alerts. The number of households impacted by BWAs was highest in 2020 (80,099), followed by 2018 (58,443) and 2021 (22,679). Total BWA vulnerability peaked in 2021 (mean 2.41, s.d. 0.35). Most parcels received at least two BWAs during the study (54.42%), with the highest intensities concentrated in West Jackson; 34.48% of parcels were at high BWA vulnerability. About 13% of schools were close to parcels with the highest BWA counts and 35% close to parcels with highest vulnerability.
• All schools (GAM results): Period 1 showed a statistically significant increase in unexcused absence rate associated with localized BWAs of about 1 percentage point (estimate 0.0099; p<2×10⁻¹⁰). City-wide BWA effects were positive but not statistically significant (estimate 0.0034; p=0.6211). Period 2 results were similar for localized BWAs (estimate 0.0096; p=0.0018) and again non-significant for city-wide alerts. Selected lag up to 2 days indicates effects can persist for as long as two days after an alert.
• School-level analyses: No statistically significant BWA effects by school type (elementary, middle, high) when aggregated by level. However, several individual schools showed significant associations. Elementary: Bates (pre-pandemic, +6% unexcused rate; ≈23 students per alert), Boyd (post-pandemic, +4%; ≈14 students), Key (post-pandemic, +6%; ≈21 students), Spann (post-pandemic, +1%; ≈5 students). Clausell showed a decrease (pre-pandemic, −7%). Middle schools showed no significant associations, though trends were positive in Period 2. High schools: Lanier showed a significant positive association in Period 2. Effects could last up to 2 days post-alert based on lag analysis.
• Covariate associations (Table 2): Higher percentages of individuals identifying as Black, Asian, Native American, Hispanic, or two or more races and higher poverty were associated with increased unexcused absence rates in Period 1. Several vulnerability covariates showed significant associations in Period 2 (e.g., race/ethnicity positive; limited English and elderly negative; disability positive), indicating complex socioeconomic relationships with attendance.
• Magnitude of impact: Authors estimate BWAs contribute to approximately 1–10% increases in unexcused absence rates in JPSD, affecting between roughly 210 and 2,100 students per alert. In 2020, with 454 alerts issued, this translates to an estimated 95,000 to 9.5 million individual student-days of school lost. Protective effects were observed in some contexts (e.g., Clausell Elementary, where all students receive free/reduced lunch).

BWAs function as acute, unpredictable environmental stressors that complicate morning routines and daily living, particularly for households with children, thereby increasing the likelihood of unexcused school absences. By controlling for neighborhood-level socioeconomic vulnerabilities and COVID-19 instructional delivery changes, the models isolate the association between BWAs and attendance. The results show that localized BWAs significantly increase unexcused absences across the district, with persistent effects up to 2 days, and that impacts concentrate in specific elementary and high schools serving predominantly non-white student bodies and communities with higher poverty. City-wide BWAs did not show significant district-wide effects, possibly reflecting differences in communication, preparedness, or other unobserved factors. These findings align with literature documenting environmental conditions elevating absenteeism and suggest that recurrent BWAs add to cumulative educational disadvantages, potentially exacerbating long-term negative outcomes such as dropout and reduced economic wellbeing. The study underscores the environmental justice dimension: communities most affected by BWAs are those already burdened by socioeconomic vulnerabilities.

This study provides evidence that frequent boil water alerts in Jackson, Mississippi, are associated with significant increases in unexcused school absences, with effects persisting up to two days and concentrated in specific schools and vulnerable communities. The work advances understanding of how water infrastructure failures affect educational outcomes and highlights the urgency of addressing root causes of poor water quality. Policy and practice recommendations include improving BWA communication (clear, consistent, accessible notices; targeted outreach), maintaining reliable public information systems, and advancing infrastructure remedies. Future research should incorporate individual- and household-level data (e.g., income, parental employment and education), assess additional educational outcomes (grades, test scores, reading comprehension), expand to surrounding communities for comparative analyses, and evaluate broader community impacts (economic, health, psychological). The authors are developing community-engaged tools and data repositories to enhance environmental health literacy and accountability.

• Population exposure estimation: The number of people per address under BWAs was estimated using dasymetric mapping (2010 census blocks scaled by OSM building footprints), which may misestimate true affected populations.
• Confounding domains: Only community-level (neighborhood) vulnerabilities were included; individual and family factors influencing absenteeism were not modeled.
• Outcomes scope: The analysis focused on unexcused absences; other educational outcomes (e.g., grades, test scores, reading) were not assessed.
• Geography: The focus was limited to Jackson; nearby communities on the same water system were not analyzed, limiting external comparisons.
• Data linkage and coverage: Not all BWAs could be matched to parcels (15.7% unmatched and excluded). School zoning and parcel matching, while high, were not perfect (98% street match; 95% parcel-school match).
• BWA duration: Explicit duration data existed only for 2015 (mean ≈2 days); duration and lag assumptions for other years may introduce error.
• Pandemic period complexities: COVID-19-related changes in instruction were modeled but may not capture all dynamics affecting attendance.